Professional Documents
Culture Documents
Final Report Print, Sameep
Final Report Print, Sameep
INSTITUTE OF ENGINEERING
PASHCHIMANCHAL CAMPUS
THESIS NO : PAS075MSDGE014
"Probabilistic load flow analysis using point estimation method: A case study of
Tandi distribution feeder"
by
Sameep Poudel
A THESIS
POKHARA, NEPAL
Bhadra, 2078
Probabilistic load flow analysis using point estimation method: A case study of
Tandi distribution feeder"
by
Sameep Poudel
Thesis Supervisor
Tribhuvan University
Pokhara, Nepal
Bhadra, 2078
ii
COPYRIGHT ©
The author has agreed that the Library, Department of Electrical Engineering,
Pashchimanchal Campus, Institute of Engineering may make this report freely available
for inspection. Moreover, the author has agreed that permission for extensive coping of
this thesis report for scholarly purpose may be granted by the supervisor who supervised
the thesis work recorded herein or, in their absence, by the Head of the Department
wherein the thesis report was done. It is understood that the recognition will be given
to the author of this report and to the Department of Electrical Engineering, Institute of
Engineering, Pashchimanchal Campus in any use of the materials of this thesis report.
Copyright or publication or other use of this report for the financial gain without
approval of the Department of Electrical Engineering, Institute of Engineering,
Pashchimanchal Campus and author’s permission is prohibited. Request for the
permission to copy or to make any other use of the material in this report in whole or
in part should be addressed to:
Head
Pokhara, Nepal
iii
CERTIFICATE OF APPROVAL
iv
ACKNOWLEDGEMENT
On the very outset of this thesis report, I would like to extend my sincere obligation
towards the Department of Electrical Engineering, Pashchimanchal Campus for
providing me the opportunity to pursue my Master’s Degree in Electrical Engineering
in Distributed Generation.
v
ABSTRACT
After successful completion of using point estimation method on IEEE-33 bus system,
Tandi distribution system as a real feeder with 122 buses is taken for the analysis using
Point Estimation method. TOD load data of Tandi feeder is collected from Tandi DCS
vi
and parsa substation. Hourly variation of solar irradiance from 9am to 5pm of each day
measured by pyranometer installed at Gairapatan, Pokhara based on beta distribution is
taken as real data for PV penetration of tandi feeder. Gram charlier expansion with
chebyshev hermite polynomial equation is consider to calculate the Probability Density
function(PDF) and Cumulative distribution function(CDF) of the random variable.
CDF of pu voltage and active power loss before and after DG penetration is compare
and analysed. Solar as a DG is connected at the different Buses and Probability of over
voltage and under voltage at different bus before and after DG penetration is analysed.
DG location is selected based on the least pu voltage and end of the lateral branch. The
result show the effect of DG penetration on Voltage and power loss of the system.
vii
TABLE OF CONTENTS
ACKNOWLEDGEMENT ......................................................................................... v
ABSTRACT ............................................................................................................. vi
viii
3.1.2 The probabilistic model of load .......................................................... 16
REFERENCES ........................................................................................................ 45
ix
LIST OF TABLES
x
LIST OF FIGURES
Figure 2. 3 Wiring diagram for grid connected solar PV system [19]. ........................ 9
xi
Figure 4. 19 Intermittency of Solar Power Generation ............................................. 39
xii
LIST OF ABBREVIATIONS
DG Distributed Generation.
PV Photo-Voltaic.
xiii
CHAPTER ONE
INTRODUCTION
1.1 Background
Rapid development of renewable energy resources play a vital role to control carbon
emission. The increasing use of energy creates a challenges for energy saving as a result
use of renewable energy resources has increases in the power system rapidly. with the
development of new technology and innovation, the use of solar energy has increases,
but due to the fluctuating nature of solar output operation and penetration problem
arises [1]. The stochastic nature of solar output is generally due to maximum generation
at the day time and no generation of power during the night. Solar output also vary
seasonally, more at spring than the winter.The climatic condition and meteorological
feature of different location also causes uncertainty generation of the solar in the
distribution system [2].
There are normally three method for PLF calculation as analytical method, simulation
method and approximation method. Simulation method mainly include Monte-carlo
simulation(MCS) [5] this changes the each random variable into the sequence of
deterministic load flow problem. MCS can be consider as the most accurate PLF
method among various other method so it is commonly used for analysing the accuracy
of various other probabilistic method [3]. The accuracy of MCS is achieve by the use
of large number of random sampling and iteration calculation which makes this method
computationally inefficient and time consuming hence for the large system with many
buses MCS is not used [6]. so to overcome this drawback, PEM can be used.
1
Analytical method include the cumulant method and convolution method.Analytical
method less computational burden with references to simulation method as it consider
the number of problem into the mathmatical assumption to obtain the accurate
information of output random variable [7].
Point estimation is the method of using sample data to calculate a single value (Point
estimate as it identifies a point in some parameter space). Point estimation uses the
value from the sample of unknown parameter to get the best fit of the output data. Each
sample data from the large number of population is use by the point estimation for the
output parameter. random sample of the population is used to approximate the value
using the mathmatical equation and assumtion to get the best fit of the output data.
There are two types of estimation as point estimation and internal estimation, point
estimation produces a single value whereas internal estimation produces a range of
value [1].
Point estimation method have good balance of accuracy and computational burden. It
need few deterministic power flow runs (it required less number of deterministic
iteration for same number of input variables). Mathmatical equation and derivation are
simple and it require less information of input random variable as compared to other
method [8-9].
Another most popular method to solve the probabilistic power flow compare to MCS
method is PEMs. Mean, variance, skewness and kurtosis are some statistical moment
used as probabilistic function for PEM. Along these 4 central moment, other 2 central
moment in terms of polynomial function can also be used in PEM for accuracy of the
method. PEM is superior than MCS method in the sense that PEM require less data,
less computational burden and it takes less computational time [7]. The accuracy of
PEM can also be improved by increasing the number of estimation point 'm' so it can
be said that 3PEM is more accuracy than 2PEM and less number of estimation point
lead to inaccuracy. It should be noted than on increasing the estimation point the higher
order moment is required but higher order moment greater than 4th order has no physical
meaning [4]. The use of higher order estimation point may arises the following
problem;
2
2) standard location and weight for m> 3 cannot be determined.
3) Higher order central moment may give negative or non real value of location
and weight.
So for improve the accuracy only large number of estimation point shouldnot be used
and on considering the less estimation point arises large inaccuracy. Hence considering
the optimum condition, 3PEM is used in my thesis work.
The availability of generation at various location and due to high demand of power
system distributed generation are connected to the network. The power output of
distributed generation like PV highly varies as the source of energy is also vary. On
integrating such network to the grid proper analysis must be done for proper operation
of the system. on using deterministic load flow various uncertainties are not consider.
For each value of bus loads, generation patterns and network topologies, large number
of computation effort are required so the use of conventional load-flow computation
seems impractical.
For such stochastic generation of the power probabilistic load flow analysis is accurate
and more effective as it consider all the uncertainties like uncertainty in load demand,
generation outage.
Large number of sample data and iteration are use by the monte carlo simulation for
each of the random output data which make the Monte-carlo simulation time consuming
and computationally inefficient for best result. For the large system on using Monte
carlo simulation require computer having high processor, greater capacity of RAM and
more memory space with large computational time. Such drawback of MCS can be
overcome by the PEM. MCS method is seems less accurate for radial distribution
system than PEM.
3
1.3 Objective
Main Objectives
Specific Objectives
• To use IEEE-33 bus system for probabilistic load flow using PEM.
• To compare the result of Point estimation method and Monte-carlo simulation
method in IEEE-33 Bus system.
• Probabilistic Load flow of IEEE-33 bus and Tandi distribution system having
122 buses using Point estimation method with PV penetration is done in
MATLAB environment.
• The Point Estimation method with deterministic load flow is done using Square
root method.
• Maximum load is consider for deterministic load flow of tandi feeder.
• Hourly variation of solar irradiation and load is modeled by beta and normal
distribution respectively.
• The various cases for simulation are no PV generation with normally distributed
load, 25% PV generated of total load at different bus to generate PDF and CDF
of voltage and power loss.
4
1.5 Report organization
• Chapter Two describe the literature review of distribution system, load flow
analysis and probabilistic load flow of power system with PV interconnection
is also discussed here. Beside this the use of point estimation method over the
monte carlo simulation method is also discussed in this chapter.
• Chapter Three describe the methodology of the work, to meet the require
objective. It include the probabilistic model of PV and load, flow chart, point
estimation method and detail about the Gram-charlier expansion.
• Chapter Four Provide the result obtained from the MATLAB environment. It
include the result of load flow and point estimation method with PV connection
on IEEE-33 Bus Radial Distribution system. The result of Point estimation
method and monte-carlo simulation on IEEE- 33 Bus system is compare and
analyse here. The result of Tandi Distribution feeder of 122 buses with PV
interconnection using point estimation method is also shown here.
• Chapter Five conclude the result and present the further work to do.
5
CHAPTER TWO
LITERATURE REVIEW
Literature review show the need for additional research. It helps to gain and clearify the
knowledge on the particular subject. It is the initial stage for all research activities as
unsolved problem can be easily found and further research can be continue. some
research activities also help to solve the controversey in a particular field. Research
makes clearification of any matter with justification as well as theoritical or practical
proof.
After the electricity is generated at the power station, it is then transmitted through the
transmission system followed by the distribution system to the end user. The electricity
is transmitted at high voltage at the transmission system which is then step down and
transmitted through the distribution system.
A drop in voltage levels results when demand for electricity exceeds the capacity of the
distribution system. Medium voltage power is transmitted to the distribution
transformers near the customer location through primary distribution lines. Distribution
transformers again lower the voltage by step down transformer to the utilization voltage
which is then passes through the secondary distribution line to the customer [10].
Radial distribution system is more applicable as loop and network system are more
expensive because more switches and more conductors is necessary to construct the
loop and network system.
6
The power from the main branch is split to sub branch and again split out from the sub
branches in radial distribution system [10]. The cost of radial distribution system is
low compare to other network so it is less reliable although radial distribution network
is prefer when substatin is located at the center of the load and due to its low cost,
simple construction and easy to expand. Radial distribution system is mostly used in
densely populated areas [11].
7
Solar is one of the mostly used Distributed Generation in power system. The penetration
of Distributed generation in the distribution system helps to improve the Voltage profile
of the system, decreases the transmission losses and also impove the efficiency. DG
penetration can replace the power transformer, distribution station, can break the power
monopoly, can supply electricity in remote areas, reduced transmission loss, reduced
price volatility, reduced transmission and distribution congestion [14].
PV array is the combination of PV module and are connected in series and in parallel.
The process of obtaining electricity directly from light energy is solar or PV generation
system. Solar mainly depend on photovoltaic effect so it is also called photovoltaic cell.
Electricity is produced from the semico nductor solar cell when light strikes on it and
potential different product across it terminal. The potential different is normally
maintained at 0.5 volt and voltage is independent to intensity of incident light whereas
current depend on intensity and area of the solar panel [17].
There are many organization and researching center that has certain standards around
the world for PV integration such including the US, Germany, Spain, etc. In the context
of Nepal, according to National Energy Crisis Control and Electricity Development
Decade, 2072 S.N 52’B’, 500 watt or above Solar PV can be connected to the national
grid through Net-Metering system. Net Metering should be bidirectional meter capable
of reading both import and export of electricity. For Net-Metering, Customers must
8
submit an application to the respective Distribution Centers along with Single Line
Diagram [18]. NEA Standards of quality of power of solar PV to be supplied by
consumer to Grid should be of following:
• Frequency – 50 Hz
• The power generated by the customers solar PV should have power factor between
0.85 lag to 0.95 lead.
• The fault clearance time for earth fault, overcurrent and frequency fluctuation should
be of standard described by IEC/IEEE
• 230V up-to 5 kW
• 400V 5 kW -40 kW
• 11kV above 40 kW
9
2.5 Deterministic load flow
The point estimation method is applied on IEEE 33 bus radial distribution system. The
load flow of such a system is carried out Using Square root method [20].
Let us consider a radial Distribution system with two node as shown in figure 3.2;
V2∠∂2
V1 ∠∂1 P2+ jQ2
R+jX
I2
jQ2 Node 1 Node 2
where,
R + jX = Branch Impedance.
V1 , V2 = Voltage magnitude
Then,
𝑃2 −𝑗𝑄2
I2 = ...(2.2)
𝑉2∗
Also,
|𝑉1 |∠ ∂1 −|𝑉2 |∠ ∂2
I2 = …(2.3)
𝑅+𝑗 𝑋
and
10
|V2| =
1
√[{(𝑃2 𝑅 + 𝑄2 𝑋 − 0.5|𝑉1 |2 )2 − (𝑅2 + 𝑋 2 )(𝑃2 + 𝑄2 )}2 − (𝑃2 𝑅 + 𝑄2 𝑋 − 0.5|𝑉1 |2 )]
2 2
…(2.5)
From above equation 2.4 and equation 2.5 gives the angle and the magnitude of the
Voltage respectively at the receiving end Bus.
The above equation for voltage magnitude and phase angle at receiving end node can
be written in generalized form as;
V(m2) =|B(j)-A(j)|1/2
Where,
B(j) = {A2(j)-[R2(j)+X2(j)]*[P2(m2)+Q2(m2)]}1/2
and
𝑃(𝑚2 )∗𝑋(𝑗)−𝑄(𝑚2 )∗𝑅(𝑗)
∂(m2) = ∂(m1) -tan-1[𝑃(𝑚 2 (𝑚 ]
2)∗𝑅(𝑗)+𝑄(𝑚2 )∗𝑋(𝑗)+𝑉 2)
similarly,
Real and Reactive power loss of respective branch can be calculated as;
j= Branch Number.
P = Active Power.
Q= Reactive Power.
11
start
Initialize value
Set convergency
criteria
Update value
Yes
Max error>
Tolerance
No
Display Result
Stop
The deterministic load flow of radial distribution system of IEEE-33 bus system and
real tandi distribution feeder is done using the square root method. The input variable
are line data which contain value of resistance and inductance, and Bus data which
contain the load at each buses. Initially the pu voltage is set one and voltage angle del
is set zero. The load flow is done to calculate the active power, reactive power, voltage
and voltage angle till the value is converge and maximum error reaches below the
tolerance value. finally the result are display and iteration of load flow stop. The detail
of deterministic load flow shown in flow chart is shown in the figure 2.5.
12
2.6 Monte carlo simulation
• Optimization
• Numerical integration
• Probability distribution.
The probabilistic method is mainly based on the uncertainty in the system due to the
stochastic generation of PV. The review of several paper based on point estimation
method has been described below;
Yulong Che[4] estimate the probabilistic moment of Probabilistic power flow using the
improved three point estimation method. IEEE-14 bus system and IEEE-118 buses with
normally distributed input variable are used for two point estimation, three point
estimation and Chebyshev inequality to obtained the result of improved three point
estimation method.
Chun Lien Su [8] is based on the probabilistic load flow analysis using point estimation
method. IEEE 6 bus system is used and result are verified using the monte carlo
simulation method. If the stochastic nature of DG output is estimated, all the line
parameter and power flow can be accurately evaluated by using 2PEM through
mathmatical equations. However, more estimation point can be taken for better result.
13
J. Emmanuel Anandraj [11] is based on probabilistic load flow of distribution network
with PV unit. Active power produced by PV and power absorbed by load are taken as
uncertainties. Probabilistic load flow is done in MATLAB Software in 30 bus radial
network using two point estimation and three point estimation scheme and obtained
result are compared using Monte- Carlo simulation method to show that three point
estimation method is better than two point estimation method.
Delgado [22] said that the variation in load demands and power fluctuation of
renewable generators cannot be considered with the deterministic power flow because
it uses specific power values so probabilistic power flow method is used here that reflect
the uncertainty of input variables. This paper extends the point estimation
method(PEM) with dispersed generation and variable power factors of wind and solar.
This paper combine the 3 point estimation scheme with cholesky decomposition to
estimate the cumulative distribution function. This system is tested on the three phase
unbalanced IEEE 123 node test system and result are compared with those obtained
result from the monte carlo simulation method.
B. K. Panigrahi [23] present the Two schemes of PEM (2m and 3m) and their
performance is evaluated by comparing it to the Monte-Carlo based PPF. Point
estimation method is applied in transmission system (30-Bus) and radial distribution
system(33-Bus). The obtained result from both the method show that the point
estimation method has better perfomance in radial networks as compare to monte carlo
simulation method.
It can be summaries from the different paper of literature review that stochastic
behaviour of DG cannot be trace by the deterministic load flow and so the probabilistic
method of load flow seems practical and accurate. The performance of point estimation
method is more accurate for radial distribution system than Monte carlo simulation
method. The output of MCS and PEM is almost same but MCS require more
computational time than PEM. The accuracy of PEM can be maintain by increasing the
number of estimation point. Hence 3 point estimation is more accurate than 2 point
estimation method and require only one more computation time than 2 point estimation
method.
14
CHAPTER THREE
METHODOLOGY
The methodology show the step by step procedure to achieve the objective of the work.
Literature review show the background information and detail about the point
estimation method. The work is divided into two part;
Data of real feeder are collected from the Tandi Distribution center and Parsa substation
which include the line data and Bus data of Tandi Distribution feeder with 122 buses
located at Chitwan district of Nepal.
Due to stochastic nature of solar irradiance(Ib ) the power output obtained from the solar
vary at different time. The variation of solar irradiance and hence the power generated
by the solar can be consider to follow the beta distribution [24]. The equation for power
output of the solar is given by;
PG = η*Ac*Ib
where,
For power generation by the PV, the probabilistic density function is given by [25];
Γ (α +β)
fPG(PG ) = Γ (α )∗Γ (β) (Ib ) α-1 *(1-Ib)β-1
Where;
15
fPG is the probability density function for power generated; α,β denote the shape
parameters of beta distribution with 12.62 and 2.21 respectively[11] and Γ is the gamma
function.
As per the ontario's standard the percentage of penetrated DG unit is limited to less than
30% [26]. According to [27] the best penetration level is shown 21% for IEEE -33 bus
distribution system considering only two factors, power loss and voltage regulation and
other factor such as economic and geographic are not considerations.
The power ( PL) absorbed by load at any load bus is random and assumed to follow a
normal distribution with certain value of mean and standard deviation [11]. The
probability density function of PL is given by:
−(𝑃𝐿−μ𝑃𝐿 )2
1 2∗(𝜎𝑃𝐿 )2
fPL(PL) = σ * 𝑒 ...(3.1)
𝑃𝐿 ∗√2𝜋
Where;
μPL denote the mean value and σPL is the standard deviation.
Equation 3.1, show the variation of load active power. Load reactive power is also
follow normal distribution function and PDF for reactive power can also be obtained
from the equation 3.1 [11].
Few point of random variable are consider in calculation the statistical information or
central moment in the point estimation method. Using these point (concentration),
output information is obtained [28].
The various uncertain parameter are the power generated by the solar and normally
distributed load active and reactive power, these uncertain parameter are consider
random variable. For three point estimation method, the concentration of random
variable can be calculated as [28];
16
Where;
The standard location ℇl,k and weight Wl,k of input random variable is obtained by
solving the below given non-linear equations 3.2 and 3.3 by using the equation derived
by the Miller and Rice [29].
1
∑𝑘𝑘=𝑙 𝑊 l,k = …(3.2)
𝑚
Where,
m is the total number of input random variables, i.e. the uncertain generator power and
the active and reactive load powers; 𝜆 l,j denotes the jth standard central moment of the
random variable ρl . The concentration point ρk,l may move away from the mean value
µρ. so in three point estimation method one of the concentration is set to mean to
maintain the accuracy [28].
For two point estimation method each random variable consider only the two point. The
statistical information can be obtained from the same equation 3.2 and 3.3 at K=2.
standard location and weight are calculated from skewness factor λl,3. Here output
move away from the mean value µρ as variation of input variable changes the vaue of
standard location and concentration Hence uniformity is not obtained. This is overcome
by three point estimation method by considering one concentration as mean value [28].
In case of 3PEM one concentration points out of three is taken as mean value by
considering one of the standard locations ξl,k as zero. Hence K = 3 is taken for 3PEM.
17
statistical information is obtained from equation 3.2 and 3.3. The standard locations ξl,k
and weight Wl,k of random variable can be obtained by;
λ𝑙,𝑘 3
ξl,k = +(-1)3-k √( λ𝑙,4 − λ 2𝑙,3 for k=1,2
2 4
ξl,3 = 0;
(−1)3−𝑘
wl,k = ξ for k=1,2
𝑙,𝑘 (ξ 1,𝑙 −ξ 2,𝑙 )
1 1
wl,3 = (𝑚 − )
λ𝑙,4 − λ2𝑙,3
Three point estimation method is more accurate than two point estimation method as
input random variable 'm' do not depend on standard location [11].
The standard central moment is denoted by 𝜆 l,j . It is useful in comparing the shape of
different probability distribution. The moment of the probability distribution about the
mean is the central moment. central moment show the variation of data from the first
raw moment [11]. The various central moment are;
where,
Skewness measure the symmetry of the distribution and study the shape of the curve.
3∗(Mean−Median)
Skewness = standard deviation
Kurtosis show the differences between the tail of the distribution with the Normal
Distribution.
1∗∑𝑛
𝑘=𝑙 (𝑋𝑘−𝑋𝑚𝑒𝑎𝑛)^4
Kurtosis= 𝑛∗σ
where,
18
n = No. of random variable.
σ= Standard deviation.
Cumulant method is an efficient method to generate the PDF of random variable. This
method has less computational burden than the convolution method. The probability
distribution having same moment have identical cumulants so moment determine the
cumulant, similarly cumulant also determine the moments. Different order cumulant
can be obtained using the cumulant generating function [22] as shown below;
The six order cumulant obtained from the six raw moment are;
k1 = µ1
k2 = µ2 - µ12
k3 = µ3 -3µ2µ1 + 2µ13
Four order cumulant is normally used, but for accuracy six order cumulant is used here
in my work. The result of the cumulant is used to obtained the PDF and CDF curve
using the Gram- charlier expansion.
19
3.4 Gram-Charlier Expansion
The density of Gram−Charlier distribution is the polynomial times the normal density.
Gram-charlier expansion find the probability distribution of random variable in term of
its cumulant. If x is mean value and µ is standard deviation σ of random variable, the
standardized variable x̅ has the form x̅ = (x−µ)/ σ . According to Gram-Charlier
expansion, the cumulative density function F (x̅) and probability density functions f(x̅)
can be written as [11];
𝐶1 𝐶2 𝐶3
F(x̅) = 𝜙(x̅) + 1! 𝜙′ (x̅) + 2! 𝜙′′(x̅) + 3! 𝜙′′′(x̅)
𝐶1 𝐶2 𝐶3
f(x̅) =𝜑(x̅) + 1! 𝜑′(x̅) + 2! 𝜑′′(x̅) + 3! 𝜑′′′(x̅)
Where,
𝜙(x̅) and φ(x̅) represent the CDF and PDF of the standard normal distribution with µ
= 0 and σ = 1, respectively, cv constant coefficients with the following form.
co = 1
c1 = c2 = 0
c3 = -β3/σ3
c4 = β4/σ4 -3
c5 = -β5/σ5 + 10 β3/σ3 …
To solve this Gram-charlier expansion equation for PDF and CDF, chebyshev hermite
polynomial equation are required [22], which are as;
He0(x) = 1
He1(x) = x
He2(x) =x2 -1
He3(x) = x3 -3x
He4(x) = x4 -6x2+3
He5(x) = x5 -10x3+15x
20
Read input
variable
k =1
l =1
compute,
Standard central moment (λ𝑙,j )
Standard location (ξl,k )
weight (Wl,k )
if l <= m
if k <= 3
step 4: for all random variable calculate standard central moment,standard location and
weight for the point.
Step 5: calculate the points for the random variable as per the estimation scheme (three
point estimation)
Step 8: For all output random variable calculate the cumulant and plot the CDF curve
using Gram-Charlier expression.
The obtained result from the point estimation method is compared with the result from
the monte carlo simulation method in IEEE-33 bus system. Active and reactive power,
minimum and maximum voltage, overvoltage and Undervoltage is taken as the
parameter for comparison. The execution time with point estimation method and monte
carlo simulation method is also compared.
22
CHAPTER FOUR
RESULT AND DISCUSSION
RAM 8GB
The various parameter for load flow analysis of IEEE-33 Bus system is taken as shown
in table 4.2.
S. N Parameter Data
1. Number of Buses 33
2. Number of Branches 32
23
The single line diagram of IEEE- 33 Bus system is shown in figure 4.1;
It is seen from Figure 4.2 that minimum per unit voltage is 0.9038 and 0.9164 at Bus
18 and 33 of IEEE-33 Bus system.The active power loss at various buses is shown in
figure 4.3.
From figure 4.3 the maximum active power loss is 52.08 Kw at branch number 2.
Active power loss also goes on decreasing as we move away from the source.
24
Figure 4. 3 Active power loss of IEEE-33 Bus system.
Active Reactive
S.N Power(KW) Power(Kvar)
From figure 4.4, it can be observed that after obtaining the PDF curve of the required
parameter we can say that what will be the variation range of that parameter for the
considered variation in input parameters. Voltage magnitude of bus-18 can be seen to
be varied approximately in the range of 0.899 to 0.91 p.u. while that of bus-33 in the
range of 0.91 to 0.924 p.u. in considerable amount.
25
Figure 4. 4 PDF curve of Bus 18 and 33
Mean value of voltage magnitude obtained for bus-18 and bus-33 from calculation are
0.9037 p.u. and 0.9164 p.u. respectively. From normally distributed pdf curve of the
both bus, it can be observed that maximum density value coincides to the corresponding
mean value. In similar way pdf curves for other buses can be obtained and can get
information as per the requirement. Hence, showing that large amount of information
can be obtained from probabilistic load flow in compare with deterministic load flow.
If all the random variable are equally distributed for all pu voltage then flatten type PDF
curve is obtained.
1.02
0.98
0.96
Pu voltage
0.94
0.92
0.9
0.88
0.86
0.84
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Bus Number
The minimum voltage and maximum voltage that can be obtained from IEEE-33 Bus
system with Normally distributed load having zero skewness and 3 kurtosis is shown
27
in the figure 4.6. The minimum and maximum voltage are 0.8994 and 0.9083 at Bus 18
of IEEE-33 Bus system.
Probabilistic load flow of IEEE-33 bus system is carried out by Monte Carlo simulation
technique for 500 iteration and in each iteration 40 samples are taken.
It is seen that the output voltage from Both the method coincide each other and have
the error less than 0.001%. The Active power Loss obtained from both the method are
almost similar with less than 3% error. The comparison of MCS and PEM is shown in
table 4.4.
Hence it can be concluded from the comparison result of Point Estimation method and
Monte- carlo simulation method that the accuracy of Monte- carlo simulation method
is also maintain by Point Estimation method and Point Estimation Method is Prefer
here for further work because,
28
4.1.3 Point Estimation Method with DG
Injecting 25% of solar at bus number 18 and 33 of IEEE-33 bus system. Bus Number
18 and 33 is selected for solar injection as it has the least pu voltage as compare to other
node. The generation from the solar is given by probability density function using beta
distribution. The value of Alpha and beta is 12.62 and 2.21 having mean value
851kw/m2 and standard deviation of 8.95%[11]. The random variable for the PEM are
Solar active power and load active and reactive power which is based on beta and
normal distribution respectively.
1.02
1
0.98
Pu Voltage
0.96
0.94
0.92
0.9
0.88
0.86
0.84
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Bus Number
Vm 0% DG Vm 25% DG
29
From the CDF curve shown in figure 4.7 the Probability of obtaining under voltage at
Bus 18 get improved from 0.8994 to 0.9267 and over voltage get improved from 0.9083
to 0.9521 after 25% of DG penetrated at Bus 18 and 33. Hence voltage get improved,
which is more clearly shown by the figure 4.8.
60
Active Power Loss(KW)
50
40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
Branches
The comparison of Active Power Loss after 25% DG penetrated on IEEE-33 Bus
system is shown in figure 4.9. It can be seen that Active Power Loss in each branches
get reduces after PV Penetration.
NO 25%
Particular
PENETRATION PENETRATION
30
It is seen from the Table 4.5 that after PV interconnection mean voltage goes on
increasing from 0.9453 to 0.961. Minimum voltage of the bus increases from 0.9037 to
0.9397 and active and reactive power loss decreasing as PV is Penetrated.
After successful completion of using Point Estimation Method with IEEE-33 bus
system, Tandi Distribution Feeder with 122 buses as a Radial distribution system is
taken for Analysis. The single line diagram of Tandi Distribution Feeder with Parsa
substation as a source at Bus 1 is shown in figure 4.10;
Table 4.6 shows the parameter of 122 bus Tandi feeder which is one of the feeder of
NEA, Ratnanagar Tandi Distribution center among 4 feeders. The feeder is supply from
the Parsa substation and has feeder length of around 115 km. The data obtained from
the Tandi feeder and corresponding Line data and Bus data is shown in Appendix 2.
31
Table 4. 6 Key Parameter of Tandi feeder.
S. N Parameter Data
The maximum KVA is 5439.52 KVA with 0.86 average power factor obtained from
the TOD load data of Tandi feeder. Hence 59.4% Transformer loading is obtained. The
Load data is obtained by substituting the value of resistance and reactance of Dog,
Rabbit and Weasel conductor.
The mean of the load parameter of tandi feeder is taken with reference to 59.4%
transformer loading and 0.86 pf and standard deviation is taken with references to
standard deviation obtained from 22 TOD load data.
The daily load curve of Tandi Feeder is shown in the figure 4.11.
3000
2500
2000
Load (KW)
1500
1000
500
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Times of day (Hours)
The TOD data of the Tandi feeder is checked for the best fit of the distribution. The
result obtained from the TOD data at Bus number 15 are;
250
200
Frequency
150
100
50
0
1 2 3 4 5 6 7 8
KVA interval
From the Table 4.7 it is seen that the Mean and Median are almost equal to each other
and the value of skewness and kurtosis nearly equal to zero and lies in the interval [-
2,2]. The negative of skewness show that the data are right skewed.
The figure 4.12 also show the histogram plot which are almost symmetrical. All the
TOD load data are checked and can be concluded that the taken data are Normally
distributed.
33
Figure 4. 13 Voltage magnitude plot of Tandi feeder.
The Pu voltage of tandi distribution feeder obtained from the deterministic load flow is
shown in the figure 4.13. The pu voltage goes on decreasing and is 0.817 at Bus 122
which is the minimum voltage at end bus of the feeder.
The result of Voltage magnitude in terms of Minimum voltage and Maximum voltage
obtained using Point Estimation Method of Tandi feeder is as shown in Table 4.8;
35
59 0.8575 0.876 120 0.8027 0.8357
60 0.86 0.8785 121 0.8022 0.8355
61 0.8588 0.8776 122 0.802 0.8347
The minimum and maximum voltage obtained from each bus of the tandi feeder is
shown in Table 4.8 and Figure 4.14 and it is seen the pu voltage goes on decreases as
we move toward the end buses and variation of pu voltage at end buses is 0.802 to
0.8347pu.
120
100
80
Probability
60
40
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Bus Number
The under voltage probability from bus 1 to bus 20 of tandi feeder is shown in the figure
4.15. There is no under voltage probability from bus 1 to bus 6 and above bus 15 to bus
122 there is 100% probability of getting under voltage before DG penetration which is
36
more clearly shown by the figure 4.15. There is no chance of over voltage probability
in tandi feeder.
The detail of Load and Loss of Tandi feeder is shown in the Table 4.9;
Active Reactive
Parameter
Power(KW) Power(KVAr)
Tandi feeder has the total load of 4674.19kw and 2773.49 kVAr. On considering the
given value of conductor of tandi feeder as shown in Appendix 2 the total line losses
are 544.904kw and 633.131Kvar, Hence 5219.09 kw and 3406.621kvar is supplies from
the substation under the normal condition when DG is not connected.
Figure 4.16 show the CDF curve of voltage magnitude of Bus 40, 80 and 122 of Real
Tandi Distribution feeder. The probability of obtaining voltage 0.82 pu and less than
0.82 pu at Bus 122 is around 90% and there is no chance of obtaining voltage greater
than 0.834 at Bus 122 before DG penetration. The Pu voltage at Bus 122 is 0.802 pu at
worst conditon and 0.834 pu at best condition. similar result can be obtained for Bus
40, 80 from figure 4.16 and for all other buses can also be calculated.
37
Figure 4. 17 CDF plot of Active Power Loss of Tandi feeder
38
no chance of loss less than 47 kvar and more than 56 kvar at Branch 2 of tandi feeder.
similarly reactive power loss at branch 3 and 4 and, and other branches of tandi feeder
can be obtained.
It is seen that the pu voltage of the Bus goes on decreasing as we move away from the
source. The Bus 122 which is the end Bus of the Tandi Distribution Feeder has the least
pu voltage i.e 0.817. To improve the Pu voltage of the system and also to decrease the
losses 25% of the solar as per ontario's standard [26] and [27] is penetrated at different
buses of the real system considering only two factors, power loss and voltage regulation
and other factor such as economic and geographic are not considerations. The
probabilistic analysis of tandi distribution feeder is done for three different scheme of
DG connection as , 25% of DG connected at Bus 122, 25% of DG connected at Bus
122,94 and 25% of DG connected at Bus 122,94,71. The Bus is selected based on
Minimum per unit voltage and end of the lateral branch.
800
PV output irradiance (W/m2)
700
600
500
400
300
200
100
0
9:13
9:32
9:51
10:1
7:00
7:19
7:38
7:57
8:16
8:35
8:54
10:2
10:4
11:0
11:2
11:4
12:0
12:2
12:4
1:01
1:20
1:39
1:58
2:17
2:36
2:55
3:14
3:33
3:52
4:11
4:30
4:49
5:08
5:27
5:46
Time of Day
The solar irradiance for each hour of the day is modeled by the Beta probability density
function based on collected data of solar irradiance from 9am to 5pm of 4 month, taken
as a reference for penetration of PV. The mean value of irradiance obtained is 0.451
KW/m2 and 61.03% standard deviation with beta distribution parameter as shown in
table 4.10.
39
Table 4. 10 Beta Distribution Parameter of PV.
The figure 4.20 show the Pu voltage plot with three different scheme of DG connection.
It is seen that Pu voltage is improved with each scheme of DG connection. The
maximum improvement in Pu voltage is seen at scheme 1 when 25% of DG is
connected at the end bus.
From figure 4.21 and 4.22 of pu voltage, it can be observed that although the minimum
voltage probability remain almost same but maximum voltage Probability is increases
for each scheme of DG connection. The maximum increases in voltage of Bus 120 is
seen at scheme 1 when 25% of DG is connected at bus 122 and maximum increases in
voltage of Bus 71 is seen at scheme 3 when 25% of DG is connected at Bus 122,94
and 71. Hence it is seen that pu voltage is improved at each scheme of DG connection.
The minimum voltage probability remain almost same although the DG is connected
because of the intermittance character of PV.
Similarly, for all other Buses of tandi feeder CDF can be plotted and probability of
minimum voltage and maximum voltage at each buses can be determined. The CDF
plot of active power loss at Branch 4 is shown in figure 4.23, It is seen that Probability
of active power loss at branch 4 is decreases with each scheme of DG connection.
Under the normal condition when DG is not connected, the mean active power loss at
branch 4 is 43 KW which can be reduced maximum and reaches to 25KW under the
best condition of DG connection and over active power loss at branch 4 is around 46
41
kw for each scheme under worst condition of DG connection which is as similar to no
DG connection scheme as shown in figure 4.23.
Hence for all other branches of tandi feeder the probability of maximum and minimum
power loss can be obtained for each scheme of DG connection.
25% DG connected at
Bus Bus
Parameter No DG Bus 122 122,94 122,94,71
42
Table 4.11 show the comparison of various parameter for each scheme after 25% DG
connected at different Bus. The Minimum voltage of 0.8182 at Bus 122 is improved to,
0.8513 when 25% of DG connected at Bus 122, 0.8469 when 25% of DG connected at
Bus 94,122, 0.8417 when 25% of DG connected at Bus 71,94,122. Hence with each
scheme of DG connection the pu voltage get improved and maximum improvement in
minimum voltage occur when 25% of DG is connected at the end bus. The active and
reactive power loss is decreases and mean minimum voltage and maximum voltage
probability is seen increases for each scheme of DG connection.
Bus 1 to Bus 20 is taken to calculate the mean under voltage probability because from
bus 20 to end bus 122 there is 100% probability of under voltage and under voltage
probability is not change although DG is connected at each scheme. 46.2% under
voltage probability at normal condition is reduced to 42.9% at scheme 1 of DG
connection which is more clearly seen by the table 4.11. Hence maximum reduction in
under voltage is achieved when DG is connected at end bus. There is no chance of over
voltage at each scheme although the DG is connected.
43
CHAPTER FIVE
CONCLUSION AND RECOMMENDATION
5.1 Conclusion
In this paper, Point Estimation method is used for Probabilistic load flow analysis of
PV This paper show the comparative result of PEM and MCS method. 42.36 sec
computational time of MCS is reduced to 2.95 sec by PEM and accuracy of MCS is
also maintain by the PEM. Hence for the large system it is efficient and fast to use PEM.
Hourly variation of solar irradiance is consider using beta distribution. Bus number
122,94 and 71 is chosen for PV interconnection. The minimum voltage of 0.8182 is
improve to 0.8513 at scheme 1 when 25% of DG is connected at Bus 122, which is the
best case considering the improvement in per unit voltage. similarly total active power
loss of 541.24 KW is reduced to 419.05 KW and reactive power loss of 628.91 KVar
is reduced to 486.63 KVar when 25% of DG is connected at Bus 122 and 94, which is
the best cases considering the minimizing the Power loss as it has maximum loss
reduction as compare to other cases. The mean minimum voltage and maximum voltage
probability is also improve for each scheme of DG connection.
PV connection in distribution system reduces the branch power loss, reduces the
undervoltage probability but increases the overvoltage probability. The use of PEM in
the real system help to calculate the undervoltage and overvoltage at each buses of the
system which is helpful in maintaining the voltage limit, calculate the maximum and
minimum power loss at each branches and power flow through each branches. Hence
probability analysis is used to view the best cases and worst cases at each buses, which
is used to analyse the uncertainty and maintain the reliability of the system.
5.2 Recommendation
The various opportunities for extending the scope of these thesis are;
[2] Wu, C., Wen, F., Lou, Y., & Xin, F. (2015). Probabilistic load flow analysis of
photovoltaic generation system with plug-in electric vehicles. International Journal of
Electrical Power & Energy Systems, 64, 1221–1228.
[4] Che, Y., Wang, X., Lv, X., & Hu, Y. (2020). Probabilistic load flow using improved
three point estimate method. International Journal of Electrical Power & Energy
Systems, 117, 105618.
[5] Liu Jun S. Monte Carlo strategies in scientific computing. Springer; 2001.
[6] Kroese DP, Taimre T, Botev ZI. Handbook of Monte Carlo methods. John Wiley &
Sons; 2013.
[7] M. Fan, V. Vittal, G. T. Heydt and R. Ayyanar, "Probabilistic Power Flow Analysis
With Generation Dispatch Including Photovoltaic Resources," in IEEE Transactions
on Power Systems, vol. 28, no. 2, pp. 1797-1805, May 2013.
[8] Mensah, L., Yamoah, J. and Adaramola, M., 2019. Performance evaluation of a
utility-scale grid-tied solar photovoltaic (PV) installation in Ghana. Energy for
Sustainable Development, 48, pp.82-87.
45
[11] J. Emmanuel Anandraj, "Point estimate method of Load Flow for distribution
network with photovoltaic generators," 2013 International Conference on Energy
Efficient Technologies for Sustainability, Nagercoil, India, 2013, pp. 24-29.
[12]www.theelectricalportal.com/2017/06/primary-distribution-system-secondary
html
[15] Xiaotian Xu, Xiao Lv, Xinyuan Zhang, Xiaotian Xu," Research on Distributed
Photovoltaic Grid-connected Voltage Cooperative Control Strategy Considering Local
Load," American Journal of Electrical and Electronic Engineering. 2020.
[17] Liu, S., Liu, P. X., & Wang, X. (2016). Stochastic Small-Signal Stability Analysis
of Grid-Connected Photovoltaic Systems. IEEE Transactions on Industrial Electronics,
63(2), 1027–1038.
[18] Nepal Electricity Authority, Course of 2074 related to energy obtained from solar
PV, Nepal Electricity Authority, 2076.
[19]www.researchgate.net/publication/318654444_Design_Estimation_of_Rooftop_G
rid-tied_Solar_Photovoltaic_System
[20] Hong, L., & Goudie, D. (1977). The Square Root Method in Power Flow
46
[22] Delgado, C., & Domínguez-Navarro, J. A. (2014). Point estimate method for
probabilistic load flow of an unbalanced power distribution system with correlated
wind and solar sources. International Journal of Electrical Power & Energy Systems,
61, 267–278.
[24] Conti. S, Raiti. S, “Probabilistic Load Flow for Distribution Networks with
Photovoltaic Generators. Part 1 – Theoretical concepts and models,” International
Conference on Clean Electrical Power, Capri, Italy, pp. 132-136 May 21st-23rd 2007.
[31] Mohammadi, M., Shayegani, A., & Adaminejad, H. (2013). A new approach of
point estimate method for probabilistic load flow. International Journal of Electrical
Power & Energy Systems, 51, 54–60.
47
[32] Ruiz-Rodriguez FJ, Hernández JC, Jurado F. Voltage unbalance assessment in
secondary radial distribution networks with single-phase photovoltaic systems. Int J
Electr Power Energy Syst 2015;64:646–54.
[33] Kroese DP, Taimre T, Botev ZI. Handbook of Monte Carlo methods. John Wiley
& Sons; 2013.
[34] Morales JM, Baringo L, Conejo AJ, et al. Probabilistic power flow with correlated
wind sources. IET Gener Transm Distrib 2010;4(5):641–51.
[35] P. Zhang and S. T. Lee, “Probabilistic load flow computation using the method of
combined Cumulants and Gram-Charlier expansion,” IEEE Trans. Power Syst., vol.
19, no. 1, pp. 676–682, Feb. 2004.
48
APPENDIX- 1 IEEE-33 BUS SYSTEM
IEEE-33 bus system load data and Bus data is as shown in the Table below.
Bus
From To R X P(KW) Q(KVAR)
No.
1 1 2 0.0922 0.0477 0 0
2 2 3 0.493 0.2511 100 60
3 3 4 0.366 0.1864 90 40
4 4 5 0.3811 0.1941 120 80
5 5 6 0.819 0.707 60 30
6 6 7 0.1872 0.6188 60 20
7 7 8 1.7114 1.2351 200 100
8 8 9 1.03 0.74 200 100
9 9 10 1.04 0.74 60 20
10 10 11 0.1966 0.065 60 20
11 11 12 0.3744 0.1238 45 30
12 12 13 1.468 1.155 60 35
13 13 14 0.5416 0.7129 60 35
14 14 15 0.591 0.526 120 80
15 15 16 0.7463 0.545 60 10
16 16 17 1.289 1.721 60 20
17 17 18 0.732 0.574 60 20
18 2 19 0.164 0.1565 90 40
19 19 20 1.5042 1.3554 90 40
20 20 21 0.4095 0.4784 90 40
21 21 22 0.7089 0.9373 90 40
22 3 23 0.4512 0.3083 90 40
23 23 24 0.898 0.7091 90 50
24 24 25 0.896 0.7011 420 200
25 6 26 0.203 0.1034 420 200
26 26 27 0.2842 0.1447 60 25
27 27 28 1.059 0.9337 60 25
49
28 28 29 0.8042 0.7006 60 20
29 29 30 0.5075 0.2585 120 70
30 30 31 0.9744 0.963 200 600
31 31 32 0.3105 0.3619 150 70
32 32 33 0.341 0.5302 210 100
33 60 40
The detail of conductor Type, Transformer size and corresponding length of Tandi
feeder is as shown in the table below;
Tr. Tr.
Conductor Length Conductor Length
SN RN Size SN RN Size
Type (km) Type (km)
(kVA) (kVA)
50
16 17 Dog 0.5 100 77 78 Rabbit 0.5 100
51
41 42 Dog 0.2 100 102 103 Dog 0.8 50
52
The Active and Reactive Power flow and Branch Active and Reactive Power Loss of
122 Buses of Tandi Feeder is shown in Table Below;
53
24 25 409.366 243.262 0.453 0.527
54
48 50 370.713 220.065 0.284 0.332
55
74 75 1717.119 1043.757 2.506 2.909
56
99 100 735.855 441.084 1.468 1.705
57
APPENDIX- 3 CENTRAL MOMENT
Bus
Mean S.D Skewness Kurtosis Mean S.D Skewness Kurtosis
No.
58
21 76.626 7.84 0.0794 1.8102 45.47 7.84 0.0794 1.8102
59
45 25.542 12.7 -0.3698 4.7706 15.16 12.7 -0.3698 4.7706
60
69 25.542 19 0.1456 7.94 15.16 19 0.1456 7.94
61
93 25.542 12.3 -0.824 3.59 15.16 12.3 -0.824 3.59
62
117 25.542 13 0.25 7.94 15.16 13 0.25 7.94
63