Download as pdf or txt
Download as pdf or txt
You are on page 1of 71

DEGREE PROJECT IN ELECTRICAL ENGINEERING,

SECOND CYCLE, 30 CREDITS


STOCKHOLM, SWEDEN 2020

Case Study on Applications of


Measurement Equipment for
Dynamic Line Rating

LINUS DAHLGREN

KTH ROYAL INSTITUTE OF TECHNOLOGY


SCHOOL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE
Examiner: Patrik Hilber (KTH)

Supervisors: Kateryna Morozovska (KTH), Erik Lejerskog (Ellevio AB)

School of Electrical Engineering and Computer Science (EECS)

KTH Royal Institute of Technology

1
Abstract
Dynamic Line Rating is the concept of rating the ampacity of power lines depending on the thermo-
dynamic conditions of the lines. It is an alternative to Static Line Ratings, which are set to maintain
network safety requirements under unfavourable operating conditions. Thus Dynamic Line Rating allows
the network operator to utilise unused capacity of the lines, and operate network assets closer to their
limits while still maintaining network safety requirements.

This thesis explores the potential of Dynamic Line Rating through an analysis of a pilot project from
Ellevio AB and Heimdall Power. The goal is to ascertain the ideal placement of monitoring equipment for
Dynamic Line Rating purposes, determine an estimate of the transmission capacity increase, and study
the economic impact of Dynamic Line Rating implementation.

Data from the measurement equipment used in the pilot project has been analysed to ensure that it is
operating correctly. This analysis showed that the line temperature measurements were malfunctioning.
Because of this, the analysis of Dynamic Line Rating in this thesis is strictly weather based.

The most critical spans along a line have been identified through a method that has been developed for
this thesis. The capacity increase is estimated to 23% when Dynamic Line Rating is implemented. An
economic analysis quantified the potential value of this capacity increase from both the perspective of a
wind power producer and a DSO. This analysis also showed that DLR show a correlation towards wind
power generation and no correlation towards the general load profile of the line.

2
Sammanfattning
Dynamisk gradering är ett alternativt sätt att gradera utrustning i elkraftsystem (luftledningar, transfor-
matorer). Den rådande normen är en statisk gradering som är satt för att säkerställa att systemet håller
sig inom säkerhetsgränserna även vid ofördelaktiga förhållanden. En dynamisk gradering beräknar last-
barheten kontinuerligt utifrån realtids eller prognosticerade mätvärden. Detta tillåter drift närmare de
faktiska säkerhetsgränserna, vilket ökar utrustningens överföringskapacitet. Det är således ett attraktivt
forskningsområde inom elkraftmarknaden, då utbyggnad av överföringskapacitet i nuläget ofta är förenat
med stora kostnader och långa handläggningstider.

I Oktober 2019 påbörjade Ellevio AB tillsammans med Heimdall Power ett pilotprojekt om dynamisk
lastbarhet. Mätutrustning utvecklad av Heimdall Power monterades på en regionalledning, för att ut-
forska potentialen hos både mätutrustningen och den dynamiska lastbarheten för ledningen i fråga. Det
här examensarbetet har som en del i pilotprojektet haft som mål att besvara frågor gällande den po-
tentiella ökningen i överföringskapacitet och de ekonomiska konsekvenserna av detta, samt analysera
mätutrustningens tillförlitlighet och optimala placering längs ledningen.

En metod för att finna de mest kritiska punkterna längs ledningen har tagits fram i samband med
examensarbetet. Metoden är baserad på en litteraturstudie av tidigare metoder, och använder tre faktorer
för att göra en riskskattning av varje span på ledningen. Metoden applicerades sedan på ledningen i
pilotprojektet för att finna de mest kritiska punkterna för att placera mätutrustning.

Analysen av mätutrustningens tillförlitlighet visade att temperaturmätningarna hos Heimdalls neuroner


inte fungerade korrekt. Detta har senare bekräftats genom tester från Heimdall.

Den potentiella överföringskapaciteten ökade i genomsnitt med 23% då dynamisk gradering implementer-
ades. En positiv korrelation mellan den dynamiska graderingen och lastprofilen från vindkraftsproduktion
i området har också påvisats. Den ekonomiska analysen visade att det potentiella ekonomiska värdet från
denna ökning beror på mängden nyetablering av vindkraft, och att värdet från implementering av dy-
namisk gradering då som mest är 4.1 MSEK.

Sammanfattningsvis har det här examensarbetet påvisat både möjligheter och utmaningar kring dynamisk
gradering. Vidare behöver ovanstående analyser göras ånyo då mer mätdata samlats in, för att kunna
bekräfta eller förkasta slutsatserna i rapporten.

3
Acknowledgements
I would like to extend my gratitude to my supervisor Kateryna Morozovska, not only for the support
throughout the project, but also for the help with choosing a topic and enabling business contacts for
the thesis.

I would also like to thank Erik Lejerskog for the opportunity to work on an interesting project, and for
the close feedback and fruitful discussions that helped shape this thesis.

My gratitude also goes out to the collegues at Ellevio and Heimdall Power, for the patience, support and
expertise throughout the project.

Finally, I would also like to extend a heartful thanks to Bo Flodin, for encouraging me and always keeping
my spirits high.

4
Contents
1 Introduction 13

2 Purpose and structure of the report 14

3 Previous work 15

4 Dynamic Line Rating 18


4.1 Joule heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Convective cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Solar heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.4 Radiative cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.5 Thermal calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.5.1 Steady state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.5.2 Dynamic state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.6 Sag theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.7 Types of Dynamic Line Rating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5 Pilot study 28
5.1 Test area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 DLR Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3 DLR Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3.1 Solar heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3.2 Radiative cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.3.3 Convective cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.3.4 Dynamic Line Rating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

6 Validation of measurement devices 33


6.1 Weather station . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.1.1 Temperature measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.1.2 Wind measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.1.3 Solar irradiance measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2 Heimdall Neuron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.2.1 Line current measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.2.2 Line temperature measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2.3 Angle of inclination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

7 Economic analysis 39
7.1 Power transmission calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7.2 Results from windpower script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.3 Net Present Value analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.3.1 Input Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.3.2 Analysis method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.3.3 NPV analysis results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

8 Identification of critical spans 46


8.1 Method for finding critical spans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
8.2 Proposed methodology for identifying critical spans . . . . . . . . . . . . . . . . . . . . . . 47
8.2.1 Macroclimatic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
8.2.2 Microclimatic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
8.2.3 Distance-to-ground analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
8.2.4 General guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
8.3 Application of methodology on VL-X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

9 Discussion 52

10 Conclusions 55

11 Future work 56

5
A Appendix 57
A.1 Main script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
A.2 IEEE thermal calculation script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
A.3 Cigre thermal calculation script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
A.4 Wind Power Injection Script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
A.5 Risk factor calculation script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

6
List of Figures
1 Comparison of thermal conducitivity calculated by the different standards . . . . . . . . . 20
2 Comparison of the IEEE and Cigre calculation methods for dynamic viscosity of air . . . 21
3 Inclined catenary span [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4 Solar heating thermal component from IEEE-738 and Cigre-601 calculation methods . . . 30
5 Radiative cooling thermal component from IEEE-738 and Cigre-601 calculation methods . 30
6 Convective cooling component from IEEE-738 and Cigre-601 calculation methods . . . . . 31
7 Dynamic Line Rating from IEEE-738 and Cigre-601 calculation methods for first week of
November . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
8 Dynamic Line Rating for the entire measurement period with IEEE-738 and Cigre-601
calculation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
9 Ambient temperature from weather station and control measurements in Arvika and Karlstad 33
10 Wind rose plots for November . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
11 Wind rose plots for January . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
12 Global radiation measured in weather station and Karlstad for November . . . . . . . . . 35
13 Global radiation measured by weather station, Karlstad and estimated by IEEE-738 for
November . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
14 Line current measurements from all neurons and a nearby substation . . . . . . . . . . . . 36
15 Measured and estimated temperature, and current 6/11 . . . . . . . . . . . . . . . . . . . 37
16 Measured and estimated temperature for November 2019 . . . . . . . . . . . . . . . . . . 37
17 Load on VL-X at 230MW of added windpower . . . . . . . . . . . . . . . . . . . . . . . . 40
18 Relative SE and relative transmitted power from windpower indices . . . . . . . . . . . . 42
19 ROI as a function of windpower for DLR and SLR . . . . . . . . . . . . . . . . . . . . . . 45
20 Residual NPV as a function of windpower for DLR and SLR . . . . . . . . . . . . . . . . 45
21 Orto-photo of VL-X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
22 Estimation of total risk along the entire line . . . . . . . . . . . . . . . . . . . . . . . . . . 51
23 The most vulnerable span along the line . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

7
List of Tables
1 List of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Greek symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Coefficients for Natural Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Forced convection coefficients in Cigre-601 method . . . . . . . . . . . . . . . . . . . . . . 22
5 Checklist table for which monitoring systems support which DLR methods . . . . . . . . 27
6 Data for VL-X conductor of type Ibis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
7 Input variables for DLR calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
8 ENS, ET and SE for different levels of added windpower . . . . . . . . . . . . . . . . . . . 41
9 Correlation between ampacity from DLR and load profiles . . . . . . . . . . . . . . . . . . 41
10 Parameters of the NPV analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
11 Distribution of investment costs for onshore wind farms . . . . . . . . . . . . . . . . . . . 43
12 Total estimated risk for each span along the line . . . . . . . . . . . . . . . . . . . . . . . 51

8
Nomenclature
List of abbreviations
ACSR Aluminium Conductor Steel Reinforced
ANM Active Netowrk Management
CFD Computational Fluid Dynamics
CIGRE Conseil International des Grands Réseaux Électriques
CSM Conductor Sag Measurement
CTE Conductor Temperature Evaluation
CTGM Clearance-to-ground Measurement
DLR Dynamic Line Rating
DSO Distribution System Operator
DTOC Delayed Time Over-current
EPE Experimental Plastic Elongation
EPRI Electric Power Research Institute
ERCOT Electric Reliability Council of Texas
FSM Full-scale Monitoring
GE General Electric Company
IEA-ETSAP Institute of Electrical and Electronics Engineers
IEEE International Energy Agency - Energy Technology Systems Analysis Program
IRENA International Renewable Energy Agency
LE Linear Elongation
MEMS Micro-electric-mechanical System
MESAN Mesoscale Analysis
NPV Net Present Value
NYPA New York Power Authority
OHL Overhead Line
ROI Return of Investment
RTTR Real Time Thermal Rating
SAW Surface Acoustic Wave
SER Seasonal Line Rating
SGDP Smart Grid Demonstration Program
SLR Static Line Rating
SMHI Sveriges Meteorologiska och Hydrologiska Institut
SPE Simplified Plastic Elongation
TM Tension Monitoring
TSO Transmission System Operator
USi Underground Systems, Inc
WF Weather Forecast Rating
WM Weather Model Rating
WRF Weather Research and Forecasting

9
List of variables
Symbol Description Unit
CN P V Net Present Value from Economic Analysis SEK
Cres Residual Net Present Value SEK
c(T ) Specific heat capacity at temperature T J/kg · K
D Conductor diameter m
d Strand diameter m
dtglimit Distance to ground limit Dimensionless
dtg50◦ Distance to ground when Ts is 50◦ C Dimensionless
EN S ENS reliability index M W/year
ET Transmitted power index M W/year
ET% Relative transmitted power index %
Gr Grashof number Dimensionless
H Horizontal component of tension N
He Elevation above sea level m
I Current A
+
I Additional load from wind power generation kA
IT Global solar radiation intensity W/m2
Ksolar Elevation adjustment factor for solar heat flux Dimensionless
ksk Skin effect factor Dimensionless
m Mass of unit length of conductor kg/m
N Added windpower Dimensionless
Nu Nusselt number Dimensionless
N u90 Nusselt number for perpendicular wind flow Dimensionless
N uδ Nusselt number adjusted for wind angle Dimensionless
P Produced windpower MW
PI Corona heating W/m
PJ Joule heating W/m
PM Magnetic heating W/m
PS Solar heating W/m
Pr Prandtl number Dimensionless
qc Convective cooling W/m
qc1 Forced convective cooling low wind component W/m
qc2 Forced convective cooling high wind component W/m
qcn Natural convection component W/m
qr Radiative cooling W/m
qw Evaporative cooling W/m
Qse Estimated solar heat flux, adjusted for elevation W/m2
R(T ) Resistance at temperature T Ω
R(Thigh ) Tabulated resistance at Thigh Ω

10
R(Tlow Tabulated resistance at Tlow Ω
Rdc DC-resistance Ω
Rs Conductor surface roughness Rs = d/[2 · (D − d)] Dimensionless
ROI Return of investment from Economic Analysis %
ROI ∗ Desired return of investment in Economic Analysis %
riskdtg Unweighed distance-to-ground risk factor Dimensionless
riskmicro Microclimate risk factor Dimensionless
risktotal Total risk factor Dimensionless
S Span length m
SE Sold electricity index SEK/year

T Temperature C

Thigh High measurement temperature C

Tlow Low measurement temperature C

Ta Ambient temperature C

Tf Film temperature around conductor C

Ts Conductor surface temperature C
VV L−X Operating voltage of VL-X kV
w Conductor weight per unit length kg/m
Y Years in operation Dimensionless
Zi Rate of inflation %
Zr Real interest rate %
Zn Nominal interest rate %

Table 1: List of variables

11
Greek symbols
Symbol Description Unit
α Solar absorptivity Dimensionless
α20 Linear temperature coefficient of resistance K −1
β20 Linear temperature coefficient of specific heat capacity K −1
δ Angle of wind incidence dimensionless
η Percentage of wind park production through VL-X %
ε Emissivity Dimensionless
ΓIC Investment cost SEK
γT C Turbine cost SEK
γCC Construction cost SEK
γOC Other cost SEK
Λ Maintenance cost SEK
λf Thermal conductivity of air W/m · K
µf Dynamic viscosity of air kg/m · s
ρf Air density kg/m3
σb Stefan-Boltzmann constant, 5.67 · 10−8 W/m2 · K 4
θ Solar angle of incidence dimensionless
Υ Measurement uncertainty for DTG risk calculations m
ζ20 Quadratic temperature coefficient of resistance K −2

Table 2: Greek symbols

12
1 Introduction
As generation and consumption patterns are changing, power system owners are facing new challenges
that need to be met to ensure a reliable and cost effective electric power supply [2]. In addition to a need
for improved capacity to meet the ever-increasing demands of the consumers, the nature of generation
is changing. Traditionally, power has been generated by sources with adjustable output (hydro, nuclear,
carbon, oil etc.) and the power grid has been built around these generation centers. As generation is
shifting towards renewable sources, which by nature are decentralized and intermittent [3], so does the
demands on the power system.

According to [4], load flow analysis often confirms that additional wind power cannot be connected at the
distribution level due to risk of overloading the existing lines. In these cases, costly grid infrastructure
investments will have to be undertaken to be able to connect wind farms.

However, new technology may allow for a different solution that would allow grid owners to postpone
or omit such major investments. Currently, the norm is to assign seasonal ampacity ratings to lines,
which are commonly referred to as Static Line Ratings. These are based on worst-case scenarios and
set to ensure that the line stays within its thermal limits. However, a different type of line rating called
Dynamic Line Rating could be implemented as an alternative to utilise previously unused capacity [5].
Dynamic rating is in addition to OHLs also an interesting area of study for other components in the
power system such as cables and transformers [2], but the major focus in this thesis will be about OHLs.

DLR uses real-time (or forecasted) measurements of different variables of interest to calculate to permis-
sible ampacity dynamically. This allows for a more accurate estimation of the grid transfer capacity. It
also increases the overall transfer capacity without having to upgrade existing lines, or build new lines
which are costly and time-consuming projects [6]. Furthermore due to the natural synergy between wind
power generation and heat dissipation from convective cooling, DLR helps increase the capacity for wind
energy integration. In addition to utilizing untapped transfer capacity, DLR also prevents premature
aging and dangerous sag-limit violations due to high conductor temperatures for when the conditions are
worse than the assumption for static rating [7]. However, the ethical advantage of facilitating renewable
generation must be weighed against the fact that a higher load on the lines yields higher losses.

Due to its potential economical and environmental benefits, DLR is identified as one of eight smart grid
metrics by the U.S. Department of Energy [8]. Both IEEE and CIGRE have developed standardized
methods and models for temperature calculations with varying weather conditions, a topic which will be
discussed later in the thesis [9],[10].

[8] argues that although some progress has been made regarding the theoretical foundation of DLR, the
key challenges are still centered on quantifying the benefits of DLR through pilot studies in operational
grids. Ellevio AB are performing such a study in collaboration with Heimdall Power in the south-west of
Sweden, and this thesis will largely be centered around quantifying the effect of DLR, and the behaviour
of the measuring equipment through said pilot study.

13
2 Purpose and structure of the report
The purpose of this report and master thesis as a whole is to investigate the potential and consequences
of implementing Dynamic Line Rating, through participation in and analysis of a case study by Ellevio
AB and Heimdall Power. This investigation will include verifying that the measurement devices used
for data collection are operating correctly. It will also include calculations of the Dynamic Line Rating
through values from measurements in the study, and a comparison of methods and standards used for
DLR calculations.

To analyse the consequences of DLR implementation, a comparison of how much additional wind power
can feasibly be connected to the grid with DLR will be made. The basis of this comparison will be an
economic analysis from the perspective of a grid owner, and a wind power producer in the area.

The report will also contain a literature study to ascertain the optimal placement of measurement equip-
ment, and an application of the methodology found through said literature study on the line in the pilot
study.

In conclusion, the purpose of the report is to answer the following questions:

• What is the increase in capacity from implementing DLR?

• What are the location of the critical spans along the line?

• What is the economic benefit from implementing DLR?

Section 3 contains a brief summary of previous work in the field. Section 4 contains a theoretical explana-
tion of the basic concepts of Dynamic Line Rating. Section 5 contains the methods for, and results from
the DLR modeling. Section 6 contains the results from the verification of the data from the measurement
devices used in the project, and Section 7 contains the method and results from the economic analysis of
DLR implementation. Section 8 contains the summary of a literature study regarding previous work on
the identification of critical spans. It also includes a proposed methodology for finding critical spans, and
the results from the application of said methodology on the line in the pilot project. Section 9 contains
a discussion of the results found in Section 5-8, Section 10 contains the conclusions that can be drawn
from the project, and Section 11 contains suggestions for future work that can build upon the work done
in this project.

14
3 Previous work
A new thermal rating approach: The real time thermal rating system for strategic overhead conductor
transmission lines by Murray W. Davis was published in 1977, and is widely considered as the first
mention of DLR in academia. There had been earlier considerations, but the development of SCADA
and improved sensor technology in the late 1970’s made DLR a real possibility [11].

In the late 1990’s, San Diego Gas & Electric installed a CAT-1 unit on a 230kV transmission line that
was a bottleneck for the system. The monitoring data was integrated into the EMS where it could be
used by operators in decision making [12].

Since 2008, E.ON and Central Networks UK uses a weather based DLR with automatic calculations based
on Cigre TB207 on a 132kV line from Skegness to Boston. Additionally, four Power Donuts are used
for direct monitoring of line temperature. The DLR system controls the wind farm generation, so that
if the line current reaches 95% of the dynamic rating, the wind farm output is curtailed. There is also
a secondary protection system in place, where a DTOC relay trips if the generation is not successfully
curtailed. Initially, E.ON expected the capacity to increase around 30% from the implementation of
DLR. The potential capacity increase proved even greater, but the ampacity was limited to ensure safe
operation of other components along the line. [13]. The temperature measurements from the Power
Donuts were used for validation of the weather based rating system. In a follow-up report from 2009, it
is shown that the theoretical temperature estimations from the weather data follow the actual conductor
temperature closely. Furthermore it is concluded that the conservative assumptions about wind direction
and solar radiation that was made in the project provided a good safety margin. It was concluded that
DLR is a cost efficient alternative to more drastic reinforcement procedures in order be able to connect
additional distributed generation to the grid[14].

In 2010, the Belgian TSO Elia implemented DLR on a 70kV network in the Ardennes region using devices
developed by Ampacimon. They chose to implement an ANM (Active Network Management) solution
which involved Elia retaining the right to curtail the wind farms’ output when necessary to guarantee
safe network operation. However, it was acknowledged that curtailment destroys the economic value of
wind power plants, and thus a solution was needed to minimize the necessary curtailment. Expanding the
existing grid in a reasonable timeframe was not an option due to regulations and budget concerns, which
lead to the option of implementing DLR to increase the capacity of the existing lines. The data output
from the Ampacimon DLR computation is used in the EMS to determine if curtailment is necessary. If
that is the case, the optimal power flow is calculated, and the wind farms receive new set points. This
process repeats itself every 15 minutes, which is motivated by the thermal inertia of the conductors. This
Dynamic Line Rating solution together with ANM resulted in a capacity increase of nearly 75%. Elia
concluded that this was a valid and quick solution to increase the capacity of a saturated network, with
positive effects for wind power integration [15].

In 2012, the German TSO Amprion GmbH performed a pilot study with two separate methods for DLR
estimation, based on direct temperature measurements and weather station measurements. The temper-
ature of the line was measured with two SAW units. It is mentioned in the technical report that the cost
of the pilot systems and technical effort was quite high, and that the effort of installation and necessary
line outage made this approach unfeasible from an economical perspective. The Amprion pilot study used
the temperature sensors to measure the validity of weather based DLR. By comparing the calculated line
temperature with the measured one, an accuracy of the DLR model was calculated. Different types of
scenarios relating to different levels of shading etc were compared, and the cumulative frequency of the
deviation between measured and calculated temperature was calculated. The most dangerous type of
deviation is a so called positive deviation, which implies that the measured temperature is higher than the
calculated temperature. This means that the DLR model underestimates line temperature, which could
lead to dangerous situations if the DLR model is allowed to control the ampacity of the line unchecked.
It was concluded that a safety margin of 5 K was necessary to account for this estimation error [16].

The Amprion pilot study also takes the dynamic system wide aspects of an increased capacity into
consideration. It mentions that higher line ampacity would affect both the system stability, as well as the
protection systems. In the latter, a high nominal current would make distinguishing fault and operational
currents from each other more difficult. The study also contained a comprehensive checklist of necessary
measures that had to be completed before DLR was applicable for the overhead line. In conclusion, the
study found that the goal of a capacity of 3150A was attainable 80% of the time [16].

15
Vattenfall has a Dynamic Line Rating pilot ongoing on a 145kV line on the west coast area of Sweden
since 2014. The DLR method is weather based, and both IEEE-738 and a precursor of Cigre-601 called
Cigre TB207 are used and compared. Conductor temperature is also measured to validate the weather
based calculations. Furthermore, the DLR is calculated with commercial software, where two different
softwares have been tested. One developed by USi that calculates the DLR from a single span weather
measurement, and one from GE where the weather measurements support the calculations performed in
a line-and-terrain model. Both these softwares use the IEEE-738 method, whereas Vattenfall performed
the Cigre TB207 calculations themselves. The different calculation methods are compared as 1 minute
averages and 15 minute averages. The result shows that at 1 minute averages, 99% of all data points
are within a ±15% deviation. For a 15 minute average, 99.4% of the data points are within a ±10%
deviation. It is thus concluded that the methods are sufficiently equivalent and that a 15 minute sliding
average is recommended [4].

Vattenfall also investigated whether MESAN-computed weather data from SMHI could be used as an
alternative to local weather measurements. Findings showed that more than 40% of the rates calculated
by data from MESAN were 20% higher than those calculated with local weather. To investigate the effect
of shading, the wind speed and angle of attack were scaled to 50% and 25%. The calculations showed
that 98% of the differences were within ±30%. The explanation for this was that the MESAN values
does not consider the sheltered environment in a line corridor. The poor accuracy of this method shows
that MESAN-computed weather data is not sufficient for DLR calculations [4].

It was also determined that the terrain model DLR calculations underestimated the single span calcula-
tions by more than 10% almost 50% of the time. This is an indication that the single span calculations
likely overestimate the transmission capacity. Furthermore the study showed that 0.2% of the calculated
values underestimated the actual conductor temperature with more than 3 K. This corresponds to a
change in sag which is less than 0.1m. In conclusion, the study showed that Cigre TB207 and IEEE-738
give equivalent results for DLR calculations. The DLR calculations are a correct estimation of the trans-
mission capacity, and a significant increase in capacity from the SLR can be seen. This capacity can be
utilized if the load current can be regulated [4].

Two studies of weather based DLR are compared in [17]. The studies were performed in an area in
North Wales in the U.K, and along Snake River Plane in Idaho U.S. The U.S study used weather
data from a number of meteorological stations combined with lookup tables for wind data generated
by CFD-simulations. It then used the IEEE-738 method to calculate the steady-state current carrying
capacity of the OHL. In the U.K. study, an inverse distance interpolation technique was used instead
of CFD-simulations to estimate wind speed, wind direction, solar radiation and ambient temperature
[18]. The U.K. study also used Monte Carlo simulations to account for the uncertainties in the system.
[17] considered this a key point as understanding the uncertainties of key variables would be essential
for success. However the U.K. method suffered from its crude wind estimation technique, and contained
large modelling errors in the summer when convective cooling is the dominating factor in the cooling of
the line. In conclusion both studies showed promising results, but suffered from inaccuracies in the wind
modelling, which decreased the confidence in the solution.

As mentioned in Section 1, the US Department of Energy has identified DLR as one of eight key metrics
in smart grid transmission and distribution infrastructure [8]. In a report from 2014, the New York Power
Authority (NYPA) and Oncor Electric Delivery Company both demonstrated DLR technologies as part of
the Smart Grid Demonstration Program (SGDP) from the US Department of Energy. The NYPA’s DLR
project was focused on assessing a range of both prototype and commercially available DLR products,
whereas ONCOR focused on demonstrating the commercial viability of established DLR technologies, and
a full-scale automatic integration in real-time system operations [19]. NYPA used conductor temperature
sensors developed by the Electric Power Research Institute (EPRI), as well as a video sagometer, a
weather station and the ThermalRate system developed by Pike Electric Inc [20]. Oncor partnered with
Nexans, who amongst other things have developed the line tension measurement system CAT-1, one of
the most established DLR-solutions on the market. CAT-1 loadcell devices were installed on 345kV and
138kV line segments, along with auxiliary communication and computation devices also provided within
the CAT-1 package [21].

The NYPA encountered reliability issues with the EPRI DLR-system, whereas Oncor encountered fewer
complications with the Nexans system. Both complications were in a large degree related to low load

16
situations, where the temperature difference between conductor and ambient air was small, which lead
to large measurement errors which affected the estimation of effective wind speed. In the end, both
projects revealed an increase in available capacity with the help of DLR. The NYPA also found a positive
correlation between wind farm output, dynamic rating and line loading. The Oncor project was largely
focused around integrating the Nexans DLR system with the ERCOT control room. DLR has been
implemented before, but this is the first time the dynamic ratings have been automatically incorporated
into the system state estimator tool [19].

E.ON installed DLR-systems in its regional grid in 2013. The 50kV network at Öland was a bottleneck
for wind power development in the area, and it was determined that expanding the network with a
new 130kV cable would have been too expensive. The DLR system on Öland employs a wide range
of measurement equipment. It contains weather stations, direct temperature measurement solutions, a
CAT-1 for strain-measurement and also direct measurements of the sag using sensors. One of the most
interesting points about the Öland system is that it is a comprehensive full-scale implementation, with
automatic control of the wind farm power as well as protection system integration. Up until 2018, the
wind park was expected through simulations to be curtailed for about 10h/year. With the DLR system in
place, it has yet to have been curtailed at all due to high temperatures in the conductors [22]. There have
however been both planned maintenance on the OHLs which have lead to curtailment, as well as software
issues with different parts of the system. In 2018, Sweden experienced an unusually warm summer. This
resulted in several spikes in the dynamic load, where the wind farm had to be curtailed. In a system
with static ratings, extreme weather conditions such as this could have resulted in potentially dangerous
loading situations.

The E.ON DLR project on Öland showed that there are financial benefits for the wind power producer
in using DLR for grid connection. For the grid owner there are advantages but also higher losses, as
well as equipment and maintenance costs for the new system components. The project showed that the
challenge is often not primarily in the DLR implementation, but rather the integration with protection
and communication systems [23].

In 2015, Idaho National Laboratory began a DLR study in cooperation with AltaLink, Alberta TSO.
The study used meteorotical and operational data to calculate the dynamic ampacity with DLR, with the
help of a weather based DLR called GLASS, which was developed by Idaho National Laboratory. GLASS
uses the IEEE-738 standard. It also uses CFD to increase the accuracy of wind data inbetween weather
stations. The largest error from CFD modelling was 13%, which would translate to an approximate
ampacity error of 6% as it is roughly proportional to the square root of the wind speed. There were several
areas being studied, and the minimum mean capacity increase that was observed was 22%. However it
is worth noting that since there were no direct measurements on the line, it is difficult to validate the
accuracy of the DLR estimation [8].

In 2015-2016, a study in Northern Ireland showed promising results regarding the use of machine learning
algorithms to forecast weather variables for DLR implementation. The authors argued that the results
proved the feasibility of computing line-rating forecasts using machine learning [24].

17
4 Dynamic Line Rating
Dynamic Line Rating is a collection of procedures that allows power system operators to operate power
lines closer to the thermal limits of the conductor. Whereas a static or seasonal line rating uses con-
servative estimates of ambient variables (such as ambient temperature, solar radiation and wind etc), a
dynamic rating uses continuously measured or forecasted variables. With knowledge of these variables,
the operator can accurately determine the permissible ampacity of a line in the present or imminent
future [25].

This can be done in a number of different ways which will be discussed in Section 4.7, however the
foundation of all dynamic rating is knowledge and control of the heat balance. For overhead lines, the
following components affect the heat balance:

• Joule heating PJ

• Solar heating PS

• Convective cooling qc

• Radiative cooling qr

• Magnetic heating PM

• Corona heating PI

• Evaporative cooling qw

There are two leading models for calculation of line heat balance, developed by IEEE and CIGRE. Both
use some or all of the aforementioned components, however the components are not calculated in the
same way for both models. Hence the most intuitive way to present these models is by a side by side
comparison for each thermal component. This section will contain the fundamental differences between
the methods. A more comprehensive comparison can be found in [25],[26].

Cigre model IEEE model [9]


PJ + PS + PM + PI = qc + qr + qw (1) PJ + PS = qc + qr (2)

The Cigre model has three components that the IEEE-model does not have, namely magnetic heating
(PM ), corona heating (PI ) and evaporative cooling (qw ). However, it is difficult to measure humidity along
the line for long distances hence evaporative cooling is often omitted in the heat balance calculations.
Furthermore for conductors made from aluminium alloy the magnetic and corona heating are sufficiently
insignificant to be omitted as well [27]. This leaves four components; joule heating (PJ ), solar heating
(PS ), convective cooling (qc ) and radiative cooling (qr ).

4.1 Joule heating


Joule heating can be described as the influx of thermal energy as a result of resistive losses in the
conductor. The definitions for joule heating per unit length differ between the IEEE and Cigre models.
In the IEEE-738 model, the joule heating is described as:

PJ−IEEE = I 2 · R(T ) [W/m] (3)

where I is the RMS-value of the current, and R(T ) is the electrical resistance of the conductor as a
function of temperature T .
R(Thigh ) − R(Tlow )
R(T ) = · (T − Tlow ) + R(Tlow ) [Ω/m] (4)
Thigh − Tlow

In (4), R(Thigh ) and R(Tlow ) are values of the conductor resistance at high and low temperature. These

18
values are either given by the conductor manufacturer, or can be found in handbooks for ACSR con-
ductors. They are widely accepted within the community and include magnetic effects, skin effect and
lay ratios. They enable calculations of the resistance at every temperature T between Tlow and Thigh
through linear interpolation [9].

The Cigre-601 method does not use empirical values to calculate R(T ) as the IEEE-738 method, but
instead uses the DC-resistance Rdc , which gives:

PJ−Cigre = I 2 · Rdc (T ) [W/m] (5)

The DC-resistance is calculated from the resistivity of the materials at the given temperature, the cross-
sectional area and the conductor temperature. The resistivity of a material at temperature Ts is:

R(Ts ) = R20 · [1 + α20 (Ts − 20) + ζ20 (Ts − 20)2 ] [Ω · m] (6)

The ζ-component is the quadratic temperature coefficient which only becomes significant at temperatures
higher than 130◦ C. Hence for DLR-calculations at standard temperatures, it may be omitted.

As the conductor is energised with alternating current, the skin effect factor is present in the AC resistance.
In these cases the Joule heat gain is:

PJ−Cigre = ksk · I 2 · Rdc (T ) [W/m] (7)

where ksk is the skin effect factor. For standard applications of conductor diameter and load frequency,
the skin effect factor is below 2% [10].

4.2 Convective cooling


Both the IEEE-738 and Cigre-601 methods divides convective cooling into two types: Natural Convection
(in some cases denoted Free Convection) and Forced Convection. Natural Convection occurs during low
wind speeds. The conductor heats surrounding air, which rises and is replaced by cool air. During higher
wind speeds, air is blown past the conductor and thus carries heat away. Both methods also recommend
that only the larger one of the components is used (and not both), as this is conservative [9],[10].

There are two equations that describes Forced Convection according to the IEEE-738 method. One
that is accurate at low wind speeds but underestimates the heat loss at high wind speeds, and one
that is accurate at high wind speeds but underestimates the heat loss at low wind speeds. The method
recommends that both equations are solved for, and the larger one is used. Equation (8) is more accurate
at low wind speeds, and (9) at high wind speeds.
0.52
qc1 = Kangle [1.01 + 1.35 · NRe ] · λf −ieee (Ts − Ta ) [W/m] (8)
0.6
qc2 = Kangle · 0.754 · NRe · λf −ieee (Ts − Ta ) [W/m] (9)
Ts is the surface temperature of the conductor, Ta is the ambient temperature around the conductor,
Kangle is the wind direction factor. It is a function of δ which is the angle between the wind direction
and the conductor axis. The wind direction factor can be seen in (10) and is at its maximum when the
wind is perpendicular to the conductor axis, i.e. at δ = ± π2 .

Kangle = 1.194 − cos(δ) + 0.194 cos(2δ) + 0.368 sin(2δ) (10)

NRe is the Reynolds number which is given by (11), and describes the magnitude of convective heat loss.
D · ρf · Vw
NRe = (11)
µf −ieee
where D is the diameter of the conductor, Vw is the wind velocity, ρf is the air density and µf −ieee is
the dynamic viscosity of air, calculated according to the IEEE-738 method. ρf and µf −ieee are calcu-
lated using the mean film temperature of the conductor boundary layer; Tf and the elevation He . The
calculations for these variables can be seen in (12) - (14).
Ts + Ta
Tf = [◦ C] (12)
2

19
1.293 − 1.525 · 10−4 · He + 6.379 · 10−9 · He2
ρf = [kg/m3 ] (13)
1 + 0.00367 · Tf
1.458 · 10−6 · (Tf + 273)1.5
µf −ieee = [kg/m · s] (14)
Tf + 383.4
λf −ieee is the thermal conductivity of air, calculated according to the IEEE-738 method. It too is a
function of Tf and can be seen in (15).

λf −ieee = 2.424 · 10−2 + 7.477 · 10−5 · Tf − 4.407 · 10−9 · Tf2 [W/m ·◦ C] (15)

The Natural Convection (or Free Convection) is described in (16) with some of the parameters used to
calculate Forced Convection [9].

qcn = 3.645 · ρf0.5 · D0.75 · (Ts − Ta )1.25 [W/m] (16)

The convective cooling according to the Cigre-601 method is seen in (17)

qc−cigre = π · λf −cigre · (Ts − Ta ) · N u [W/m] (17)

where λf −ieee is the thermal conductivity of air calculated by the Cigre-601 method at the film temper-
ature Tf as seen in (18) and N u is the Nusselt number, which is calculated in different ways depending
on type of convection.

λf −cigre = 2.368 · 10−2 + 7.23 · 10−5 · Tf − 2.763 · 10−8 · Tf2 [W/m ·◦ C] (18)

The two methods have two different ways to calculate the thermal conductivity at film temperature,
hence the different subscripts for λf . As seen in Figure 1, these are not interchangable. The IEEE-738
method consistently estimates the thermal conductivity higher than the Cigre-601 method. Due to this
discrepency, the methods should employ individual calculation of thermal conductivity according to their
own standard, and a general expression that applies for both methods should not be used.

Figure 1: Comparison of thermal conducitivity calculated by the different standards

As mentioned previously, the Nusselt number is calculated differently depending on the type of convection.
In the case of Natural Convection, the Nusselt number is calculated as seen in (19).

N unat = A · (Gr · Pr )m (19)

where Gr is the Grashof number (20), and P r is the Prandtl number (21).

D3 · (Ts − Ta ) · g
Gr = µ (20)
(Tf + 273) · f −cigre
ρf

20
Figure 2: Comparison of the IEEE and Cigre calculation methods for dynamic viscosity of air

cf · µf −cigre
Pr = (21)
λf −cigre

The dynamic viscosity of air is also calculated differently from the IEEE-738 method, as seen in (22).
However for the range of temperatures that is relevant for the applications in this report, the two methods
are interchangable as seen in Figure 2.

µf −cigre = (17.239 + 4.635 · 10−2 · Tf − 2.03 · 10−5 · Tf2 ) · 10−6 [kg/m · s] (22)

The coefficients A and m in (19) vary depending on the product of the Grashof and Prandtl number.
The different cases can be seen in Table 3. For forced convection, the Nusselt number for perpendicular

Range of (Gr · Pr ) A m
From To
10−1 102 1.02 0.148
102 104 0.850 0.188
104 107 0.480 0.250
107 1012 0.125 0.333

Table 3: Coefficients for Natural Convection

wind flow is calculated in a similar manner, with Equation (23).


n
N u90 = B · NRe (23)

The Reynolds number is calculated in the same way as for the IEEE-738 method as in (11). B and n
depends on the roughness Rs , and the Reynolds number, and can be seen in Table 4. There is a third
set of values for B and n, that applies for smooth conductors. However as the conductors in this project
are stranded, this set is omitted.

The Nusselt number for perpendicular wind flow is then adjusted for wind direction factor. Hence the
Nusselt number for forced convection is calculated according to the equation seen in (24).
(
(0.42 + 0.68 · (sin(δ))1.05 ) · N u90 if δ ≤ 24◦
N uδ = (24)
0.90
(0.42 + 0.58 · (sin(δ)) ) · N u90 if δ > 24◦

21
Stranded Conductors Stranded Conductors
Rs ≤ 0.05 Rs > 0.05
NRe B n B n
100 - 2,650 0.641 0.471 0.641 0.471
2,650 - 50,000 0.178 0.633 0.048 0.800

Table 4: Forced convection coefficients in Cigre-601 method

The Cigre-601 method makes a point to mention that turbulence has a considerable effect on the con-
vection cooling rate, but is very difficult to assess for real overhead lines. At high winds, considerable
errors can occur in dynamic rating calculations due to factors such as variability of wind, obstacles and
bundled conductors.

In conclusion the Cigre-601 method recommends the use of the higher of the natural and forced convection
values determined in Equation (19) and Equation (23). Since the convective cooling factor is proportional
to the Nusselt number, it is convenient to compare which Nusselt number is the largest, and then use
that for the calculation of the convective cooling factor by inserting the largest Nusselt number into (17)
[10].

4.3 Solar heating


Solar heating is the heating component that comes from the sun shining onto the conductor. The heat
energy transfer from solar heating depends on a number of factors such as the sun’s position, the heat
intensity, absorptivity of the conductor and the surface area of the conductor. The expression for solar
heating according to the IEEE-738 method can be seen in (25) [9].
Ps−ieee = α · Qse · sin(θ) · D0 [W/m] (25)
where α is the absorptivity of the conductor. The absorptivity depends on the condition of the conductor.
A new, shiny conductor will reflect most of the sun’s energy whereas an older, blackened conductor will
absorb the sun’s energy instead. α lies in the range of 0 ≤ α ≤ 1 where 0 indicates full reflection and 1
full absorption. The angle θ is effective angle of incidence for the sun rays and depends on a trigonometric
function of solar altitude and solar azimuth. Qse is the solar heat flux adjusted for elevation
Qse = Ksolar Qs [W/m2 ] (26)
where:
Ksolar = 1 + 1.148 · 10−4 · He − 1.108 · 10−8 · He2 (27)
is the elevation adjustment factor, and Qs is the heat flux intensity at sea level, which depends on the
solar altitude and the quality of the atmosphere (if the conductor is located in an industrial or clear
atmosphere).

D0 is the projected area of the conductor, which depends on not only the diameter of the conductor, but
also it’s orientation compared to east-west. Further derivations for how to calculate solar altitude, solar
azimuth, heat flux intensity at sea level etc can be found in [9].

The IEEE-738 method estimates the solar heat flux from the type of atmosphere the conductor is in, and
the sun’s position in the sky. Hence there aren’t any actual measurements for these calculations, but the
heat flux is calculated from the latitude, day of the year and time of day. It does not take into account
variable factors such as cloud coverage.

The Cigre-601 method calculates the solar heat gain from the outer diameter of the conductor, the
absorptivity of the surface and the global radiation intensity, as seen in (28).
Ps = α · I T · D [W/m] (28)

According to the Cigre-601 method, measurement devices for global radiation intensity are reliable and
indexpensive, and suitable for line monitoring systems and dynamic line rating calculations. However, the

22
user should be adviced that the global radiation received by the conductor may vary along the line due to
factors such as shading, cloud coverage, orientation of the line, reflectance from ground, etc [10]. Global
radiation consists of both direct radiation and diffuse radiation. While the estimation-based method of
IEEE-738 only considers direct radiation, the global radiation measurements of Cigre-601 considers both
direct and diffuse solar radiation.

4.4 Radiative cooling


Radiative heat transfer occurs due to the emission of electromagnetic waves into the surrounding medium.
The conductor has a higher temperature than the surrounding medium, and therefore it radiates thermal
energy at a rate that is proportional to the temperature difference to the power of 4 [27]. The equation
for the radiative cooling component according to the IEEE-738 method can be found in (29).
 
Ts + 273 4 Ta + 273 4
qr−ieee = 17.8 · D · ε ( ) −( ) [W/m] (29)
100 100
Emissivity - denoted by ε - is the radiative material property which describes the rate the conductor
emits heat. It increases with age according to the relation shown in (30) [28].
Y
ε = 0.23 + 0.7 (30)
1.22 + Y
where Y is the number of years in operation. It is necessary to mention that the aforementioned equation
is a rough estimate, as the aging process is complicated and difficult to assess with absolute precision.

The equation presented in the IEEE-738 method is in fact a numerical representation of a more general
representation of radiative heat transfer, which employs the Stefan-Boltzmann constant. This represen-
tation is presented in the Cigre-601 method seen in (31) [10].

qr−cigre = π · D · σB · ε · [(Ts + 273)4 − (Ta + 273)4 ] [W/m] (31)

A comparison of the (29) and (31) shows that


 
Ts + 273 4 Ta + 273 4
17.8 · D0 · ε ( ) −( ) ≈ π · D · σB · ε · [(Ts + 273)4 − (Ta + 273)4 ] =⇒
100 100

17.8 (32)
≈ σB · π =⇒
1004

1.78 · 10−7 ≈ 1.78128 · 10−7


It’s clear that the two representations are equivalent as the discrepancy from the approximation in the
IEEE-method is as little as 0.128 · 10−9 . Hence it can be said that

qr−cigre = qr−ieee (33)

and that, since the Cigre method is slightly less approximative, for this project

qr = qr−cigre [W/m] (34)

4.5 Thermal calculations


Regardless of which method is used, the thermal balance equation is the same. Henceforth the calculations
will yet again be presented in a general state, as they are applicable for both methods. When calculating
the thermal behaviour of the line, there are two cases to consider. The steady state, which can be
considered a momentary observation of the system behaviour, and the dynamic state which considers the
thermal behaviour of the system for some period of time dt [25], [9], [10].

4.5.1 Steady state


The steady state method utilises the basic thermal balance presented in 4 and seen in (2). To calculate
the ampacity, (3) can be inserted into (2) which gives:

I 2 · R(T ) + Ps = qc + qr (35)

23
The ampacity is then solved for as a function of the thermal components:
s
qc + qr − Ps
I= [A] (36)
R(T )

4.5.2 Dynamic state


The steady state method is a quick and simple way to determine the permissible ampacity for a given
set of weather parameters, however it fails to take into consideration the thermal inertia of the line [25].
If one instead would like to calculate the temperature of the line, or the time for a temperature shift of
the line during dynamic conditions, the thermal inertia of the line must be considered. The change in
temperature can then be expressed with the equation in (37)[10].

dTs
m·c· = PJ + Ps − qc − qr (37)
dt
The mass m and specific heat capacity c determines the amount of energy required to heat a unit length
of the conductor one degree ◦ C (or K). The mass density can be considered constant for the range of
relevant temperatures for this project. The specific heat capacity is temperature dependent as seen in
(38), however for the ranges of conductor temperature expected in this project (≤ 200◦ C), it can also be
considered constant [29].
c(T ) = c20 · [1 + β20 · (T − 20)] [J/kg · K] (38)
The Cigre-601 method suggests that this equation can be solved numerically to calculate the temperature
of the line, with 5-15 minute time steps as suitable. However, it is not suitable to use this method for
fault current applications. In these cases, due to the short time-frame of a high current fault, it is more
suitable to consider adiabatic conditions. As this is not relevant for the project, this method of calculation
is omitted [10].

4.6 Sag theory


The sag of the line is a parameter of great interest for power system operators. It is the distance from
the fastening point of the line, to the vertex of the catenary of the line. This is directly connected to the
distance-to-ground of the line, which in Sweden is regulated by [30]. The minimum allowable distance
to ground is 7 + s meters for populated areas and 6 + s for unpopulated areas. s is an additional safety
margin dependent on the voltage of the system. S = 0.4m for a directly earthed 132kV system, which
would yield total safety limits of 7.4m and 6.4m respectively.

In this section, a method to calculate the sag of a line will be presented. This method is presented in
Cigre TB324. Cigre makes a point to underline that regardless of the calculation method, a clearance
buffer is still required to account for modeling undercertainties when designing new lines, or uprating old
ones [1].

When subjected to uniform loading per unit length, an overhead line takes the form of a catenary between
the support points of the surrounding towers. The shape of the catenary is a function of the weight of the
conductor per unit length, w, and the horizontal component of tension, H. The sag itself is a function of
these parameters, as well as the span length , S and elevation difference between the fastening points.

The catenary equation can be seen in (39), where y(x) is the height of the catenary above the vertex,
and x is the horizontal distance from the vertex.

H h w·x i
y(x) = · cosh ( )−1 [m] (39)
w H

The sag Ψ is found at x = S/2. The expression for sag can thus be seen in (40).
 
H w·S
Ψ= · cosh ( )−1 [m] (40)
w 2·H

24
Figure 3: Inclined catenary span [1]

This is correct for level spans, where the fastening points are at the same height in both ends. For inclined
spans, the vertex will be shifted off-center, and as such the new position of the vertex must be calculated.
The equation for calculating the vertex for an inclined span is seen in (41).
 
S H h/2
XR = − sinh−1   [m] (41)
2 w H
wsinh S/2
H/w

XR is the distance from the lowest fastening point to the vertex, and h is the difference in elevation
between the fastening points. A representation of an inclined span can be seen in Figure 3. The variable
that in the figure is denoted as D has in this report been called Ψ, in order to avoid confusion with the
diameter of a conductor.

The sag can now be calculated depending on the reference point. The sag with respect to the lowest
fastening point can be found by inserting XR into (39). To get the sag from the highest fastening point,
one must instead insert XL = S − XR [1].

To accurately account for the dynamic conditions of an OHL, the elongation of the line during such
conditions must be estimated. To do so, Cigre TB324 has grouped such methods into three types of
models; Linear Elastic (LE) model, Simplified Plastic Elongation (SPE) model, and Experimental Plastic
Elongation (EPE) model. For the scope of this project, the LE-model will be sufficient and as such it
is the only model that will be presented here. The curious reader is referred to [1] for a more rigorous
description of SPE and EPE models.

The LE model omits two types of elongation factors, which are ”settlement and strand deformation”
and ”long-time or high tension plastic strain”. Because of this, the LE model is an underestimation of
the actual sag (and require clearance buffers for safe operation). The parameter of interest to estimate
for this project is the linear thermal strain of an ACSR conductor. The composite coefficient of linear
thermal expansion for an ACSR conductor can be calculated according to (42).
     
EA AA ES AS
αas = αA + αS [K −1 ] (42)
EAS AAS EAS AAS

The thermal elongation can then be calculated as seen in (43)

∆L = αAS · ∆Ts · L [m] (43)

The sag of the conductor can then be estimated by (44).


r
3 · S · (L − S)
Ψ= [m] (44)
8

25
In a Master Thesis from 2011, Elisabet Lindberg estimated the correlation between temperature and
distance to ground on a 130kV line belonging to Vattenfall. It was determined that the distance to
ground decreased (i.e. the sag increased) by 3.7 cm per degree. However, the line in question had a
diameter of 39.2mm whereas the line in this project has a diameter of 19.9mm. Furthermore the sag-
thermal relationship also depends on the span length. Thus the results are not directly applicable for the
area in this project [31].

4.7 Types of Dynamic Line Rating


There are many types of methods for line rating, some of which fall within the scope of Dynamic Line
Rating. Kateryna Morozovska and Patrik Hilber have proposed a categorization of line rating methods
in [32]. The categories are:

• Static line rating (SLR): A standard fixed rating, which is specified by international or national
standards, and conservative weather assumptions.

• Seasonal rating (SER): A summer/winter rating or in some cases summer/autumn/winter rating,


where the rating is set separately for a span of time based on a conservative estimate of the typical
weather conditions of the period.

• Weather model (WM): A rating based on averaged weather data for several years. More accurate
than the seasonal rating.

• Weather forecast (WF): A rating based on real time weather that data that is collected near
the conductor and used to create a weather forecast. The ratings are set according to the forecast.

• Conductor temperature evaluation (CTE): When the conductor temperature is measured


with the help of a temperature sensor.

• Tension monitoring (TM): When the tension of the line is monitored, by placing load cells in
series with the insulator strings. Most tension systems requires installation of weather monitoring
equipment for further evaluation of system parameters and calculation of the ampacity of the line.

• Conductor sag measurement (CSM): A more advanced system that can actually measure the
sag of the line. By placing such equipment in the worst case parts of the power system, it can be
operated within safety margins.

• Clearance-to-ground measurement (CTGM): A new type of monitoring system for OHLs


that does not measure sag, but clearance-to-ground through imaging hardware either mounted on
the line or on the ground below the line. It is a more relevant measurement as it gives direct
information about the clearance to ground, which is a limiting factor in power system operation.

• Full scale monitoring (FSM): This method can be a combination of several of the aforementioned
methods. The main feature of this category is the placement of sensors along the line. However
with present technology, this is expensive.

Given these categories, the different types of DLR mentioned in this report can be assigned to the correct
category. This categorization can be seen in Table 5.

26
Monitoring System Name WF CTE TM CSM CTGM FSM
ADR Sense X
CAT-1 [21] X X
EPRI Sensor [20] X X X1
EPRI Sagometer [20] X
Heimdall Neuron X X1
Power Donut [33] X X1
SAW Technology [16] X
ThermalRateTM [20] X

Table 5: Checklist table for which monitoring systems support which DLR methods

1 These types of measuring equipment all estimate sag by using a 3-axis accelerometer to determine the inclination of

the line at their point of suspension.

27
5 Pilot study
Ellevio AB has in collaboration with Heimdall Power launched a pilot study to determine the viability
of using DLR as a cost efficient way to improve the transmission capacity. The area of the study
is located in the south west of Sweden, where Heimdall Neurons are mounted on a 24km long 130kV
regional line. A weather station is placed in the nearest substation to measure ambient temperature, wind
speed, wind direction and global radiation. A collection of scripts have been developed in the statistical
computing language R to analyse the potential increase in transmission capacity, and validate the Neuron
measurements. This section contains an explanation of the test area, the calculation methods and the
scripts used to estimate the Dynamic Line Rating and thermal components using both the IEEE-738 and
the Cigre-601 calculation methods, and a discussion regarding said results.

5.1 Test area


The line in the pilot study is located in the south-west of Sweden. It connects generation from nearby
wind farms and imported power from Norway to an adjacent urban center. Future wind power integration
into the network is impeded by the capacity of the line in question, which in this report will be referred
to as VL-X. A pre-study for the pilot determined that the terrain height for the line varies within 58-208
meters above sea level. The wind direction is mostly from south-west, but is highly variable close to the
ground due to shading from terrain and vegetation. The line is an ACSR-conductor of the type Ibis. The
relevant parameters for the line can be seen in Table 6.

Diameter
Area # threads Thread Core Line Weight Tensile strength DC-resistance Heat capacity
Al Fe
2
mm Al Fe mm mm mm mm kg/km kN Ω/km J/K · kg
234 26 7 3.14 2.44 7.32 19.9 812 70.53 0.1434 788.6

Table 6: Data for VL-X conductor of type Ibis

In addition to the conductor data presented in Table 6, there is additional data used as input variables
for the DLR calculations. These can be seen in Table 7. As mentioned in Section 4, the absorptivity
and emissivity are used to determine solar heating and radiative cooling respectively. The latitude is
an input parameter in the estimation of solar heating used in the IEEE-738 method. The elevation and
line orientation are used in both IEEE-738 solar heating estimations and convective cooling calculations.
The atmosphere type is relevant for the solar heating estimation for the IEEE-738 method, and is set
to clear as the line is located in a rural area with low air pollution levels. The aluminum temperature
coefficient is used for the entire conductor (although part of it is steel) to simplify calculations. It is used
to estimate the linear temperature dependence for the resistivity of aluminum.

5.2 DLR Calculations


As shown in Section 4.7 there are many methods for DLR. In this section, the reasoning behind the choice
of calculation method used in the project, and an explanation of said method is presented.

The choice of method is based on the available data for the line. The available measurement equipment
measures the weather conditions (ambient temperature, wind speed and direction, global radiation) and
the line temperature and sag. This allows for three types of DLR; WF, CTE or CSM. Early in the
project, either of these methods were a possibility. However during the analysis and validation of the
measurement devices seen in Section 6, it was determined that the line temperature measurements of the
Neuron were not working correctly. Furthermore the angle measurement from which the sag of the line
was derived was in need of calibration. Neither of these issues were fixed before the end of the project,
hence neither CTE nor CSM was applicable. They are however still interesting methods, and exploration
of said methods are mentioned in Section 11 as areas for future study.

The method of elimination thus yielded a weather based DLR method as the only possible alternative for
the duration of the project. It is based on the thermal components calculated either from the IEEE-738

28
Variable Value
Absorptivity 0.8
Emissivity 0.8
Latitude 59◦
Elevation 70 m
Type of Atmosphere Clear
Line orientation 270◦
Aluminum temperature coefficient 0.00429 Ω/K
Line temperature limit 50◦ C
SLR - Winter 531 A
SLR - Summer 344 A

Table 7: Input variables for DLR calculations

or Cigre-601 method (described in detail in Section 4). For this project, both the IEEE-738 and Cigre-
601 methods are used and compared. It is worth mentioning that as no direct measurements of either
temperature or sag are available at this time, the validation of the DLR methods (in terms of accuracy of
the temperature estimation of the line) is not possible. This too is an important aspect to study further
once either the temperature or sag measurements (or both) are working correctly, and suggested as a
topic for future work in Section 11.

The DLR calculations are realised in a series of scripts in the statistical computing environment R.
The contents of the scripts are shown in Appendix A.1, and merely a brief summary is presented here.
There is a main script where the DLR calculation is located, called main DLR. The main script calls
upon a separate calculation script where the functions for calculating the thermal components and the
temperature-dependent resistivity are contained, shown in Appendix A.2 and A.3. These functions take
weather data and line parameters as input, and return the heating or cooling parameter of the line in the
unit W/m. The main script iterates with an internal time variable, which is compared with the time-
stamps of the weather data to find the weather conditions at a certain time. The weather conditions are
then used in the thermal component functions, and the thermal components are then used to calculate the
DLR. Once an iteration is complete, the internal time variable is increased and the iteration is repeated.
Thus the script iterates between two chosen UTC date-times to find the DLR between said times. A
sliding window average of the last 15 minutes is calculated in each iteration, in order to obtain a more
stable output. This is recommended in [10] as the raw data output is highly variable due to the strong
influence of the also highly variable convective cooling.

For the present application, the calculation scripts take Excel datasheets of past measurements from the
weather station as input data. However the script could be adjusted to operate in real-time (or close to
real time) with little effort, and has consciously been built to iterate through the data as were it collected
in real time.

5.3 DLR Results


In order to calculate the DLR of the line, the thermal components are calculated. This section will
contain the output data for these component for a week in November to give a clearer picture of the
nature and relative importance of these components. In addition to this, the DLR and the percentage
change in capacity will be shown in this section also.

5.3.1 Solar heating


The solar heating for the first week of November is shown in Figure 4.

It can be seen that the IEEE-738 estimation is consistently higher than the Cigre-601 calculation. Fur-
thermore the amplitude of this component is peaking around 10 W/m for the IEEE-estimation and 3

29
Figure 4: Solar heating thermal component from IEEE-738 and Cigre-601 calculation methods

W/m for the Cigre calculation. Hence the relative difference is quite large, but the actual difference and
overall influence of this component is small.

5.3.2 Radiative cooling


A number of things of interest can be seen for the radiative cooling. The radiative cooling for the first
week of November can be seen in Figure 5.

Figure 5: Radiative cooling thermal component from IEEE-738 and Cigre-601 calculation methods

The first interesting thing to note is that the methods are seemingly interchangable for the radiative
cooling estimation. This is to be expected as seen Section 4.4, as the methods are virtually identical.
Another interesting thing to note is that the radiative cooling is fairly stable between 14-16 W/m for the
entire week. This too is to be expected since it only depends on the ambient temperature, which is not
prone to rapid changes.

5.3.3 Convective cooling


It can be seen that there is a strong correlation between how the convective cooling is calculated in
Figure 6, where the convective cooling for the first week of November is plotted.

30
Figure 6: Convective cooling component from IEEE-738 and Cigre-601 calculation methods

This component is highly variable, as it depends on two highly variable inputs in the form of wind speed
and wind direction (and a less variable third component in ambient temperature). Despite the high
variability, both methods return essentially the same data for the convective cooling. There is a baseline
around 21 W/m which is due to the Natural Convection. The convective cooling peaks around 150 W/m
for the plotted week and 241 W/m for the entire dataset.

5.3.4 Dynamic Line Rating


The Dynamic Line Rating is naturally the most important output from the DLR calculation script. The
DLR from the same week as the previously shown thermal components can be seen in Figure 7.

Figure 7: Dynamic Line Rating from IEEE-738 and Cigre-601 calculation methods for first week of
November

It can be seen that despite the 15 minute sliding average, the dynamic rating is still highly variable.
However it is significantly less variable than the raw convection cooling component seen in Figure 6,
partially due to the sliding window averaging, partially due to a dampening effect from the other thermal
components. It can also be seen that there is a strong correlation between the IEEE-738 and Cigre-601
methods.

31
The DLR for the entire measurement period can be seen in Figure 8. Two anomalies can be seen, one in

Figure 8: Dynamic Line Rating for the entire measurement period with IEEE-738 and Cigre-601 calcu-
lation methods

the DLR (magenta and green) and one in the load (black). The load signal anomaly is due to an outage
of the current measurement system (Heimdall Neurons) between the 2nd to the 18th of December. This
is not an issue for the DLR calculations as the load is not used for these. It would have been an issue
for temperature estimation, and also DLR calculations should they have used the line temperature or
inclination, but as that is currently not the case this outage in the measurement equipment is of no
concern for the DLR calculations.

The second anomaly is seen between the 28th of December and the 3rd of January. This is due to an
outage in the weather station. The DLR calculation script is built to be able to deal with these types
of outages. It continuously checks the difference between the internal iteration clock and the weather
data time stamp. If this difference exceeds 15 minutes, the SLR rating is imposed as a safety measure.
The alternative would be a dynamic rating stuck at the last measurement values, which would be highly
unsafe and of course unacceptable in actual operation. Hence what can be seen in this time period is
both the IEEE-738 and Cigre-601 rating being set to the SLR for the duration of the outage. This does
not break the calculation script, nor does it set the rating at an unsafe level. The only inconvenience
with this malfunction is a temporary decrease in capacity for the line.

The increase in capacity for the line compared with using the SLR is 23% for the measurement period.
It is the same for both IEEE-738 and Cigre-601 based DLR.

32
6 Validation of measurement devices
This section contains an analysis of the validity of the measurement devices. There are two types of mea-
surement devices in the pilot project; a weather station that measures ambient temperature, wind speed,
wind direction and global radiation, and the Heimdall Neurons which measure current, line temperature
and sag (through measuring the angle of inclination at the top of the span). Each parameter for each
measurement device is estimated separately, to determine if they are working correctly.

6.1 Weather station


The weather station is of the type HY-WDS6SE. Each of its measurements is analysed in this section.

6.1.1 Temperature measurements


The ambient temperature is measured by a MEMS sensor. It has the accuracy of ±0.2% and a resolution of
0.1◦ C. By comparing the temperature measurements from the weather station with similar measurements
from nearby SMHI weather stations, the validity of the temperature measurements can be verified. If
the measurements from the weather station deviates significantly from the control measurements, it may
not be functioning properly and the confidence in said measurements should decrease.

In Figure 9 the ambient temperature from the weather station and SMHI measurements in Karlstad
and Arvika are plotted for the month of November. It can be seen that the temperature correlates well
with the control measurements. Some deviations due to regional differences are to be expected, but the
temperatures are often within 1◦ C. This shows that the temperature measurements from the weather
station are working correctly, and can be trusted in the DLR estimation.

Figure 9: Ambient temperature from weather station and control measurements in Arvika and Karlstad

6.1.2 Wind measurements


The wind speed and direction are measured with an ultrasonic sensor. The wind speed has an accuracy
of ±3% and a resolution of 0.01 m/s. The wind direction has an accuracy of ≤ 3◦ and a resolution of
1◦ . These measurements are harder to verify due to their inherent variability. Hence a larger degree of
leniency is allowed in this verification.

To attempt to verify the validity of the wind measurements, the data is compared with measurements
from the closest SMHI station (in Arvika) for two separate months. The comparison for the month of
November can be seen in Figure 10. By comparing the wind roses for Arvika and the weather station, it
can be seen that the wind speed at the weather station is generally lower than in Arvika. Furthermore
there is some correlation between the wind direction from the east, despite the difference in magnitude.
The wind influx from the north in Arvika is not seen at all at the weather station, which may be attributed

33
(a) Weather station measurements (b) Arvika measurements

Figure 10: Wind rose plots for November

to possible shading of the weather station. Given the variability of wind, and the strong correlation of
wind from the east, this comparison gives no reason to doubt the wind measurements.

The same comparison for January can be seen in Figure 11. For January, the average wind speed was
higher at the weather station than in Arvika. Furthermore the wind directions correlate poorly, which
yet again may be attributed to the variability of wind. In conclusion there is no reason to believe that the

(a) Weather station measurements (b) Arvika measurements

Figure 11: Wind rose plots for January

wind measurements are not functioning properly from this analysis. Both wind speed and wind direction
are expected to vary a lot even in short distances, hence nothing in these measurements are out of the
ordinary. There are some concerns that the weather station may be shaded from the north, however this
is not a severe issue as it will simply yield a conservative DLR estimation when the wind comes from the
north.

6.1.3 Solar irradiance measurements


The solar irradiance is measured with a silicon sensor mounted at the top of the weather station. It
has an accuracy of ±5% and a resolution of 1 W/m2 . It is compared with measurements from Karlstad
to determine if the measured values show a reasonable correlation. The comparison between the global
radiation measured in the weather station and in Karlstad can be seen in Figure 12. It is compared for
the month of November.

It can be seen that there is good correlation between the different measurements. The solar irradiance is
often (but not always) higher in Karlstad, but not outside what can be expected from regional deviations.
Furthermore the time when global radiation is registered matches very well. The days of low irradiance
around November 25th also matches very well, which further suggests that the measurements from the
weather station are correct.

34
Figure 12: Global radiation measured in weather station and Karlstad for November

Another interesting comparison that can and made in this section is that between the measured global
irradiance, and that estimated by the IEEE-738 standard. Since the IEEE-738 estimation does not take
shading from cloud coverage etc into account, it should be higher than the measured global radiation at
all times. This comparison can be seen in Figure 13. It is clear that this is indeed the case.

Figure 13: Global radiation measured by weather station, Karlstad and estimated by IEEE-738 for
November

6.2 Heimdall Neuron


The Heimdall Neuron has three main measurements that are of interest for the applications in this project.
It measures line current, line temperature and the angle of inclination at the position of the neuron, which
is close to the fastening point of the span. Each of these measurements is analysed in this section.

6.2.1 Line current measurements


In this section, the line current measurements are compared with current measurements from a nearby
substation. The current measurements from the substation can be considered to be credible, hence if
there is a strong correlation then the line current measurements from the neurons can be considered

35
correct. The current measurements from the neurons and the substation can be seen in Figure 14. It is
clearly shown that the measurements correlate strongly, and hence the current measurements from the
neurons can be considered correct.

Figure 14: Line current measurements from all neurons and a nearby substation

6.2.2 Line temperature measurements


Early in the project, the neuron line temperature measurements were a point of interest, as they seemed
to display strange characteristics. More specifically, the temperature measurements suggested that the
line temperature was below the ambient temperature for a large portion of the time. This sparked a
rigorous investigation regarding the validity of the temperature measurements.

The verification of the temperature measurements is done with a reversed DLR-calculation script. In
a straight forward DLR-calculation, a temperature limit is set and the ampacity is calculated with the
help of (36). However if the current is known, this equation can be reversed so that the current is
used to calculate the temperature instead. The aforementioned equation assumes steady state, which
is not a valid assumption to make for temperature calculations since the temperature does not change
instantaneously, nor does the input parameters stay constant for extended periods of time. Hence rather
than calculating the temperature the line would have if it would be subject to a specific set of static
conditions for a sufficiently long period of time, another means of calculation must be used. Both [9] and
[10] have suggestions on how to do this, and such an approach was taken for this project.

The calculation script for temperature estimation utilises an iterative method that considers the thermal
inertia of the line, and calculates the change in temperature for a given time step. The temperature
change for a given time follows (45). It’s essentially a rewritten form of (37), where the net thermal
energy for a given period of time is divided by the thermal inertia which is given by m · c. The script
utilizes a start guess, and then iterates toward the true value of what the temperature should be for the
line, given the load and weather conditions.

(PJ + Ps − qc − qr ) · dt
dTs = [K] (45)
m·c

This method has a few drawbacks. Firstly the start guess means the script requires a few iterations to
reach the true temperature value. Secondly the resolution of the time-steps limits the accuracy of the
estimation. Ideally, these time-steps should be infinitely small in order to detect variations in the net
energy instantly. However this is impractical (and impossible), partially due to the required computation
power, partially because the input data has time-steps of around 30 seconds. A third drawback is the
input data itself. The temperature estimation is only as good as the input data, and since the weather
station is located around 200 meters from the closest neuron, there may be some differences in measured

36
and actual weather conditions for the line. With all these drawbacks in mind, this is still the best possible
way to estimate the temperature of the line, given the available input data.

The temperature for the day in November with the highest load can be seen in Figure 15. It is plotted with
the load to show how the estimated and measured temperature correlates with the load (which drives the
joule heating thermal component). It is clear that the neuron temperature measurement underestimates
the temperature, given how it is below the estimated temperature for the entire day. Furthermore it
is a clear indication of incorrect operation when the load is around 400 A (which is around 80% of the
static rating) for an extended period of time, but the measured line temperature is around 2◦ C, which
is around 4 degrees above the ambient temperature at this time.

Figure 15: Measured and estimated temperature, and current 6/11

The estimated temperature and measured temperature for the entire month of November is plotted in
Figure 16 to show the trends over a larger span of time. It can be clearly seen that the measured
temperature is consistently below the line temperature estimation. The estimation oscillates more than
the actual line temperature, since it is not an exact representation of the line temperature, but rather a
rough estimation. With that being said, it’s clear that the temperature measurements of the neurons are
not correct.

Figure 16: Measured and estimated temperature for November 2019

37
6.2.3 Angle of inclination
The neuron measures the angle of inclination at the fastening point in the span with a 3-axis accelerometer.
It returns an angle between 0◦ and 30◦ . Heimdall have not yet estimated the measurement accuracy, but
early projections suggest it is within 0.1◦ . The resolution is 0.01◦ . This angle is then used as an input
for a sag estimation model developed by Heimdall, which returns a digital representation of the line. The
sag, line profile, clearance to ground etc. can be ascertained from this model.

At the time of the project, Heimdall communicated that the neurons were in need of additional calibration,
and that the output data was therefore not accurate. It was planned to gather data to calibrate the
neurons, but due to unforeseen events it was not possible to carry out this task before the end of the
project. Due to this, the angle of inclination measurements may not be used in DLR estimation until
this issue has been resolved.

38
7 Economic analysis
An important aspect of DLR is to achieve some type of economic benefit for an investor. To determine
the economic potential of the implementation of DLR on VL-X, an economic analysis has been performed
which is presented in this section. Both the perspective of a grid owner and a windpower generator is
analysed, as both are relevant and also closely interconnected.

It’s been shown in Section 5.3.4 that DLR increases the transmission capacity of a line. However increasing
the transmission capacity has no value on it’s own, unless the load on the line is also subsequently
increased. Put simply, there’s no point increasing the capacity if you don’t use it. In this economic
analysis, the additional transmission capacity is used to connect and transmit additional windpower
in the area. Hence the economic benefits comes from increased windpower generation in the area. For
Ellevio, the economic benefit is rooted asset acquistions of infrastructure required for the grid connection.
For a windpower generator it is centered around being able to transmit more energy with the existing
infrastructure. It’s worth mentioning that DLR has other applications than those presented in this
economic analysis, and other ways to generate profit.

A script was made to calculate the relevant parameters to be able to make an economic analysis of DLR
profitability. It’s explained briefly in Section 7.1 and the code can be seen in Appendix A.4.

7.1 Power transmission calculations


To simulate the effect of increased windpower generation on VL-X, the load on VL-X from the windpower
generation is calculated. Once the impact of 1MW of windpower is calculated (in terms of current on
VL-X), said impact can be scaled to simulate N MW of added windpower. Measurements of windpower
output has been collected from a nearby wind park. To simplify the calculations, it is assumed that all
added windpower is located in the same wind park, and follows the same transmission patterns as the
already existing production (essentially the load is added to the existing load). The wind park where the
output power data is gathered has a nominal power of 68MW, hence the power output is divided by 68
to obtain the power output from 1MW of windpower, P .

Experienced network planners from the area has derived that 70% of the current from the wind park
passes through VL-X. This percentage shifts depending on the load flow in the area, but to simplify the
calculations it’s assumed to be constant. The operating voltage of VL-X is 135kV, and due to the 3-phase
transmission a factor of the square root of three must be considered. This yields (46), which shows the
increase in load on VL-X from N MW of additional windpower in the wind park.

N ·P ·η
I+ = √ [kA] (46)
3 · VV L−X

In (46), I + is the additional load on VL-X, N is the MW of windpower, P is the generation of 1MW
in the wind park at a certain time, η is the percentage of the wind park production that passes through
VL-X, VV L−X is the operating voltage of VL-X, which in this case is 135kV. With this equation, the load
on VL-X for different levels of windpower in the area can be calculated. Thusly, the load of the line for
every point in time for the duration of the windpower generation dataset can be calculated as a function
of added windpower N .

The transmission capacity is given by the DLR (or for comparison, the SLR). A script calculates the
difference between the DLR and the load on the line, for every time stamp of the available data. If
the load exceeds the transmission capacity, the windpower output must be curtailed. The percentage of
curtailment throughout the available data, as well as the ENS is then calculated. The latter is simply
the sum of the load that exceeds the transmission capacity, as seen in (47).
(
X 5 (I + I + − DLR) · VV L−X · 10−6 if I + I + ≥ DLR
EN S = 60 [M W ] (47)
0 if I + I + < DLR
5
Each data point in the input data corresponds to a 5 minute period. The 60 factor converts output to
MWh. It must be taken into consideration that the DLR and the load is given in Amperes, the voltage
in Volts, but the ENS in MW, hence the 10−6 factor converts the W output to MW. It must also be

39
taken into consideration that the ENS reliability index generally is calculated as a yearly average on the
form of kWh/year. The summation in (47) would give the total energy not served for the duration of
the available data. Hence the calculated ENS is divided by the amount of hours in the datasheet and
multiplied with 8760 so that it returns the ENS in the form of MWh/year.

In addition to the ENS, it’s of interest to know the opposite; the energy supplied. Or rather, the energy
that is produced at the wind park and then successfully transmitted without the need for curtailment.
This will for the remainder of this report have the notation ET , or Energy Transmitted. It’s calculated
in (48).
X 5
ET = ( · P · N ) − EN S [M W ] (48)
60

However for economic applications, it’s not primarily the transmitted energy that is of interest, but the
money earned selling the transmitted energy. This is also calculated in the script. It’s calculated by
calculating the transmitted energy for each data point, and then multiplying said energy with the current
Nordpool Elspot market price for SE3. The implementation in R is shown in Appendix A.4, however the
aforementioned information is sufficient to understand how the sold electricity parameter is calculated.
In the final step, the sold electricity (SE) parameter is divided with the number of hours of the datasheet
and multiplied with 8760, to yield the sold electricity per hour (SEK/year). This parameter can then be
used directly to calculate the income for a windpower generator.

7.2 Results from windpower script


The results from the wind power calculation script is presented in this section. In Figure 17 the load
on the line together with the SLR and DLR can be seen. In Figure 17 the windpower in the area is

Figure 17: Load on VL-X at 230MW of added windpower

298MW. This includes the 68MW that is already a part of the base current, and an added 230MW which
is added onto the base current to produce the blue line. The black line represents the base current and
the magenta is the DLR.

The two anomalies previously mentioned in Section 5.3.4 are present in this plot as well. The first anomaly
is due to the base load being ”fixated” at around 280A for an extended period of time while the neurons
are unoperational. This causes a likely overestimation of the load from added windpower which in turn
leads to an overestimation of the curtailment and ENS for both DLR and SLR. The second anomaly due
to the weather station blackout causes additional curtailment for DLR (if one assumes the DLR will be
higher than the SLR) Without said malfunction, the ENS of DLR would have been less than what is
estimated. However one may argue that this is not an overestimation, since equipment malfunctions are
part of reality, and such an event is being taken into account here.

40
The results from the calculation of the indices ENS, ET and SE can be seen in Table 8. There are two
main points of interest to analyse from the indices in Table 8. Firstly it’s the difference in SE at different
levels of windpower with and without DLR. This will portray the yearly difference in income from sold
electricity if DLR is implemented. This can be seen in Figure 18a.

ENS [MWh/year] ET [MWh/year] SE [MSEK/year]


Windpower [MW] DLR SLR DLR SLR DLR SLR
5 0 0 19092 19092 5.7 5.7
25 0 0 95459 95459 28.6 28.6
50 0 0 190918 190918 57.3 57.3
75 0 0 286376 286376 85.9 85.9
100 52.7 1289.9 381783 380546 114.5 114.0
125 405.5 5094.0 476889 472201 143.0 141.2
150 1926.3 12771.0 570827 559982 171.0 167.3
175 7009.6 28150.3 661203 640062 198.0 191.6
200 13911.3 48609.7 749760 715062 224.4 214.4
225 22890.8 72367.1 836239 786763 250.2 236.3
250 34535,5 98398.8 920054 856190 275.1 257.5
275 49604.6 126149.3 1000443 923899 299.1 278.1
300 68654.9 155283.2 1076848 990224 322.0 298.3

Table 8: ENS, ET and SE for different levels of added windpower

The second point of interest is the portion of produced energy by the wind park that is successfully
transmitted. This is closely connected to the SE (the relative differences in SE are due to different
amounts of transmitted energy depending on the rating). The quotient of transmitted energy can be
calculating using the wind energy indices as seen in (49). It will be referred to as relative transmitted
power.
ET
ET% = (49)
ET + EN S
The relative transmitted power can be seen in Figure 18b. Both figures in Figure 18 show how the impact
of DLR becomes more prominent as the line load increases from additional windpower.

The Pearson Correlation Coefficient is calculated for both types of DLR and the load from windpower
generation, and the original load profile (without additional windpower generation). It can be seen that
the DLR shows a stronger correlation to windpower generation than the original load profile.

Rating Correlation wind generation Correlation load profile


DLR 0.5310 -0.0236

Table 9: Correlation between ampacity from DLR and load profiles

7.3 Net Present Value analysis


A Net Present Value analysis has been undertaken to determine the economic incentives of deploying
DLR. It is based on the methods presented in [34]. Both the perspective of a windpower producer and a
DSO is considered.

7.3.1 Input Values


The parameters taken into consideration can be seen in Table 10.

41
(a) Relative SE between DLR and SLR as a function of (b) Relative transmitted power for DLR and SLR line
added windpower rating methods

Figure 18: Relative SE and relative transmitted power from windpower indices

Parameter Value
Turbine cost 9 MSEK/MW
Construction cost 0.54 MSEK/MW
Other cost 0.54 MSEK/MW
Vendor Commission 3%
Operation cost 4% of Turbine Cost
Equipment life span 25 years
Inflation rate 1.2%
Real discount interest rate 4%
Nominal discount interest rate 5.25%
Desired ROI 8%

Table 10: Parameters of the NPV analysis

In a report from Energimyndigheten, the turbine cost is estimated to 9.3 MSEK/MW [35]. However this
report is from 2016 and technological progress has been made since then. Furthermore, representatives
from Rabbalshede Power have given figures in the vicinity of 48 MSEK for 5-7MW (roughly estimated).
This gives a span of 6.9-9.6 MSEK/MW. With this in mind, 9 MSEK/MW was determined to be suffi-
ciently accurate and conservative for the estimation. Vendor commission, equipment life span and desired
ROI are also based on figures provided by Rabbalshede Power.

The inflation rate for the past 10 years in Sweden has been 1.1% [36]. Hence an inflation rate of 1.2%
was chosen as a conservative estimate. The real discount interest rate is more difficult to estimate, since
it is set by companies on an individual basis. In both [37] and [38] this is set to 4%. Hence a 4% discount
interest rate is chosen with reservations that these differ from company to company.

The nominal discount interest rate is calculated with (50), where Zn is the nominal discount interest
rate, Zr is the real interest rate and Zi the rate of inflation. The report from Energimyndigheten uses
interest rates between 5-8%, and the nominal discount rate used in this report is within that span [35].

(1 + Zn ) = (1 + Zr )(1 + Zi ) (50)

42
The figures from Rabbalshede Power regarding investment costs states that a rough estimate is 63 MSEK
for a 5-7MW unit including installation, construction, grid connection etc. 48 MSEK of this is the turbine
cost, whereas the additional 15 MSEK is attributed to the additional costs. Hence 76% of the cost is pure
turbine cost. A report from IRENA suggests that distribution of investment costs are as seen in Table
11 [39]. It can be seen in 11 that a 76% is within the span presented by IRENA and IEA-ETSAP. This

Cost share %
Wind turbine 64-84
Grid connection 9-14
Construction 4-10
Other capital 4-10

Table 11: Distribution of investment costs for onshore wind farms

leaves 24% distributed over grid connection, construction and other. In this project, 12% is attributed
to the grid connection as it is in the middle of the span. That leaves 6% for both construction and other
capital. The grid connection cost is omitted in Table 10, as it is not an input variable in the analysis.
Instead, the grid connection cost (or rather the possible grid connection cost) is the desired output from
the analysis.

The maintenance cost was set to 4% as it was also set to 4% in [38]. With the aforementioned param-
eters this is 360 kSEK/MW,year. This is calculated differently in [35], where it is assumed to be 148
SEK/MWh. This yielded costs of 565 kSEK/MW,year for the given data. It’s difficult to estimate which
of these estimates is the most accurate, and ideally data from windpower producers in the area should be
used for these calculations. However as this is not available for this project, 360 kSEK/MW,year is chosen
as the maintenance cost. It should then be taken into account that this seems to be a comparatively
optimistic estimate. Furthermore the operational costs are assumed to be linear for this project. This is a
simplification, however it is outside the scope of this project to determine a more advanced maintenance
scheme, and it would also make the results a lot more complicated to understand.

Another note regarding the input values is the lack of a residual cost. Neither a positive nor a negative
residual cost has been used for the analysis. The accuracy of this estimation is yet to be determined.
As of right now, no reliable data on the residual cost of windpower equipment in the area is available.
This is quite difficult to estimate, and furthermore it is irrelevant for the focus of this analysis. Including
a residual value would increase or decrease the NPV equally for both DLR and SLR, and since this
economic analysis is based on finding the differences between them it is not a necessary point of focus.
With that being said, a Master Thesis from 2015 on this topic suggests that the decommissioning cost
will be higher than the residual value at the end of the equipment lifetime [40]. This suggests that the
decision to omit a residual value from the calculations is optimistic, and the actual NPV is probably
belowed the estimated value.

7.3.2 Analysis method


As mentioned previously in Section 7.3, the method used for the economic analysis is based on the method
presented in [34]. It calculates the NPV based on incomes, costs, the nominal discount interest rate and
economic lifetime of the equipment. The investment cost is calculated according to (51).
ΓIC (N ) = (γT C + γCC + γOC ) · N (51)
ΓIC is the Investment Cost, γT C is the Turbine Cost, γCC is the Construction Cost, γOC is Other Cost
and N is the number of MW. As mentioned in Section 7.3.1, the grid connection cost is omitted as an
input value, as it is in fact an output value to be determined by the analysis.

The NPV is calculated in (52) where y is the number of years from present, k is the lifetime of the
equipment, SE is the value of sold electricity for a year as shown in Table 8. γV C is the vendor commission
and Λ is the Operational Cost. The Operational Cost is 4% of the Turbine Cost as shown in (53).
k
X
CN P V (N ) = (Zn−y · SE(N )(1 − γV C ) − Λ(N )) − ΓIC (N ) (52)
y=0

43
Λ(N ) = 0.04 · γT C · N (53)

The NPV CN P V (N ) is then used for two things. It is used to calculate the ROI directly (without taking
grid connection cost into account) which is of interest to the windpower producer. This calculation can
be seen in (54).
CN P V (N )
ROI(N ) = (54)
ΓIC (N )
It is also used to calculate the residual NPV once the desired ROI has been returned to investors. This
calculation has several steps shown in (55).

CN P V (N )
ROI ∗ = =⇒
ΓIC (N ) + Cres
CN P V (N ) = ROI ∗ · ΓIC (N ) + ROI ∗ · Cres =⇒ (55)

CN P V (N ) − ROI ∗ · ΓIC (N )
Cres =
ROI ∗
ROI ∗ is the desired ROI of 8%. Cres is the residual net present value when a cut of 8% has been taken
from the net present value. Or if one wishes to turn it around, it is the limit that can be spent on
connection costs while still maintaining a ROI of ROI ∗ .

This method of calculation is chosen for two reasons. Firstly the connection costs vary from project to
project, and it seems useful for all parties involved to have a numerical limit for the connection costs where
a project is no longer economically feasible. The second reason is based on the economical prerequisites
of a DSO regarding wind farm connections. An interview with the Head of Connections Industry &
Production at Ellevio AB, Per Selldén forms the basis of said prerequisites:

When a windpower producer connects to the grid (or upgrades an existing connection) they are responsible
for the cost of said connection. However the electrical infrastructure built for said connection (lines, cables,
substations) becomes part of the DSO’s assets. So if for example a 50 MSEK investment is required to
build new lines to connect a wind park, the windpower producer would be responsible for the cost of
50 MSEK, whereas the DSO would gain 50 MSEK (or in the vicinity thereof) in assets. Hence it is
obvious that connecting windpower to the grid is a very good deal for the DSO. The limiting factor is
the economic viability for the windpower producer, which is why this analysis is centered around finding
this limit, and also the margin above said limit.

In summary, the economic analysis yields three results. The relative ROI, the absolute ROI, and the net
present value above 8% of the investment costs. These results are a function of added windpower in the
area. The relative ROI, is the percentage of the investment that is earned in profit over the lifetime of
the equipment. The absolute ROI is the sum of the cash flow throughout the life time subtracted by the
investment cost, and the net present value above 8% of the investment costs (called residual NPV in this
report) is a combination of the two.

7.3.3 NPV analysis results


The ROI as a function of added windpower is shown in Figure 19. It can be seen that the relative ROI
is the same for both rating schemes until the 85MW mark when the ROI for the SLR starts to drop. At
this level, the power is being curtailed as the load on VL-X occasionally reaches the static limit. The
relative ROI for DLR also begins to drop off as slight curtailment occurs from the 85MW mark, but the
effect is not significant until around 130-140MW.

The absolute ROI rises linearilly until it begins to drop off as the effects of curtailment becomes increas-
ingly severe. For SLR, the absolute ROI peaks at 96 MSEK which occurs at around 120MW of added
windpower. At this level of windpower, the relative ROI is around 8%. The same results for DLR are
140 MSEK at around 130MW, for which the relative ROI is around 9.2%.

The residual net present value can be seen in Figure 20. For analysis of the latter it is of interest to
determine the location and magnitude of the peaks for each respective method. For DLR it’s 18.8 MSEK
at 130MW and for SLR it’s 14.7 MSEK at 85MW.

44
Figure 19: ROI as a function of windpower for DLR and SLR

Figure 20: Residual NPV as a function of windpower for DLR and SLR

45
8 Identification of critical spans
In order to accurately and safely provide a dynamic rating for a line, the locations along the line where
monitoring equipment should be placed must be chosen with great care. The line temperature along a
line may vary, as transmission lines are poor temperature equalisers, and while the load is often uniform
along the line, the weather conditions are not. Air temperature and solar heating are fairly uniform along
a line, however wind speed and wind direction are highly variable [41].

Cigre TB498 states that the data to be checked for line rating should be the worst case between maximum
temperature and maximum sag. This is referred to as a critical span. In real applications, the as-
design relationship between these two variables is no longer valid, due to the complexity of in-field
conditions. Furthermore the location of this span varies, depending on current weather conditions. Hence
the challenge of finding critical spans along the line is centered around finding the spans that are most
likely to have the highest line temperature, the lowest distance-to-ground, or both [41].

A method on how to find critical spans has been developed with the help of literature study of previous
methods, and discussions with experienced network planners. It has then been applied in practice to find
critical spans along VL-X.

8.1 Method for finding critical spans


In [42], a novel heuristic for identifying the number and location of critical monitoring spans is put forth.
It’s based on a Mesoscale Weather Model, and a statistical analysis of the thermal capacities in each
span, along a 325km line in northern Chile. The optimal monitoring solution is then determined, and
compared to an equidistant monitoring strategy.

The authors in [42] acknowledge that wind speed has the greatest influence on the thermal capacity of
the line. The study in [42] uses a WRF-model to model the wind speed along the line to identify the areas
with the lowest wind speed. In addition to wind, temperature is also modeled. However it is mentioned
that the WRF model can not account for local site characteristics, such as vegetation and obstacles.

For a 325km long line, it may be necessary to use a method such as the one presented in [42], however for
a shorter regional line such as the one in this project, it should be possible to take the aeolian variations
of microclimates along the line into consideration. In a study from 1988, the author mentions that the
placement of the monitoring equipment was determined from span orientation and wind sheltering. The
first of these factors is dependent on the average wind direction in the area, whereas the second takes
microclimates along the line into consideration [43].

A study in Northern Ireland chose to monitor points in high, medium and low wind areas to give a
representation of system behaviour at unmeasured spans. The majority of the measurement points were
placed in low wind/sheltered areas [44].

In [45], the authors reference the method presented in [42] and point out a weakness; the method only
takes the thermal capacity of the line into consideration, and does not consider the sagging level. Hence
the method in [45] employs a sag model for the conductor, and suggests placement of the monitoring
equipment based on distance-to-ground limits. This model showed that the heuristic presented in [42] re-
quired less monitoring devices, but also underestimated the sag several times. The methodology presented
in [45] did not exceed sag limitations at any time during the analysed period.

The authors in [46] critizes the methodology used in [45] for not considering the span-topography and
terrain levels. They state that since these factors are not considered, the actual sag across the line may
differ from the estimations, and the distance-to-ground estimation is therefore unreliable. Instead, a
methodology that includes the surrounding terrain, as well as the sag and geo-spatial weather conditions
is proposed.

The methods employed in [42], [43], [44], [45], [46] considers one or more of the following factors:

• Macroclimatic conditions
The macroclimatic conditions are the macro-scale weather conditions around a line. The primary
variables of interest are wind speed and wind direction, whereas the secondary variable of interest

46
is ambient temperature. Areas that are subjected to lower wind speeds, wind directions parallel
to the line and higher temperatures will have a higher local line temperature than other areas.
Analysis of historical weather data can determine the location of these areas.

• Microclimatic conditions
The microclimatic conditions are the local weather conditions around the line caused primarily
by wind shading. This can be caused by vegetation around the line or terrain that shelters the
line from wind. This factor is difficult to assess since detailed knowledge about the surrounding
vegetation around the line is rare. Knowledge of the surrounding terrain is easier to estimate, but
the effect it will have on the wind incidence is still complicated and difficult to predict. However
both [4] and [46] emphasize the importance of this factor. Either expert knowledge from network
planners can be used to identify critical spots along the line, or additional safety margins must be
implemented to account for this factor.

• Sag conditions
The sag is closely connected to the distance-to-ground which is the primary limiting factor for line
loading. By deriving the placement of monitoring devices from sag limits instead of temperature
limits, situations where sag and distance-to-ground limits are exceeded are minimized or removed
completely, as argued in [45]. In essence a factor of uncertainty is removed, and the factor most
relevant to ampacity rating limits is considered directly in the monitoring device placement.

There is another advantage to considering the sag conditions for monitoring device placement.
Regardless of the method used for DLR calculations, there will be a degree of uncertainty that
increases with distance from the monitoring devices. Furthermore along a line, the distance-to-
ground and hence the allowable sag will vary. By placing monitoring devices in the spans that are
closest to these limits when the line is at its maximum allowable temperature, one can avoid distance-
to-ground infringements due to estimation uncertainty in the input data for DLR calculations.

8.2 Proposed methodology for identifying critical spans


In this section, a methodology to determine the critical points along a line is proposed. The methodology
is based on previous work discussed in Section 8, and its applicability to shorter regional lines as the one
in this project. The methods discussed in Section 8 all propose a general method that place monitoring
devices based on a statistical confidence measure. Furthermore they all share the same weakness, which
is that they consider a partial picture of the surroundings. In the proposed methodology, a more detailed
approach is suggested. The factors considered in [42]-[46] are used to form a basis, but ultimately the
placement of the monitoring equipment is up to the discretion of the network planner. This methodology
can be considered a risk analysis to determine the most vulnerable locations along a line, rather than a
placement strategy. This is more time consuming than the methods proposed in [42]-[46], but compensates
by utilising the expert knowledge of network planners, considering all 3 main factors and having manual
decisions as the ultimate decider. This approach is backed up by Cigre TB498 which underlines that a
global solution to the number of critical spans and thus number of monitoring devices to install is very
difficult to estimate, and that the answer must be adapted to each line separately [41].

The methodology produces a list of the most vulnerable locations, and it is then up to the network
planner to determine what level of security to strive for. The quality of the identification of critical spans
depends entirely on the quality of the available data.

The steps of the methodology is presented below.

8.2.1 Macroclimatic analysis


In this step the macroclimatic conditions around the line are considered. The factors of interest are
primarily wind speed and direction, and the secondarily temperature. MESAN-models and similar,
newer versions are suitable to model the climate along the entire line, and not just around adjacent
weather stations. CFD is also a possible alternative for macroclimatic modeling.

Once historical macroclimatic conditions have been estimated along the line, they can be used to calculate
the DLR as a function of time for the different sections of the line. It’s suggested that the line is divided
into sections based on the square-grid provided by the Mesoscale-model. The areas can then be arranged

47
from most to least vulnerable from a macroclimatic perspective, by determining the location and duration
of the minimum dynamic rating as a function of time, compared with the average dynamic rating during
said time.

The most vulnerable area can be considered to have an elevated macroclimatic risk factor. The relative
risk difference between sections should be determined in relation to the relative difference of the line
rating, as well as the duration of said difference. For example, a section that is rated 20% below the
average DLR for 50% of the time will have a higher risk factor than a section that is 5% lower 25% of
the time. It is up to the discretion of the network planner to quantify the weight of this risk factor, as it
depends heavily on the line conditions. Long lines through a very dynamic climate are more vulnerable
for macroclimatic variability than short lines through stable climates, hence the risk must be managed
differently on a case to case basis.

8.2.2 Microclimatic analysis


The microclimatic analysis estimates the impact of wind shading from terrain and vegetation around the
line. It is stated in [4] that this is a major factor of uncertainty in DLR estimation, hence by placing
monitoring equipment in the most shaded areas the risk of overestimating the dynamic capacity of the
line is minimized. This factor is very difficult to estimate, since it’s rare to have detailed knowledge about
the vegetation along an entire line. In the situations where this is not available, this part of the analysis
will have to be omitted. In these cases the wind cooling factor would instead have to be multiplied with
some type of shading factor. Furthermore as the uncertainty of the DLR estimation increases, rigorous
safety margins would have to be implemented.

When data of the vegetation and terrain around the line is available, it should be used to estimate a risk
factor from the wind shading. Through assigning values corresponding to different levels of shading to
the line sections, the most and least vulnerable areas from a microclimatic perspective can be determined.
The magnitude of this risk factor is yet again left for the network planner to determine, as it depends on
the specific situation. Areas that generally see a large amount of wind are naturally more vulnerable to
differences in local capacity due to shading, as the convective cooling accounts for a larger amount of the
heat transfer than for areas that see little wind. Hence this risk factor need to be determined based on
the type of vegetation and prevalence of wind in the area around the line.

8.2.3 Distance-to-ground analysis


As mentioned in Section 8, there is an inherent degree of uncertainty in all DLR estimations. This
uncertainty increases with the distance from the monitoring equipment. Furthermore as the distance-to-
ground varies along the line, certain sections will be closer to regulatory limitations than others. It is
imperative that the accuracy of the DLR estimations is as high as possible in these sections, hence the
distance-to-ground can be considered a risk factor to consider for the placement of monitoring equipment.
It is the distance-to-ground when the line is at its thermal limit that is to be considered.

In the proposed method, the risk is estimated on a 0-10 scale, where 0 corresponds to the lowest risk and
10 the highest. The risk factor is calculated as the quotient between the safety limit distance according
to [30], and a measurement uncertainty and the measured distance to ground at the temperature limit of
the line. The measurement uncertainty corresponds to the uncertainty of the method used to determine
the distance-to-ground, and the DLR temperature modeling uncertainty corresponds to the estimated
uncertainty of the DLR method. This uncertainty is 5 K in [16], and 3 K in [4]. The variable to consider
from this is the difference in sag from said uncertainty, i.e. how much more the line will sag if the
temperature increases by 3 or 5 K. The measurement uncertainty and the change in sag from the DLR
temperature modeling uncertainty are added together to form the measurement uncertainty variable Υ.
The calculation of the unweighed distance to ground factor is shown in (56) where dtglimit is the safety
limit for the distance to ground (typically 6.4m for unpopulated areas), Υ is the measurement uncertainty
and dtg50◦ C is the measured distance to ground at 50◦ C.

dtglimit + Υ
riskdtg = (56)
dtg50◦

A weight factor is then calculated by dividing 10 with the maximum value for unweighed risk, so that

48
when said value is multiplied with the weight factor it will become the highest value possible in the scale
(i.e. 10). All other risk values are weighed according to the same factor, which provides a continuous
relative estimation for each span along the line.

8.2.4 General guidelines


An analysis of the aforementioned factors will provide a vulnerability assessment for each span along a
line. It is then up to the network planner to use this information to place the monitoring devices in
locations that provide a safe worst-case estimation of the line. However there are additional factors that
should be considered for the placement of monitoring equipment. The monitoring equipment should also
be placed so that:

• The entire geographical area is covered

• The different directions of the line is covered

The reason for this is the same in both cases. There may be a section of the line that is subject to
adverse weather conditions for most of time. However if all monitoring devices are placed in this area,
and none in the low-risk area the line capacity will be overestimated during the rare weather conditions
when the cooling conditions are more favourable in the high-risk areas. Hence in order to ensure safe
operation for all weather conditions, the aforementioned factors must be considered as well. This means
that the network planner may be forced to place a monitoring device in a medium-risk area rather than
a high-risk area to get a better geographical spread.

8.3 Application of methodology on VL-X


As part of the project, the methodology presented in the previous section is applied on the line VL-X.
This section contains explanations of if and how analysis of different kinds was performed and what
results were found. There is also several suggestions for future work within this topic in section 11. The
R-script used to calculate the risk factor for each span can be seen in Appendix A.5.

For the macroclimatic analysis, it must be taken into consideration that the line is 24km long. The grid
size of MESAN-data for the available area is 11km by 11km, which would mean that only 3-4 weather
zones can be analysed. The difference between adjacent zones is likely to be negligible, and furthermore
even if it is not, monitoring equipment should be placed with geographical spread, i.e. in both zones. Even
though the differences in weather are spread over a continuous area, the data used for a macroclimatic
analysis is discrete. Hence it’s acknowledged that for this line, the macroclimatic analysis will not affect
the placement of monitoring devices, and can therefore be omitted.

The microclimatic condition have been modeled with the help of laser and ortho-imaging of the line.
The images are taken by a helicopter that flies along the line and generates an image of the line and its
surroundings. An example of an ortho-photo of the line and its surroundings can be seen in Figure 21.
The main purpose of the data is to estimate the vegetation around and under the line to determine if
and when maintenance is required to clear the line corridor, but it can also be used to estimate other
risk factors around the line.

These images are available for the entire length of the line. It can be seen that an area has been marked
in green. This area contains vegetation that is high enough to shade the line (or strike the line if it were
to fall). The white vertical line represents the separation of spans, and the three horizontal lines are the
conductors of the line.

By analysing these images, the degree of shading around the line has been estimated. The shading has
been graded on a scale of 0-5 where 0 represents 0% of shading, 1 represents 1%-20% of shading, and 5
represents 100% shading. This grading is done for each side of the line. For example, the grading for
the span seen in Figure 21 would be 3 on the left side, and 0 on the right side. In the total grading, this
would represent a 3/10. This grading is later used for the total risk analysis. Hence, with the help of
these images a risk estimation for the microclimate on a 0-10 scale for each span along the line.

The distance-to-ground and temperature of the line is measured through laser scanning and thermal
imaging. The position of the line and thus the distance-to-ground at the thermal limit can then be

49
Figure 21: Orto-photo of VL-X

determined.

The measurement uncertainty of the distance-to-ground of the line is based partially on the precision
of the laser scanning and thermal imaging, and partially on the precision of the DLR-estimation. By
adding these factors, a distance corresponding to the uncertainty is found. For this analysis, a precision
of 3 K was assumed for the DLR estimation. Ideally, this should be calculated through comparison with
measurements of line temperatures, however since that is not available the uncertainty from [4] is used.
How much the sag differs from a temperature change of 3 K depends on the length of each span. Elisabeth
Lindberg has found this change to be 3.7cm per degree [31], and Cigre TB498 found this relation to be
3.6cm per degree [41]. However for the sake of this rough risk assessment, it should be sufficient to take
a conservative approach and say that the sag changes with 4cm per degree, which would lead to a 12cm
uncertainty from the 3 K uncertainty in the DLR estimation. The measurement uncertainty from the
line modeling is even more difficult to estimate as both the uncertainty of the thermal imaging, and the
line position imaging must be taken into account. Daniel Öhman, CTO of Visimind who supply the
data claims that the measurement uncertainty from the line position is at most 25 cm (depending on the
vegetation below the line). The thermal uncertainty is even harder to estimate as the line is very thin.
Furthermore a high degree of absorptivity makes the line appear colder than the surroundings, which is
clearly inaccurate. When this is the case, the line temperature is estimated as the ambient temperature.
Hence the thermal uncertainty depends on the load of the line (as a high load would mean an elevated
line temperature). For this project, this has been estimated as 5 K. This thermal uncertainty can be
minimised if line temperature measurement equipment is installed on the line.

As there is not yet sufficiently detailed knowledge about the influence of the different risk factors, they
are evenly weighed when calculating the total risk factor for each span along the line. The total risk
factor is calculated as the sum of the microclimatic and distance-to-ground risk factors divided by 2,
which yields an estimation of risk between 0-10 for each span along the line.
riskmicro + riskdtg
risktotal = (57)
2
The total estimated risk along the entire line can be seen in Figure 22. It can be seen that the most
vulnerable span is span 9, where the risk is estimated to a 9/10. This span has a large degree of shading
and the lowest distance-to-ground clearance of the entire line. The span can be seen in Figure 23. The
naming convention of spans is based on the number of the left pole in the span, so span 9 corresponds
to the span between pole 9 and 10. The complete set of total risk estimates for the entire line can be
found in Table 12. It can then be used by the network planner as a support when placing monitoring
equipment along the line. This is a rough, yet useful visualisation of the most vulnerable areas along the
line. The methodology and especially the risk estimation can be improved with additional future work.
What could be done; and how, is discussed further in Section 11.

50
Figure 22: Estimation of total risk along the entire line

Figure 23: The most vulnerable span along the line

Span Total estimated risk


9 10-9
4 14 29 38 81 84 9-8
6 7 10 22 24 26 31 35 37 53 55 73 85 91 92 104 111 8-7
8 25 30 32 36 66 70 74 80 82 83 90 110 7-6
5 19 20 21 23 27 41 47 57 65 68 77 103 109 6-5
15 17 28 33 34 42 45 50 54 71 72 79 86 93 94 99 112 118 119 121 5-4
1 2 18 39 40 43 46 48 52 56 60 61 100 102 105 106 114 117 120 123 124 4-3
3 12 51 59 63 64 67 69 75 76 96 97 98 101 108 113 115 122 3-2
11 13 49 62 87 116 2-1
16 44 58 78 88 89 95 107 1-0

Table 12: Total estimated risk for each span along the line

51
9 Discussion
The first task of the project was to determine which measurement data could be trusted. The results
in Section 6 show that the measurements from the weather station seem to be functioning correctly,
except for an anomaly in the radiative cooling analysis in Figure 5. A dip in the radiative cooling after
approximately 2 hours into the data show that something is out of order. Analysis of the input data show
anomalies in the ambient temperature data. Before (and after) the anomaly, the ambient temperature
is around -0.7 ◦ C. At the point of the anomaly, the temperature is registered as 5.625 ◦ C. In addition
to the temperature anomaly, the pressure measurement (which is constantly 1008 hPa for the rest of the
data) registers 1012 hPa. Therefore it is reasonable to believe that there is a malfunction in the ambient
temperature data collection at this time, that also affects the pressure measurements. The reason behind
this is unclear at this time. Furthermore the anomaly is rare enough to have no effect on the DLR,
especially since it is averaged over a 15 minute sliding window. Therefore the weather station can be
considered to provide sufficiently accurate measurements for DLR applications

It is shown in Section 6.2.2 that the temperature measurements of the Heimdall Neuron are not working
correctly. This information was passed on to Heimdall who performed tests on the Neuron. It was
confirmed by Heimdall that the temperature measurements were indeed not working correctly, and work
is currently in progress for a fix of this issue for the next version of the Neurons. Furthermore, Heimdall
communicated that the angle of inclination measurements were in need of calibration. This calibration
was not completed before the end of the project, hence both the temperature measurements and angle
of inclination measurements were not available for calculation or verification of the DLR.

The results in Section 6, limits the DLR method used in Section 5 to a weather based DLR. The plots of
the thermal components in this section show the nature and importance of each component. It can be
seen that the convective cooling is the most important, and also the most variable component. Second
in importance is the radiative cooling, and last and least is the solar heating. One may think that
the IEEE-738 method of estimating the global radiation is superior to the Cigre-601 method of using
measurements, as it yields a conservative estimation and requires less measurement equipment (i.e. it is
cheaper to implement). However this depends on the DLR application. If temperature measurements
of the line are used to calculate the effective wind speed for the line, then said wind speed will depend
on an equilibrium in the heat balance equation. If there is an overestimation on the heating side (from
global radiation as seen in Figure 4 and Figure 13), this will lead to a subsequent overestimation on
the cooling side (effective wind speed). As this is scaled to the temperature limit of the line, the effect
from this overestimation will increase for the convective cooling as it is temperature dependent, whereas
it will not for the solar heating, as it is not temperature dependent. Hence the cooling of the line will
be overestimated, which will in turn lead to an overestimation of the DLR. This can lead to dangerous
operating conditions if this is not taken into consideration in the DLR system design [41]. It is difficult to
say that one method of solar heating estimation is objectively better than the other. Rather, the method
must be chosen with consideration of the actual DLR system it is meant to operate in.

The DLR is plotted in Figure 7 and Figure 8. The plots show several things. Firstly both the IEEE-
738 and Cigre-601 method yield equivalent results. Secondly it can be seen that DLR yields a capacity
increase of around 23% for the measurement period. For the weekly plot in Figure 7, it can be seen that
the IEEE-738 rating drops below the Cigre-601 rating slightly at the same time each day. This is due to
the differences in solar heating estimation. The IEEE-738 method overestimates the solar heating which
leads to an underestimation in the overall DLR. The periodicity where the IEEE-738 rating drops below
the Cigre-601 rating matches with the periodicity of the solar heating parameter (for when it is nonzero).

Although the increase in capacity is 23%, this does not show the full picture regarding the benefits
of DLR. As the capacity increase is intermittent, it matters not only how much, but also when this
capacity increase occurs. In Section 7, the synergy between windpower generation in the vicinity and
the transmission capacity of the line is analysed. The hypothetical load on VL-X at 230MW of added
windpower in the area is shown in Figure 17. It can be seen that the peaks often line up, which suggests
synergy between windpower generation in the vicinity and DLR on VL-X. Despite this synergy, the
aforementioned capacity of 23% does not equal a corresponding potential increase in actual transmitted
power. How much of the capacity increase actually amounts to transmitted energy depends on the
strength of the correlation between the DLR and the load profile.

The correlation between DLR type and load is shown in Table 9. These results show a correlation

52
of around 0.5 towards the windpower generation, and no correlation to the original load profile. The
correlation between the load profile and the DLR is a vital piece of information regarding the viability of
DLR implementation. If the DLR were to show a weak or nonexistent correlation towards the load profile
on the line, little of the additional capacity would be utilised. For the Pearson Correlation Coefficient,
a 0 corresponds to no correlation. This means that there is no discernible correlation, i.e. that the load
and DLR are operating independently. This is not great but also not terrible. A worse alternative would
be a strong negative correlation. That would imply the DLR is falling when the load is rising. This
would mean the capacity of the line is high when the load is low and vice versa. For such a situation, the
DLR would serve no purpose. In this project, there is a strong correlation between DLR and windpower
generation and no correlation beteen DLR and the load profile. Thus, it can be said that DLR shows
great potential for increasing windpower transmission capacity, and moderate potential for increasing the
transmission capacity of the predominant load profile.

The economic analysis in Section 7 is undertaken to quantify the benefits of DLR implementation in terms
of monetary parameters. However the potential benefit depends on which parameter in the NPV analysis
results is being considered. It’s clearly shown in Figure 19 that DLR increases how much windpower can
be added to the system while maintaining an acceptable profitability. It’s also a reasonable assumption
to assume that an investor wants both a high relative and and absolute ROI. However it’s difficult to
generalise which of these metrics should be prioritized above the other.

If the highest possible relative ROI is the primary goal, then windpower investments of up to 85MW
fullfill that goal if SLR is applied, with the end point also having the highest absolute ROI within the
acceptable span. Hence 85MW can be considered the optimal investment for SLR. The corresponding
optimal point if DLR is applied is 130MW (as the effect of curtailment is not very severe up to this
point).

If the highest possible absolute ROI is the primary goal, the maximum point is located at 120MW for
SLR. However this is a terrible investment. It’s shown in Figure 19 that the absolute ROI rises very slowly
after 95MW. This means that a linearilly increasing investment cost is required, but the absolute ROI
is virtually remains constant. Hence the highest reasonable investment (if absolute ROI is the primary
goal) is 95MW for SLR. At this level the relative ROI is around 9%. If DLR is applied, this optimal
point is around 160MW. For the same reasons as the SLR case, this is a poor investment, and anything
above 140MW is essentially increasing the investment with barely any additional NPV.

As seen above, there are many ways to look at the results. This complicates a simple answer to the
question of the value of DLR. A correct answer to this question would be that it depends on what the
goal of the investment is. Depending on the goal, DLR may be a useful tool to increase the profitability
of the investment, however it is not guaranteed. For all investments below 85MW, it’s shown that DLR
is not required. The same ROI can be achieved with a SLR. At this level, the residual NPV seen in
Figure 20 is 14.7 MSEK for both DLR and SLR. For all investments above this, all analysis metrics are
higher for DLR than SLR. The peak of the residual NPV for DLR is 18.8 MSEK located at 130MW. For
all power levels above 130MW, investments are becoming increasingly less sound. It’s clear that DLR
allows the producer to connect more windpower, and the peak to peak difference in residual NPV for
the optimal investment points for SLR and DLR is 4.1 MSEK. Hence this can be argued to be the value
of DLR for both windpower producers and grid owners (due to how closely connected the profitability
between producers and grid owners is).

One could argue that the maximum acceptable cost of DLR is 18.8 MSEK if 130MW is added. It’s
important to keep in mind that the acceptable cost is in fact a function of added windpower shown in
Figure 20. For example at 100MW a DLR system would not be allowed to cost more than 16.9 MSEK if
a relative ROI of 8% is to be upheld. Furthermore if the DLR system for a 100MW implementation were
to cost more than 2.2 MSEK, it would be a better investment for a windpower producer to add 85MW
windpower and use SLR.

It’s clear that the results from the economic analysis can show many different things depending on the
investment goals. Hence the 4.1 MSEK value is just one way of looking at the potential value of DLR.
Furthermore there are many other scenarios and applications than the one in this economic analysis where
DLR can be useful and profitable, that has not been covered by this economic analysis. This should be
taken into consideration when evaluating the results from this analysis.

53
There are a few things that play a part in the interpretation of the results. One is that the capacity
analysis has been made from a single point in the system, i.e. VL-X. That VL-X gains the capacity for
130 MW windpower does not necessarily mean that the entire system does. It’s possible that another
part of the system becomes a bottleneck at less than 130 MW, hence this analysis does not give the full
picture. If another bottleneck would limit the system before 85 MW, the DLR on VL-X would be in vain.
This must be kept in mind when looking at the results from this analysis.

Another thing to consider is the season which the data for the analysis has been gathered. The analysis is
based on around 3 months worth of data from autumn and winter. In [3] it was shown that the capacity
increase was greater for the summer than winter, hence the benefit of DLR is expected to increase when
summer ratings are taken into consideration. The economic analysis spans the equipment life span of 25
years. It must be kept in mind that ratings and cash flow for 25 years are projected from 3 months worth
of data. To obtain a higher confidence in the results, this analysis should be done again after data for at
least a year has been gathered, to account for seasonal variability in both ratings and weather conditions.
This is suggested as future work for the project.

The method developed in the identification of critical spans is very rough, and many points of improvement
have been identified and can be found in Section 11. With that being said, it successfully provides the
network planner with a rough picture of which spans are the most likely to exceed the thermal and
distance-to-ground limits. However, additional data is required to accurately estimate the effectiveness
of the proposed method. Once temperature and sag measurements are available for the line, they will
prove very useful not only for the verification of all conclusions based on the DLR, but also the method
for identfying critical spans.

54
10 Conclusions
The validation of the measurement devices in this project have shown that the weather station is working
well enough to facilitate weather based DLR, whereas the line temperature measurements have been con-
firmed to be malfunctioning. Through the analysis performed in this project, the problem was discovered
and Heimdall are currently working on a fix for this issue.

The DLR analysis showed that DLR results in a capacity increase of 23% for the analysed time period.
Furthermore an analysis into the transmission capacity given an increased load due to additional wind
power in the area showed the importance of also estimating the correlation between the DLR and the
load, as this has an impact on the effective transmission capacity. The correlation was shown to be higher
for DLR and wind power generation than the general load profile of the line.

For the completion of the project, a series of scripts have been developed in R to be used for the different
types of analysis present in the project. These scripts are scalable, and can be used to analyse the pilot
project for its entire duration.

The economic analysis was able to quantify the benefit of DLR for different actors. This analysis suggests
a potential value of DLR as 4.1 MSEK, but there are other points of view that would result in a different
value. Furthermore, the optimal level of investment is determined as 85MW for SLR and 130 for DLR.

A new method to identify critical spans for placement of measurement devices has been developed through
a literature study. This method was applied to the line in the pilot project, and a risk estimation for
each span along the line was obtained. This estimation will be used at a later stage in the pilot project,
as a basis for the placement of additional measurement equipment along the line.

In conclusion the project answered many questions regarding the prerequisites for DLR both in a practical
and economical sense. The work done in this project has aided both Ellevio AB and Heimdall Power in
their analysis of the pilot project, and will continue to do so in the future. However when looking at the
conclusions drawn in this project, it’s important to remember that they are drawn from a relatively small
amount of data, and that the modelling accuracy has not yet been verified with direct line measurements.
An analysis using the methods (and scripts) in this project must be done with at least a year’s worth
of data, and direct line measurements before the conclusions can be confirmed with a high degree of
confidence.

55
11 Future work
The current method for estimating the shading of a line due to the surrounding microclimate is quite
basic and rough. It could be developed further in a number of areas which should increase the accuracy
of the total risk estimation altogether.

Another aspect of the microclimatic analysis which can be improved is the correlation between the side
of the shading and the predominant wind direction. Weighing the shading risk values based on how large
portion of wind that side sees would improve the risk analysis.

A third point of improvement for the microclimatic analysis would be to increase the resolution of the
current 0-5 scale to a continuous scale based on the exact number of shading points per span. This data
currently exists and can be observed in GIS-programs.

Direct measurements of line temperature should be used to determine the accuracy of the DLR estimation,
and also find safety margins for operation based on this accuracy. Line temperature measurements can
also be used for DLR calculations directly, which should also be investigated further.

The confidence in the measurement data and the results from said data increases with the amount of
available data. To increase confidence in the results and conclusions drawn from them, the DLR and
economic analysis should be done in one year’s time, and the results should be compared.

56
A Appendix
A.1 Main script
#Main DLR−c a l c u l a t i o n s c r i p t . This s c r i p t c a l c u l a t e s t h e dynamic l i n e
r a t i n g by c a l l i n g on
#t h e r m a l c a l c .R f o r t h e t h e r m a l c o e f f i c i e n t s .

#This s e c t i o n c a l l s upon t h e t h e r m a l c a l c .R t o l o a d t h e f u n c t i o n s i n t o t h e
workspace
s o u r c e ( ” ˜ / Rcode /IEEE t h e r m a l c a l c .R” )

#This s e c t i o n c o n t a i n s t h e v a r i a b l e s o f t h e l i n e
a <− 0 . 8 #A b s o r p t i v i t y = 0 . 8
E <− 0 . 8 #E m i s s i v i t y = 0 . 8
D <− 0 . 0 1 9 9 # Diameter 1 9 . 9mm
l a t <− 5 9 . 9 #K a r l s t a d l a t i t u d e 5 9 . 9 d e g r e e s
H e <− 7 0 #E l e v a t i o n 5 0m above s e a l e v e l
atmtype <− 1 #Assuming c l e a r atmosphere
Z l <− 2 7 0 # L ine azimuth 9 0 d e g r e e s
R 2 0 <− 0 . 1 4 3 4 ∗ 1 0ˆ(−3 ) #R e s i s t a n c e a t 2 0 d e g r e e s c e l s i u s , Ohm / m
t e m p c o e f f <− 0 . 0 0 4 2 9 # Temperature c o e f f i c i e n t o f aluminium

#This l o o p i t e r a t e s through t h e data s h e e t s from a s t a r t POSIXct−d a t e t o an


end POSIXct−d a t e .
t i c ( ) #S t a r t t i m i n g
s t a r t d a t e <− a s . POSIXct ( ” 2 0 1 9−1 1−0 1 0 0 : 0 0 : 0 0 ” , ”UTC” ) #This d a t e s a c t a s
s t a r t and end−p o i n t s
enddate <− a s . POSIXct ( ” 2 0 2 0−0 2−0 6 0 0 : 0 0 : 0 0 ” , ”UTC” ) #and can be s e t t o
choice
t e m p l i m i t <− 5 0 #The t e m p e r a t u r e l i m i t o f t h e l i n e i s 5 0 d e g r e e s
T s <− t e m p l i m i t
i <− s t a r t d a t e #I t e r a t i o n v a r i a b l e
a v g l e n g t h <− 1 5 #a v g l e n g t h d e t e r m i n e s how many minutes t o a v e r a g e t h e DLR
s l i d i n g window from
#This i n i t i a l i z e s t h e data frame t h a t s a v e s t h e r e l e v a n t data from each
iteration
DLR data . d f <− data . frame ( matrix ( n c o l = 5 , nrow = 0 ) )
DLR avg data . d f <− data . frame ( matrix ( n c o l = 2 , nrow = 0 ) )
w h i l e ( i <= enddate ) {
#The which f u n c t i o n c a l c u l a t e s t h e i n d e x o f t h e d a t a s h e e t −time c l o s e s t t o
i t e r a t i o n time
i n d e x=which . min ( abs ( w e a t h e r d a t a $Time−i ) )
# i n d e x 2 = which . min ( abs ( k a r l s t a d d a t a $ ‘ Tid (UTC) ‘− i ) )
T a <− w e a t h e r d a t a $ ‘ Ambient temperature ‘ [ i n d e x ]
V w <− w e a t h e r d a t a $ ‘ Wind speed ‘ [ i n d e x ]
V ang <− w e a t h e r d a t a $ ‘ Wind a n g l e ‘ [ i n d e x ]
q c = max( c o n v e c t i v e c o o l i n g (D,V w, V ang , Z l , T s , T a , H e ) )
#q c = c o n v e c t i v e c o o l i n g c i g r e (D,V w,V ang , T s , T a , H e , Z l )
q s = s o l a r h e a t i n g ( a , H e , D, l a t , i , atmtype , Z l )
#q s = s o l a r h e a t i n g ( a , D, w e a t h e r d a t a $ ‘ G l o b a l r a d i a t i o n ‘ [ i n d e x ] )
q r = r a d i a t i v e c o o l i n g (D, E , T s , T a )
R = r e s i s t a n c e (R 2 0 , t e m p c o e f f , T s )
I = s q r t ( ( q c+q r−q s ) /R)
#I f t h e weather s t a t i o n has not updated i t s time f o r t h e l a s t 1 5 minutes ,
i t ’ s l i k e l y t o be m a l f u n c t i o n i n g . This
#i s c o n s i d e r e d u n s a f e o p e r a t i o n and t h e ampacity i s s e t t o t h e s t a t i c
l i m i t u n t i l t h e weather s t a t i o n s t a r t s r e s p o n d i n g
i f ( abs ( d i f f t i m e ( w e a t h e r d a t a $Time [ i n d e x ] , i , u n i t s = ” s e c s ” ) ) >=5 ∗ 6 0 ) {
I=5 3 1

57
}
DLR data . d f <− r b i n d (DLR data . df , c ( i , I , q c , q s , q r ) )
DLR avg data . d f <− r b i n d (DLR avg data . df , c ( i , t a i l ( s l i d e d b l (DLR data . d f
[ , 2 ] , ˜mean ( . x ) , . b e f o r e = a v g l e n g t h ) , n=1 ) ) )
#Add one minute t o t h e i t e r a t i o n v a r i a b l e
i=i+6 0}
colnames (DLR data . d f ) <− c ( ”Time” , ” Ampacity ” , ” C o n v e c t i v e c o o l i n g ” , ” S o l a r
heating ” , ” Radiative cooling ” )
colnames (DLR avg data . d f ) <− c ( ”Time” , ” Ampacity ” )
DLR data . d f $Time <− a s . POSIXct (DLR data . d f $Time , o r i g i n = ” 1 9 7 0−0 1−0 1 ” , ”UTC
”)
DLR avg data . d f $Time <− a s . POSIXct (DLR avg data . d f $Time , o r i g i n = ” 1 9 7 0−0 1−
0 1 ” , ”UTC” )
#C a l c u l a t i n g p e r c e n t a g e b e n e f i t o f DLR
DLRbenefit <− DLR avg data c i g r e . d f $ Ampacity / 5 3 1
p e r c e n t a g e b e n e f i t <− sum ( DLRbenefit ) / l e n g t h ( DLRbenefit )

beep ( 2 )
t o c ( ) #Stop t i m i n g f u n c t i o n

A.2 IEEE thermal calculation script


#This s c r i p t c o n t a i n s f u n c t i o n s t o c a l c u l a t e t h e t h e r m a l c o e f f i c i e n t s
a c c o r d i n g t o t h e IEEE s t a n d a r d

#When t h i s s c r i p t i s s o u r c e d by main ( ) , N=0 and w=0 a r e i n i t i a l i z e d . These


must be non−n u l l n u m e r i c s
#i n o r d e r t o run t h e s c r i p t l a t e r on
N=0
w=0
#R a d i a t i v e c o o l i n g f u n c t i o n i s c a l l e d from arguments Diameter , E m i s s i v i t y ,
#Wire temperature , Ambient t e m p e r a t u r e
r a d i a t i v e c o o l i n g <− f u n c t i o n (D, E , T s , T a ) {
q r=1 7 . 8∗D∗E∗ ( ( (T s+2 7 3 ) / 1 0 0 ) ˆ 4 −((T a+2 7 3 ) / 1 0 0 ) ˆ 4 )
return (q r )
}

#C o n v e c t i v e c o o l i n g f u n c t i o n i s c a l l e d from arguments D = diameter , V w =


wind s p e e d
#V ang = e f f e c t i v e a n g l e o f wind i n c i d e n c e , T s = Wire temperature , T a =
ambient t e m p e r a t u r e
#H e = e l e v a t i o n . Returns a l l t h r e e c o n v e c t i v e c o o l i n g components , t h e main
( ) program has t o
#d e t e r m i n e which i s t h e l a r g e s t .
c o n v e c t i v e c o o l i n g <− f u n c t i o n (D, V w, V ang , Z l , T s , T a , H e ) {
a n g l e <− a n g l e o f a t t a c k (V ang , Z l )
K a n g l e = 1 . 1 9 4 − c o s ( deg 2 rad ( a n g l e ) ) + 0 . 1 9 4∗ c o s ( 2∗ deg 2 rad ( a n g l e ) ) + 0 . 3
6 8 ∗ s i n ( 2 ∗ deg 2 rad ( a n g l e ) ) #Wind d i r e c t i o n f a c t o r
T f i l m = (T s+T a ) / 2 #Temperature o f c o n d u c t o r a t boundary l a y e r
rho f = ( 1 . 2 9 3 − 1 . 5 2 5 ∗1 0ˆ{−4} ∗H e+6 . 3 7 9 ∗ 1 0ˆ{−9 } ∗H e ˆ 2 ) / ( 1+0 . 0 0 3 6 7∗T f i l m
) #a i r d e n s i t y
mu f = 1 . 4 5 8∗ 1 0ˆ{−6 } ∗ (T f i l m+2 7 3 ) ˆ 1 . 5 / (T f i l m+3 8 3 . 4 ) #dynamic v i s c o s i t y
of air
k f = 2 . 4 2 4∗1 0ˆ{−2}+7 . 4 7 7 ∗1 0ˆ{−5} ∗T f i l m −4 . 4 0 7 ∗ 1 0ˆ{−9 } ∗T f i l m ˆ 2 #t h e r m a l
conductivity of air
N r e = D∗ rho f ∗V w/mu f #Reynolds number
q c 1 = K a n g l e ∗ ( 1 . 0 1+1 . 3 5 ∗N r e ˆ 0 . 5 2 ) ∗k f ∗ (T s−T a ) #Low wind s p e e d f o r c e d
convection
q c 2 = K a n g l e ∗0 . 7 5 4 ∗N r e ˆ 0 . 6∗k f ∗ (T s−T a ) #High wind s p e e d f o r c e d
convection

58
q cn = 3 . 6 4 5∗ rho f ˆ 0 . 5 ∗Dˆ 0 . 7 5 ∗ (T s−T a ) ˆ 1 . 2 5 #N a t u r a l c o n v e c t i o n

r e t u r n ( c ( q c 1 , q c 2 , q cn ) )
}

#S o l a r h e a t i n g f u n c t i o n
#The s o l a r h e a t i n g f u n c t i o n c a l c u l a t e s t h e s o l a r h e a t i n g c o e f f i c i e n t from
arguments
#a = a b s o r p t i v i t y , H e = e l e v a t i o n , D = diameter , l a t=l a t i t u d e , UTCdate =
t h e d a t e i n UTC,
#atmtype = a b i n a r y o p e r a t o r t o d e s c r i b e t h e type o f atmosphere . #Z l =
azimuth o f t h e l i n e

s o l a r h e a t i n g <− f u n c t i o n ( a , H e , D, l a t , UTCdate , atmtype , Z l ) {


#C a l c u l a t i n g s o l a r d e c l i n a t i o n
N = a s . numeric ( s t r f t i m e ( UTCdate , format = ”%j ” ) ) #C a l c u l a t e s t h e day o f
t h e y e a r from POSIXct v a l u e
#C a l c u l a t i n g s o l a r azimuth
w = deg 2 rad ( ( hour ( UTCdate )−1 2 ) ∗ 1 5 ) #C a l c u l a t i n g hour a n g l e w a s r a d i a n
Z c = s o l a r azimuth (N, w, l a t ) #Function c a l l f o r s o l a r azimuth
H c = s o l a r a l t i t u d e (N, w, l a t ) #Function c a l l f o r s o l a r a l t i t u d e
Q s e = h e a t f l u x i n t e n s i t y (H c ,H e , atmtype )
t h e t a = a c o s ( c o s ( deg 2 rad (H c ) ) ∗ c o s ( deg 2 rad ( Z c )−deg 2 rad ( Z l ) ) )
q s = a ∗ Q s e ∗ s i n ( t h e t a ) ∗ D #This s e c t i o n s h o u l d p o s s i b l y be changed
so that a l l q s < 0 = 0
i f (q s < 0){
q s <− 0
}
return (q s )
}

#This f u n c t i o n c a l c u l a t e s t h e s o l a r azimuth . I t i s a s e p a r a t e f u n c t i o n a s
i t o n l y n e e d s t o be c a l c u l a t e d
#once e v e r y hour . Having t h i s f u n c t i o n o u t s i d e t h e s o l a r h e a t i n g
c a l c u l a t i o n s a v e s computations , a s i t
#can be c a l l e d whenever w c h a n g e s
s o l a r azimuth <− f u n c t i o n (N, w, l a t ) {
s o l d e c = deg 2 rad ( 2 3 . 4 6 ∗ s i n ( deg 2 rad ( ( 2 8 4+N) / 3 6 5∗ 3 6 0 ) ) )
c h i = s i n (w) / ( s i n ( deg 2 rad ( l a t ) ) ∗ c o s (w)−c o s ( deg 2 rad ( l a t ) ) ∗ tan ( s o l d e c ) )
i f (w < 0 ) { #N e s t l e d i f −s t a t e m e n t s t o d e t e r m i n e s o l a r azimuth c o n s t a n t
based on day and time
i f ( c h i >= 0 ) {
C=0
} else {
C=1 8 0
}
} else {
i f ( c h i >= 0 ) {
C=1 8 0
} else {
C=3 6 0
}
}
Z c = C + rad 2 deg ( atan ( c h i ) ) #S o l a r azimuth IN DEGREES
return (Z c )
}
#The same t h a t i s t r u e f o r t h e azimuth i s a l s o t r u e f o r t h e a l t i t u d e
s o l a r a l t i t u d e <− f u n c t i o n (N, w, l a t ) {
s o l d e c = deg 2 rad ( 2 3 . 4 6 ∗ s i n ( deg 2 rad ( ( 2 8 4+N) / 3 6 5∗ 3 6 0 ) ) )

59
H c=rad 2 deg ( a s i n ( c o s ( deg 2 rad ( l a t ) ) ∗ c o s ( s o l d e c ) ∗ c o s (w)+s i n ( deg 2 rad ( l a t ) ) ∗
sin ( soldec ) ) )
r e t u r n (H c ) #Return s o l a r a l t i t u d e IN DEGREES
}

#The same i s t r u e f o r t h e h e a t f l u x i n t e n s i t y , i t i s dependent on t h e type


o f atmosphere
#and t h e s o l a r a l t i t u d e , which i s c a l c u l a t e d i n t h e s o l a r a l t i t u d e f u n c t i o n
and must be ran b e f o r e t h i s
h e a t f l u x i n t e n s i t y <− f u n c t i o n (H c ,H e , atmtype ) {
i f ( atmtype == 1 ) { #I f c l e a r atmosphere
Q s = −4 2 . 2 3 9 1 + 6 3 . 8 0 4 4∗H c+(−1 . 9 2 2 0 ) ∗H c ˆ 2+3 . 4 6 9 2 1 ∗ 1 0ˆ(−2 ) ∗H c ˆ 3 + (−
3 . 6 1 1 8 ∗ 1 0ˆ(−4 ) ) ∗H c ˆ 4 + 1 . 9 4 3 1 8 ∗ 1 0ˆ(−6 ) ∗H c ˆ 5 + (−4 . 0 7 6 0 8 ∗ 1 0ˆ(−9 ) ) ∗H
cˆ6
} e l s e { #E l s e i n d u s t r i a l atmosphere
Q s = 5 3 . 1 8 2 1 + 1 4 . 2 1 1 0∗H c+ 6 . 6 1 3 8 ∗ 1 0ˆ(−1 ) ∗H c ˆ 2+(−3 . 1 6 5 8 ∗ 1 0ˆ(−2 ) ) ∗H c
ˆ 3 + 5 . 4 6 5 4 ∗ 1 0ˆ(−4 ) ∗H c ˆ 4 + (−4 . 3 4 4 6 ∗ 1 0ˆ(−6 ) ) ∗H c ˆ 5 + 1 . 3 2 3 6 ∗ 1 0ˆ(−8 )
∗H c ˆ 6
}
#C o r r e c t i n g f o r a l t i t u d e
K s o l a r = 1 + 1 . 1 4 8∗1 0ˆ(−4 ) ∗H e+1 . 1 0 8 ∗ 1 0ˆ(−8 ) ∗H e ˆ 2
Q s e=K s o l a r ∗Q s
r e t u r n (Q s e )
}

#Function t o c a l c u l a t e t e m p e r a t u r e dependent r e s i s t a n c e from s u r f a c e


t e m p e r a t u r e and aluminum temperature −r e s i s t a n c e c o e f f i c i e n t
r e s i s t a n c e <− f u n c t i o n (R 2 0 , t e m p c o e f f , T s ) {
R T=R 2 0 ∗ ( 1+t e m p c o e f f ∗ (T s − 2 0 ) )
r e t u r n (R T)
}

A.3 Cigre thermal calculation script


#This s c r i p t c o n t a i n s f u n c t i o n s t o c a l c u l a t e t h e t h e r m a l c o e f f i c i e n t s
a c c o r d i n g t o t h e Cigre −6 0 1 method
#Convection h e a t l o s s
c o n v e c t i v e c o o l i n g c i g r e <− f u n c t i o n (D,V w,V ang , T s , T a , H e , Z l ) {
#Film t e m p e r a t u r e around c o n d u c t o r
T f =(T s+T a ) / 2
#Thermal c o n d u c t i v i t y
lambda f=2 . 3 6 8∗1 0ˆ(−2 )+7 . 2 3 ∗ 10ˆ(−5 ) ∗T f −2 . 7 6 3 ∗1 0ˆ(−8 ) ∗T f ˆ 2
#Dynamic v i s c o s i t y o f a i r
#D e n s i t y o f a i r g i v e n e l e v a t i o n H e
rho f = ( 1 . 2 9 3−1 . 5 2 5 ∗ 1 0ˆ(−4 ) ∗H e+6 . 3 7 9 ∗ 1 0ˆ(−9 ) ∗H e ˆ 2 ) / ( 1+0 . 0 0 3 6 7∗T f )
mu f = ( 1 7 . 2 3 9 + 4 . 6 3 5 ∗1 0ˆ{−2} ∗T f − 2 . 0 3 ∗ 1 0ˆ{−5 } ∗ T f ˆ 2 ) ∗ 1 0ˆ{−6 }
#Kinematic v i s c o s i t y o f a i r
v f = mu f / rho f
#Reynolds number
Re = V w ∗ D/v f
#The c o e f f i c i e n t s f o r N u s s e l t s number depends on t h e r o u g h n e s s o f t h e
conductor . Since
#t h i s code i s o n l y meant t o a n a l y s e one type o f conductor , i t w i l l be ”
hard−i n p u t ” i n t o t h i s
#f u n c t i o n t o s a v e u n n e c e s s a r y g l o b a l v a r i a b l e s i n f u n c t i o n c a l l s . Rs = 0 .
09
#The B and n c o e f f i c i e n t s a r e c h o s e n depending on t h e Reynolds number
i f ( Re < 2 6 5 0 ) {
B = 0 . 641
n = 0 . 471

60
} else {
B = 0 . 048
n = 0.8
}

#N u s s e l t number a t p e r p e n d i c u l a r wind
Nu 9 0 = B∗Reˆn
#N u s s e l t number a t g i v e n wind d e l t a
a n g l e <− a n g l e o f a t t a c k (V ang , Z l )
i f ( a n g l e <= 2 4 ) {
Nu d e l t a = ( 0 . 4 2 + 0 . 6 8 ∗ s i n ( ( deg 2 rad ( a n g l e ) ) ) ˆ 1 . 0 8 ) ∗ Nu 9 0
} else {
Nu d e l t a = ( 0 . 4 2 + 0 . 5 8∗ s i n ( ( deg 2 rad ( a n g l e ) ) ) ˆ 0 . 9 ) ∗Nu 9 0
}
#N a t u r a l c o n v e c t i o n
#P r a n d t l number , c f i s s p e c i f i c h e a t o f aluminum
c f = 921
P r = c f ∗mu f / lambda f
G r = Dˆ 3 ∗ (T s−T a ) ∗ 9 . 8 0 7 / ( (T f+2 7 3 ) ∗v f ˆ 2 )
#A and m i n n a t u r a l c o n v e c t i o n e q u a t i o n depends on p r o d u c t o f P r and G r
PG <− P r ∗G r
i f (PG < 1 0 ˆ 2 ) {
A <− 1 . 0 2
m <− 0 . 1 4 8
}
i f (PG >= 1 0 ˆ 2 & PG < 1 0 ˆ 4 ) {
A <− 0 . 8 5
m <− 0 . 1 8 8
}
i f (PG >= 1 0 ˆ 4 & PG < 1 0 ˆ 7 ) {
A <− 0 . 4 8
m <− 0 . 2 5
}
i f (PG >= 1 0 ˆ 7 ) {
A <− 0 . 1 2 5
m <− 0 . 3 3 3
}
Nu nat = A∗PGˆm

#Find maximum e x p r e s s i o n t o c a l c u l a t e c o n v e c t i v e h e a t l o s s
Nu <− max(Nu nat , Nu d e l t a )

#E x p r e s s i o n o f c o n v e c t i v e h e a t l o s s
q c = p i ∗ lambda f ∗ (T s−T a ) ∗Nu
return (q c )
}

#This f u n c t i o n c a l c u l a t e s t h e a n g l e o f a t t a c k o f t h e wind , which i s a


between 0−p i r a d i a n s a t a l l t i m e s
a n g l e o f a t t a c k <− f u n c t i o n (V ang , Z l ) {
a n g l e 1 = abs (V ang − Z l )
a n g l e 2 = abs (V ang − ( Z l+1 8 0 ) )
a n g l e 3 = abs (V ang − ( Z l − 18 0 ) )
a n g l e o f a t t a c k = min ( c ( a n g l e 1 , a n g l e 2 , a n g l e 3 ) )
return ( angle of attack )
}
#R a d i a t i v e c o o l i n g
r a d i a t i v e c o o l i n g <− f u n c t i o n (D, E , T s , T a ) {
q r = p i ∗D∗ 5 . 6 7 ∗ 1 0ˆ(−8 ) ∗E∗ ( (T s+2 7 3 ) ˆ 4−(T a+2 7 3 ) ˆ 4 )

61
return (q r )
}

#S o l a r h e a t i n g f u n c t i o n . I t t a k e s i n p u t s a <− a b s o r p t i v i t y , H e <−
e l e v a t i o n above ground ,
#D <− d i a m e t e r o f conductor , l a t <− l a t i t u d e o f conductor , l o n g <−
l o n g t i t u d e o f conductor ,
#UTCdate <− c u r r e n t d a t e and time , #atmtype <− c l o u d c o v e r a g e f a c t o r , Z l
<− azimuth o f c o n d u c t o r
#F r e d <− r e f l e c t i v i t y o f t h e ground
s o l a r h e a t i n g e s t i m a t e <− f u n c t i o n ( a ,H e , D, l a t , long , UTCdate , atmtype , Z l , F
ref ){
N = a s . numeric ( s t r f t i m e ( UTCdate , format = ”%j ” ) ) #C a l c u l a t e s t h e day o f
t h e y e a r from POSIXct v a l u e
#C a l c u l a t i n g s o l a r azimuth
w = deg 2 rad ( ( hour ( UTCdate+l o n g ∗ 4∗ 6 0 )−1 2 ) ∗ 1 5 ) #C a l c u l a t i n g hour a n g l e w a s
radian
s o l d e c = deg 2 rad ( 2 3 . 3 ∗ s i n ( 2 ∗ p i ∗ ( ( 2 8 4+N) / 3 6 5 ) ) ) #S o l a r d e c l i n a t i o n
H c = s o l a r a l t i t u d e (N, w, l a t , s o l d e c ) #Function c a l l f o r s o l a r a l t i t u d e
I b0 = N s ∗ ( 1 2 8 0∗ s i n (H c ) ) / ( s i n (H c )+0 . 3 1 4 ) #D i r e c t s o l a r r a d i a t i o n a t
sea l e v e l
I by = I b0 ∗ ( 1+1 . 4∗ 1 0ˆ(−4 ) ∗H e ∗ ( 1 3 6 7 / I b0−1 ) ) #D i r e c t s o l a r r a d i a t i o n a t
elevation H e
I d = ( 4 3 0 . 5−0 . 3 2 8 8 ∗ I by ) ∗ s i n (H c ) #D i f f u s e s o l a r r a d i a t i o n
s o l az = s o l a r azimuth (w,H c , s o l d e c )
e t a = a c o s ( c o s (H c ) ∗ c o s ( deg 2 rad ( Z l )−s o l az ) )
p s <− a ∗D∗ ( ( I by∗ ( s i n ( e t a )+p i / 2∗F r e f ∗ s i n (H c ) ) )+I d∗ ( 1+p i / 2 ∗F r e f ) )
i f (H c < 0 ) {
p s <− 0
}
return (p s )
}

s o l a r a l t i t u d e <− f u n c t i o n (N, w, l a t , s o l d e c ) {
H c=a s i n ( c o s ( deg 2 rad ( l a t ) ) ∗ c o s ( s o l d e c ) ∗ c o s (w)+s i n ( deg 2 rad ( l a t ) ) ∗ s i n (
soldec ) )
r e t u r n (H c ) #Return s o l a r a l t i t u d e IN DEGREES
}
s o l a r azimuth <− f u n c t i o n (w,H c , s o l d e c ) {
s o l az = a s i n ( c o s ( s o l d e c ) ∗ s i n (w) / ( c o s (H c ) ) )
r e t u r n ( s o l az )
}

s o l a r h e a t i n g <− f u n c t i o n ( a , D, I T) {
p s <− a ∗D∗ I T
return (p s )
}

#Function t o c a l c u l a t e t e m p e r a t u r e dependent r e s i s t a n c e
r e s i s t a n c e <− f u n c t i o n (R 2 0 , t e m p c o e f f , T s ) {
R T=R 2 0 ∗ ( 1+t e m p c o e f f ∗ (T s − 2 0 ) )
r e t u r n (R T)
}

A.4 Wind Power Injection Script


#This s c r i p t c a l c u l a t e s t h e c u r t a i l m e n t a t d i f f e r e n t l e v e l s o f i n t e r m i t t e n t
generation .
#I t r e q u i r e s t h e g l o b a l v a r i a b l e s ” s t a r t d a t e ” and ” enddate ” which a r e
i n i t i a l i z e d i n t h e main

62
#f u n c t i o n t o run . I t a l s o r e q u i r e s t h e wind farm g e n e r a t i o n data t o be
loaded i n t o the
#g l o b a l environment . ” s t a r t d a t e ” and ” enddate ” a c t a s p o i n t e r s f o r which
parts of the
#wind farm g e n e r a t i o n data t o u s e i n t h e c a l c u l a t i o n s

p backa $ ‘ A c t i v e power ‘ [ p backa $ ‘ A c t i v e power ‘ > 0 ] <− 0

#e t a i s t h e p o r t i o n o f c u r r e n t from p backa t h a t p a s s e s through t h e l i n e , 7


0%
e t a <− 0 . 7
#The v o l t a g e o f t h e l i n e i s 1 3 0kV , m u l t i p l y with 1 0 0 0 t o go from kA t o A.
The n o t a t i o n i s t h a t
#power i n j e c t e d i n t o t h e l i n e from backa i s measured , s o t h e s i g n o f p
backa must be r e v e r s e d
wind c u r r e n t <− (−p backa $ ‘ A c t i v e power ‘ ) ∗ e t a ∗ 1 0 0 0 / ( 1 3 5 ∗ s q r t ( 3 ) )
p backa <− data . frame ( p backa $Time , p backa $ ‘ A c t i v e power ‘ , wind c u r r e n t )
colnames ( p backa ) <− c ( ”Time” , ” A c t i v e power ” , ” Current ” )

#The wind park has a c a p a c i t y o f 6 8MW. Thus we can c a l c u l a t e t h e c u r r e n t


f o r 1MW a s 1 / 6 8 th o f wind c u r r e n t

#The added wind power i s c a l c u l a t e d f o r e v e r y MW t o a l i m i t


c a l c l i m i t = 3 0 0 #The c a l c u l a t i o n i s l i m i t e d a t 2 3 2MW o f added wind power
n <− 5 #I t s t e p s 1 MW p e r i t e r a t i o n
m <− 1 #P o s i t i o n o f column f o r
s t e p s i z e = 5∗60
f r e e c a p a c i t y . d f <− data . frame ( matrix ( n c o l = c a l c l i m i t%/%5 ,
nrow = ( a s . numeric ( enddate )−a s . numeric
( s t a r t d a t e ) )%/%s t e p s i z e ) ) #C r e a t e
data frame t h e s i z e o f e x p e c t e d
calculation
l i n e l o a d . d f <− data . frame ( matrix ( n c o l = c a l c l i m i t%/%5 ,
nrow = ( a s . numeric ( enddate )−a s . numeric (
s t a r t d a t e ) )%/%s t e p s i z e ) )
produced power . d f <− data . frame ( matrix ( n c o l = c a l c l i m i t%/%5 ,
nrow = ( a s . numeric ( enddate )−a s .
numeric ( s t a r t d a t e ) )%/%s t e p s i z e )
)
f r e e c a p a c i t y . d f [ , 1 ] <− data . frame ( s e q ( s t a r t d a t e , enddate−1 , s t e p s i z e ) )
l i n e l o a d . d f [ , 1 ] <− data . frame ( s e q ( s t a r t d a t e , enddate−1 , s t e p s i z e ) )
produced power . d f [ , 1 ] <− data . frame ( s e q ( s t a r t d a t e , enddate−1 , s t e p s i z e ) )
#f r e e c a p a c i t y c i g r e . d f <− f r e e c a p a c i t y . d f #Same s i z e a s IEEE−7 3 8 c a p a c i t y
#l i n e l o a d c i g r e . d f <− l i n e l o a d . d f
w h i l e ( n−1 < c a l c l i m i t ) {
i <− s t a r t d a t e #S t a r t v a r i a b l e f o r time i t e r a t i o n through d a t a s e t
k <− 1
w h i l e ( i < enddate ) {
i n d e x 1 <− which . min ( abs (DLR avg data . d f $Time−i ) )
i n d e x 2 <− which . min ( abs ( p backa $Time − i ) )
i n d e x 3 <− which . min ( abs ( data 4 2 7 $DateTime − i ) )
# i n d e x 4 <− which . min ( abs (DLR avg data c i g r e . d f $Time−i ) )
f r e e c a p a c i t y . d f [ k ,m+1 ] = DLR avg data c i g r e . d f $ Ampacity [ i n d e x 1 ] −(p
backa $ Current [ i n d e x 2 ] ∗n/ 6 8+data 4 2 7 $ Current [ i n d e x 3 ] )
l i n e l o a d . d f [ k ,m+1 ] = p backa $ Current [ i n d e x 2 ] ∗n/ 6 8+data 4 2 7 $ Current [ i n d e x
3]
produced power . d f [ k ,m+1 ] = (−p backa $ ‘ A c t i v e power ‘ [ i n d e x 2 ] ) ∗n/ 6 8
# f r e e c a p a c i t y c i g r e . d f [ k ,m+1 ] = DLR avg data . d f $ Ampacity [ i n d e x 1 ] −(p
backa $ Current [ i n d e x 2 ] ∗n/ 6 8+data 4 2 7 $ Current [ i n d e x 3 ] )

63
# l i n e l o a d . d f [ k ,m+1 ] = p backa $ Current [ i n d e x 2 ] ∗n/ 6 8+data 4 2 7 $ Current [ i n d e x
3]
i <− i+s t e p s i z e
k=k+1
}
m = m+1
n = n+5
}
colnames ( f r e e c a p a c i t y . d f ) <− c ( ”Time” , p a s t e ( s e q ( 5 , c a l c l i m i t , 5 ) , ”MW” ) )
colnames ( l i n e l o a d . d f ) <− c ( ”Time” , p a s t e ( s e q ( 5 , c a l c l i m i t , 5 ) , ”MW” ) )
colnames ( produced power . d f ) <− c ( ”Time” , p a s t e ( s e q ( 5 , c a l c l i m i t , 5 ) , ”MW” ) )
beep ( 2 )

#This s e c t i o n c a l c u l a t e s t h e % o f c u r t a i l m e n t f o r each l e v e l o f wind power


integration ,
#and t h e e n e r g y not s e r v e d (ENS) , compared t o DLR and SLR
f r e e c a p a c i t y . d f [ i s . na ( f r e e c a p a c i t y . d f ) ] <− 0
c u r t a i l m e n t e n s <− data . frame ( matrix ( nrow = n c o l ( f r e e c a p a c i t y . d f )−1 , n c o l =
10 ) )
d u r a t i o n = a s . numeric ( d i f f t i m e ( enddate , s t a r t d a t e , u n i t s = ” h o u r s ” ) )
n <− 2
SLR = 5 3 1 #S t a t i c l i n e r a t i n g i s s e t t o 5 3 1
w h i l e ( n <= n c o l ( f r e e c a p a c i t y . d f ) ) {
c u r t a i l m e n t <− 0 #I n i t i a l i s i n g v a r i a b l e f o r amounts o f time power has
been c u r t a i l e d
e n s <− 0 #I n i t i a l i s i n g v a r i a b l e f o r amount o f e n e r g y not s e r v e d
t r a n s m i t t e d power <− 0
s o l d e l e c t r i c i t y <− 0
s o l d e l e c t r i c i t y s l r <− 0
c u r t a i l m e n t s l r <− 0
e n s s l r <− 0
t r a n s m i t t e d power s l r <− 0
k=1
w h i l e ( k <= nrow ( f r e e c a p a c i t y . d f ) ) {
#The ” Transmitted power ” i n t h i s v a r i a b l e i s a temporary l o c a l v a r i a b l e
f o r t h i s i t e r a t i o n . I t i s kept a s
#t h e summation o f s o l d e l e c t r i c i t y must be i n c r e m e n t a l ( a s t h e p r i c e
v a r i e s with time ) . The produced power
#v a r i a b l e i s not used d i r e c t l y , s i n c e t h a t would mean i t would be
changed i n t h e i f −s t a t e m e n t s . This would be
#h i g h l y u n d e s i r a b l e hence a temporary v a r i a b l e t r a n s m i t t e d power i s
used i n s t e a d . This v a r i a b l e i s o v e r w r i t t e n
#below t h e loop , when t h e t r a n s m i t t e d power f o r t h e e n t i r e column i s
summed
t r a n s m i t t e d power <− produced power . d f [ k , n ]
t r a n s m i t t e d power s l r <− produced power . d f [ k , n ]
i f ( f r e e c a p a c i t y . df [ k , n ] < 0 ) {
c u r t a i l m e n t = c u r t a i l m e n t +1
ens = ens + f r e e c a p a c i t y . df [ k , n ]
t r a n s m i t t e d power <− t r a n s m i t t e d power + f r e e c a p a c i t y . d f [ k , n ] ∗ 1 3 5 ∗ 1 0
ˆ−3

}
i f ( l i n e l o a d . d f [ k , n ] > SLR) {
c u r t a i l m e n t s l r = c u r t a i l m e n t s l r+1
e n s s l r = e n s s l r + SLR−l i n e l o a d . d f [ k , n ]
t r a n s m i t t e d power s l r <− t r a n s m i t t e d power s l r +(SLR−l i n e l o a d . d f [ k , n ] )
∗ 1 3 5 ∗ 1 0ˆ−3
}

64
MWh p r i c e <−e l s p o t p r i c e s . d f $ ’SE3 ’ [ e l s p o t p r i c e s . d f $ Date == a s . Date (
f r e e c a p a c i t y . d f $Time [ k ] ) &
a s . numeric ( s u b s t r ( e l s p o t p r i c e s . d f $ Hours , 1 , 2 ) )
== hour ( f r e e c a p a c i t y . d f $Time [ k ] ) ]
s o l d e l e c t r i c i t y = MWh p r i c e ∗ t r a n s m i t t e d power ∗ 5 / 6 0+s o l d e l e c t r i c i t y
s o l d e l e c t r i c i t y s l r = MWh p r i c e ∗ t r a n s m i t t e d power s l r ∗ 5 / 6 0+s o l d
electricity slr
# s a v e e l e c t r i c i t y [ n , k ] <− s o l d e l e c t r i c i t y
k=k+1
}
#The t r a n s m i t t e d power i s t h e produced power minus t h e e n e r g y not s e r v e d .
The produced
#power i s i n MW, t h e ENS i s n e g a t i v e and i n A, by m u l t i p l y i n g with 1 3 5kV ,
t h e e n s i s then
#i n kW. Hence i t i s m u l t i p l i e d with 1 0ˆ−3 t o r e p r e s e n t MW. Each sample
p o i n t i s 5 minutes ,
#hence by m u l t i p l y i n g with 5 /6 0 we c o n v e r t t h e t r a n s m i t t e d power output
t o MWh. The
#t r a n s m i t t e d power becomes t h e power t r a n s m i t t e d f o r t h e e n t i r e d u r a t i o n
between s t a r t d a t e
#and enddate
t r a n s m i t t e d power <− ( sum ( produced power . d f [ , n ] )+e n s ∗ 1 3 5 ∗1 0ˆ(−3 ) ) ∗5 / 6 0
t r a n s m i t t e d power s l r <− ( sum ( produced power . d f [ , n ] )+e n s s l r ∗ 1 3 5 ∗ 1 0ˆ(−3 ) )
∗5/60
c u r t a i l m e n t e n s [ n−1 , 1 : 1 0 ] <− c ( c u r t a i l m e n t /nrow ( f r e e c a p a c i t y . d f ) ,( − e n s ∗5 ∗
1 3 5 ∗1 0ˆ(−3 ) / 6 0 ) ∗8 7 6 0 / d u r a t i o n , t r a n s m i t t e d power ,
t r a n s m i t t e d power / d u r a t i o n , c u r t a i l m e n t s l r
/nrow ( f r e e c a p a c i t y . d f ) ,
(− e n s s l r ∗5 ∗ 1 3 5 ∗1 0ˆ(−3 ) / 6 0 ) ∗ 8 7 6 0 / d u r a t i o n ,
t r a n s m i t t e d power s l r , t r a n s m i t t e d power
s l r / duration ,
s o l d e l e c t r i c i t y ∗8760/ duration , s o l d
e l e c t r i c i t y s l r ∗8760/ duration )
n=n+1
}
colnames ( c u r t a i l m e n t e n s ) <− c ( ”%C u r t a i l m e n t ” , ”ENS [MWh/ y e a r ] ” , ”Power
s u p p l i e d [MWh] ” , ”Power s u p p l i e d [MWh/ y e a r ] ”
, ”%C u r t a i l m e n t SLR” , ”ENS SLR [MWh/ y e a r ] ” ,
”Power s u p p l i e d SLR [MWh] ” , ”Power
s u p p l i e d SLR [MWh/ y e a r ] ” ,
” S o l d E l e c t r i c i t y [ SEK/ y e a r ] ” , ” S o l d
E l e c t r i c i t y SLR [ SEK/ y e a r ] ” )
rownames ( c u r t a i l m e n t e n s ) <− s e q ( 5 , c a l c l i m i t , 5 )

A.5 Risk factor calculation script


#This s c r i p t p e r f o r m s an a n a l y s i s o f c r i t i c a l p o i n t s a l o n g a l i n e

VL X <− r e a d e x c e l ( ” ˜ / Exjobb /Data/ vl −x p r o f i l e . x l s x ” )

i <− 1
r u r a l l i m i t <− 6 . 4 #The l i m i t f o r a 1 3 0kV l i n e i n r u r a l a r e a s i s 6 . 4m
pop l i m i t <− 7 . 4 #The l i m i t f o r a 1 3 0kV i n p o p u l a t e d a r e a s i s 7 . 4m
u n c e r t a i n t y <− 0 . 5 #The DLR and DTG−measurement u n c e r t a i n t y i s e s t i m a t e d a s
0 . 5m
dtg r i s k f a c t o r <− NULL
w h i l e ( i < n c o l (VL X) ) {
#This s e c t i o n d e t e r m i n e s t h e r i s k f a c t o r due t o d i s t a n c e −to−ground . I f
t h e d i s t a n c e t o ground i s below t h e a l l o w a b l e

65
#l i m i t + a f a c t o r o f u n c e r t a i n t y , t h e dtg r i s k f a c t o r i s s e t t o 1 , i f t h e
d i s t a n c e t o ground i s below t h e a l l o w a b l e
#l i m i t t h e r i s k f a c t o r i s s e t t o 2
dtg r i s k f a c t o r [ i ] <− ( ( r u r a l l i m i t+u n c e r t a i n t y ) / ( a s . numeric (VL X[ 4 , i+1 ] )
)−1 ) ∗ 3+1
i f (VL X[ 8 , i+1 ] == 1 ) { #I f i n p o p u l a t e d area , u s e p o p u l a t e d boundary
instead
dtg r i s k f a c t o r [ i ] <− ( ( pop l i m i t+u n c e r t a i n t y ) / ( a s . numeric (VL X[ 4 , i+1 ] )
)−1 ) ∗3+1
}
i=i+1
}

s h a d i n g r i s k f a c t o r <− a s . numeric (VL X[ 6 , 2 : n c o l (VL X) ] )+a s . numeric (VL X[ 7 , 2


: n c o l (VL X) ] )
#The r e l a t i v e w e i g h t o f t h e dtg r i s k f a c t o r i s s e t s o t h e max r i s k f a c t o r
equals 10 .
dtg r e l a t i v e w e i g h t <− 1 0 /max( dtg r i s k f a c t o r )
dtg r i s k f a c t o r <− dtg r i s k f a c t o r ∗ dtg r e l a t i v e w e i g h t
dtg r i s k f a c t o r [ dtg r i s k f a c t o r < 0 ] <− 0
t o t a l r i s k f a c t o r <− ( s h a d i n g r i s k f a c t o r + dtg r i s k f a c t o r ) / 2
t o t a l r i s k f a c t o r [ t o t a l r i s k f a c t o r < 0 ] <− 0
VL X <− r b i n d (VL X, c ( ”DTG Risk F a c t o r ” , dtg r i s k f a c t o r ) )
VL X <− r b i n d (VL X, c ( ” Shading Risk F a c t o r ” , s h a d i n g r i s k f a c t o r ) )
VL X <− r b i n d (VL X, c ( ” T o t a l Risk F a c t o r ” , t o t a l r i s k f a c t o r ) )

References
[1] Cigre TF B2.12.3, “Sag-tension calculation methods for overhead lines,” Technical Brochure 324,
April 2016.

[2] C. J. Wallnerström, P. Hilber, P. Söderström, R. Saers, and O. Hansson, “Potential of dynamic


rating in sweden,” 07 2014.

[3] M. Merante, “Application of dynamic rating to improve transportation capability of the power
systems connected to wind power plants,” Master thesis, KTH Royal Institute of Technology, 2016.

[4] E. A.Bergström, “Experiences from dynamic line rating pilot installations in vattenfall distributions
145 kv system in sweden,” 09 2017.

[5] S. Talpur, C. J. Wallnerström, and P. Hilber, “Dynamic line rating for wind power,” Journal of
Renewable and Sustainable Energy, vol. 8, p. 19, 02 2016.

[6] C. J. Wallnerström, Y. Huang, and L. Söder, “Impact from dynamic line rating on wind power
integration,” IEEE Transactions on Smart Grid, vol. 6, no. 1, pp. 343–350, Jan 2015.

[7] L. Dawson and A. M. Knight, “Applicability of dynamic thermal line rating for long lines,” IEEE
Transactions on Power Delivery, vol. 33, no. 2, pp. 719–727, April 2018.

[8] B. P. Bhattarai, J. P. Gentle, T. McJunkin, P. J. Hill, K. S. Myers, A. W. Abboud, R. Renwick, and


D. Hengst, “Improvement of transmission line ampacity utilization by weather-based dynamic line
rating,” IEEE Transactions on Power Delivery, vol. 33, no. 4, pp. 1853–1863, Aug 2018.

[9] “Ieee standard for calculating the current-temperature relationship of bare overhead conductors,”
IEEE Std 738-2012 (Revision of IEEE Std 738-2006 - Incorporates IEEE Std 738-2012 Cor 1-2013),
pp. 1–72, Dec 2013.

[10] Cigre WG B2.42, “Guide for thermal rating calculations of overhead lines,” Technical Brochure 601,

66
Dec 2014.

[11] M. W. Davis, “A new thermal rating approach: The real time thermal rating system for strate-
gic overhead conductor transmission lines – part ii: Steady state thermal rating program,” IEEE
Transactions on Power Apparatus and Systems, vol. 96, no. 3, pp. 810–825, May 1977.

[12] S. Uski-Joutsenvuo, R. Pasonen, and S. Rissanen, “Maximising power line transmission capability
by employing dynamic line ratings - technical survey and applicability in finland,” VTT Technical
Research Centre of Finland, Tech. Rep., February 2013.

[13] H. T. Yip, C. An, M. Aten, and R. Ferris, “Dynamic line rating protection for wind farm con-
nections,” in 2008 IET 9th International Conference on Developments in Power System Protection
(DPSP 2008), March 2008, pp. 693–697.

[14] M. Aten and R. Ferris, “Dynamic line rating protection for wind farms overview of project,” in IET
Conference on Substation Technology 2009: Analysing the Strategic and Practical Issues of Modern
Substation, Nov 2009, pp. 1–29.

[15] P. Schell, J. Lambin, B. Godard, H.-M. Nguyen, and J.-L. Lilien, “Using dynamic line rating to
minimize curtailment of wind power connected to rural power networks,” 2011.

[16] R. Puffer, M. C. Schmale, M. Scheufen, B. Rusek, and C. Neumann, “Area-wide dynamic line ratings
based on weather measurements,” 2012.

[17] D. M. Greenwood, J. P. Gentle, K. S. Myers, P. J. Davison, I. J. West, J. W. Bush, G. L. Ingram,


and M. C. M. Troffaes, “A comparison of real-time thermal rating systems in the u.s. and the u.k.”
IEEE Transactions on Power Delivery, vol. 29, no. 4, pp. 1849–1858, Aug 2014.

[18] J. Dooley, “Two-dimensional interpolation of irregularly spaced data using polynomial splines,”
Physics of the Earth and Planetary Interiors, vol. 12, no. 2, pp. 180 – 187, 1976. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/0031920176900467

[19] W. Wang and S. Pinter, “Dynamic line rating systems for transmission lines - topical report,” U.S.
Department of Energy, Tech. Rep.

[20] A. Philips, B. Clairmont, D. Childs, D. Reuger, D. Douglass, J. Bell, and D. Birrell, “Evaluation of
instrumentation and dynamic thermal ratings for overhead lines,” Electric Power Research Institute,
Tech. Rep., October 2013.

[21] T. Goodwin, “Dynamic line rating - oncor delivery smart grid program,” Oncor Electric Delivery
Company, Tech. Rep., August 2013.

[22] C. Ahlrot, Network protection system Öland, E.ON, 8 2018.

[23] A. Estanqueiro, C. Ahlrot, J. Duque, D. Santos, J. Gentle, A. Abboud, K. Morozovska, P. Hilber,


L. Soder, and T. Kanefendt, “Dlr use for optimization of network design with very large wind (and
vre) penetration,” 10 2018.

[24] J. L. Aznarte and N. Siebert, “Dynamic line rating using numerical weather predictions and machine
learning: A case study,” IEEE Transactions on Power Delivery, vol. 32, no. 1, pp. 335–343, Feb 2017.

[25] A. Arroyo, P. Castro, R. Martı́nez, M. Mañana, A. Madrazo, R. Lecuna, and A. Gonzalez, “Com-
parison between ieee and cigre thermal behaviour standards and measured temperature on a 132-kv
overhead power line,” Energies, vol. 8, pp. 13 660–13 671, 12 2015.

[26] L. Staszewski and W. Rebizant, “The differences between ieee and cigre heat balance concepts for
line ampacity considerations,” in 2010 Modern Electric Power Systems, Sep. 2010, pp. 1–4.

[27] K. Morozovska, “Dynamic rating of power lines and transformers for wind energy integration,”

67
Disseration, KTH Royal Institute of Technology, 2018.

[28] Tan Loc Le, M. Negnevitsky, and M. Piekutowski, “Expert system application for the loading
capability assessment of transmission lines,” IEEE Transactions on Power Systems, vol. 10, no. 4,
pp. 1805–1812, Nov 1995.

[29] E. M. Greitzer, Z. S. Spakovszky, and I. A. Waitz, “2.4 Specific Heats.” [Online]. Available:
https://web.mit.edu/16.unified/www/FALL/thermodynamics/notes/node18.html

[30] K. Reenaas, Elsäkerhetsverkets föreskrifter och allmänna råd om hur elektriska stark-
strömsanläggningar ska vara utförda, 3rd ed., Elsäkerhetsverket, 6 2016.

[31] E. Lindberg, “The overhead line sag dependence on weather parameters and line current,” Master
thesis, Uppsala University, 2011.

[32] K. Morozovska and P. Hilber, “Study of the monitoring systems for dynamic line
rating,” Energy Procedia, vol. 105, pp. 2557 – 2562, 2017, 8th International Conference
on Applied Energy, ICAE2016, 8-11 October 2016, Beijing, China. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S1876610217307981

[33] Power Donut 3: PD3 Instrumentation Platform for overhead transmission lines, Underground Sys-
tems Inc., 3A Trowbridge Drive, Bethel, CT, USA, 1 2017.

[34] C. J. Wallnerström and P. Hilber, Reliability Analysis and Asset Management Applied to Power
Distribution, 1st ed. Edita Bobergs AB, 3 2014, ch. 6, pp. 65–87.

[35] D. Kuhlin, K. Eriksson, and M. Stenkvist, “Produktionskostnader för vindkraft i sverige,” En-
ergimyndigheten, Tech. Rep., 2016.

[36] SCB, “Inflation i sverige 1831-2019.” [Online]. Available: https://www.scb.se/hitta-


statistik/statistik-efter-amne/priser-och-konsumtion/konsumentprisindex/konsumentprisindex-
kpi/pong/tabell-och-diagram/konsumentprisindex-kpi/inflation-i-sverige/

[37] E. Kreutz and E. Peterson, “Lönsamhet i vindkraft - en fallstudie av lönsamheten i vindkraftverk


belägna i gns,” Bachelor Thesis, Swedish University of Acricultural Science, 2018.

[38] K. Mohammadi, A. Mostafaeipour, Y. Dinpashoh, and N. Pouya, “Electricity generation and energy
cost estimation of large-scale wind turbines in jarandagh, iran,” Journal of Energy, vol. 2014, pp.
1–8, 11 2014.

[39] P. Leko and M. Koyama, “Wind power: Technology brief,” IRENA and IEA-ETSAP, Tech. Rep.,
2016.

[40] J. McCarthy, “Wind farm decommissioning: A detailed approach to estimate future costs in sweden,”
Master Thesis, Uppsala University, 2015.

[41] Cigre WG B2.36, “Guide for application of direct real-time monitoring systems,” Technical Brochure
498, June 2012.

[42] M. Matus, D. Saez, M. Favley, C. Suazo-Martinez, J. Moya, G. Jimenez-Estevez, R. Palma-Behnke,


G. Olguin, and P. Jorquera, “Identification of critical spans for monitoring systems in dynamic
thermal rating,” IEEE Transactions on Power Delivery, vol. 27, no. 2, pp. 1002–1009, April 2012.

[43] P. M. Callahan and D. A. Douglass, “An experimental evaluation of a thermal line uprating by
conductor temperature and weather monitoring,” IEEE Transactions on Power Delivery, vol. 3,
no. 4, pp. 1960–1967, Oct 1988.

[44] J. Black, S. Connor, and J. Colandairaj, “Planning network reinforcements with dynamic line ratings
for overhead transmission lines,” in 45th International Universities Power Engineering Conference

68
UPEC2010, Aug 2010, pp. 1–6.

[45] J. Teh and I. Cotton, “Critical span identification model for dynamic thermal rating system place-
ment,” IET Generation, Transmission Distribution, vol. 9, no. 16, pp. 2644–2652, 2015.

[46] S. Talpur, T. Lie, and R. Zamora, “Application of dynamic thermal rating: Over-
head line critical spans identification under weather dependent optimized sensor place-
ment,” Electric Power Systems Research, vol. 180, p. 106125, 2020. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0378779619304444

69
TRITA-EECS-EX-2020:574
www.kth.se

You might also like