Efficient Available Transfer

Capability Analysis Using Linear

Methods
November 7, 2000
PSERC Internet Seminar
Jamie Weber
weber@powerworld.com
Power Systems Software Developer
PowerWorld Corporation
Urbana, IL
http://powerworld.com

What is Available Transfer

Capability (ATC)?

Some of you may be familiar with the terms

Then ATC is defined as

This talk will not cover these terms.

Total Transfer Capability (TTC)

Capacity Benefit Margin (CBM)
Transmission Reliability Margin (TRM)
Existing Transmission Commitments
Etc

ATC = TTC CBM TRM Existing TC

We will really be covering the calculation of TTC,
but lets not get caught up with the nomenclature.
2

Available Transfer Capability

In broad terms, lets define ATC as

ThemaximumamountofadditionalMWtransfer
possiblebetweentwopartsofapowersystem

Additionalmeansthatexistingtransfersareconsideredpartof
thebasecaseandarenotincludedintheATCnumber

Typicallythesetwopartsarecontrolareas

What does Maximum mean?

Canreallybeanygroupofpowerinjections.

Nooverloadsshouldoccurinthesystemasthetransfer
isincreased
Nooverloadsshouldoccurinthesystemduring
contingenciesasthetransferisincreased.

Computational Problem?

Assume we want to calculate the ATC by

incrementing the transfer, resolving the power
flow, and iterating in this manner.

Assume 10 is a reasonable guess for number of

iterations that it will take to determine the ATC

We must do this process under each contingency.

This means we have 10*600 power flows to solve.

If it takes 30 seconds to solve each power flow (a
reasonable estimate), then it will take 50 hours to
complete the computation for ONE transfer
direction!

Assume we have 600 contingencies.

Why is ATC Important?

Its the point where power system reliability

meets electricity market efficiency.

ATC can have a huge impact on market
outcomes and system reliability, so the
results of ATC are of great interest to all
involved.

Example: The Bonneville Power

Administration (BPA)

BPA operates a HUGE capacity of hydro-electric

generating stations

Example: The Grand Coulee Dam has a capacity of

6,765 MW (its one dam!)

Most of BPAs capacity is along the Columbia

River which starts in Canada
As a result, how Canada utilizes its part of the
Columbia River has a huge impact on the ability
of BPA to utilize its Hydro Units along the river

The Columbia River Basin

Canada
Columbia River
BPA

California

Columbia River Basin

The United States and Canada operate the

Columbia River under a Treaty Agreement
To state the Treaty in highly over-simplified terms

Canada has built and operates Columbia River Dams to

the benefit of the United States (i.e. BPAs hydro units)
BPA must make all attempts to give Canada access to
markets in the US (i.e. California)

This means BPA is always trying to ship power

across its system between California and Canada.
Huge amount of money is at stake

During the first 3 months of 2000, BC Hydro sold over

$1 billion in electricity to California!

Linear Analysis Techniques in

PowerWorld Simulator
An overview of the underlying
mathematics of the power flow
Explanation of where the linearized
analysis techniques come from
9

AC Power Flow Equations

Full AC Power Flow Equations

N

Pk 0 V g kk Vk Vm g km cos k m bkm sin k m PGk PLk

2
k

m 1
mk

Qk 0 V b Vk Vm g km sin k m bkm cos k m QGk QLk

2
k kk

Solution requires iteration of equations

V

m 1
mk

P
V
Q

P
Q

J 1

Note: the large matrix (J) is called the Jacobian

10

Full AC Derivatives

Real Power derivative equations are

Pk
VkVm g km cos k m bkm sin k m
m

Pk
Vk g km sin k m bkm cos k m
Vm

N
Pk
Vk Vm g km sin k m bkm cos k m
k
m 1

N
Pk
2Vk g kk Vm g km cos k m bkm sin k m
Vk
m 1

mk

mk

Reactive Power derivative equations are

Q k
V k V m g km cos k m bkm sin k m
m

Q k
V k g km sin k m bkm cos k m
V m

N
Qk
Vk Vm g km cos k m bkm sin k m
k
m 1

N
Qk
2Vk bkk Vm g km sin k m bkm cos k m
Vk
m 1

mk

mk

11

Decoupled Power Flow Equations

Make the following assumptions

k m 0
cos k m 1
sin k m 0

Vk 1

rkm xkm
g km 0

Derivates simplify to
N
Pk
bkm
k m 1

Pk
bkm
m

N
Qk
2bkk (bkm )
Vk
m 1

Qk
bkm
Vm

mk

mk

Pk
0
Vk

Pk
0
Vm
Qk
0
m

Qk
0
k

12

B and B Matrices
P
B'

Q
B ''
V

Define

Now Iterate the decoupled equations

and

P
V B Q
B

' 1

'' 1

What are B and B? After a little thought, we can

simply state that

BistheimaginarypartoftheYBuswithallthe

shunttermsremoved
BistheimaginarypartoftheYBuswithallthe
shunttermsdoublecounted
13

DC Power Flow

The DC Power Flow equations are simply

the real part of the decoupled power flow
equations

Voltages and reactive power are ignored

Only angles and real power are solved for by
iterating

' 1

14

Bus Voltage and Angle Sensitivities

to a Transfer
Power flow was J P
1

Q
solved by iterating V

Model the transfer as a change in the injections P

Buyer: T 0 0 PF 0 PF 0 PF 1
Seller: T 0 PF 0 PF 0 0 PF 1
Assume buyer consists of
N

Bf

Bg

h 1
N

Sx

Sy

z 1

Bh

Sz

85% from bus 3 and 15% from bus 5, then

TB 0

0.85

0.15

Assume seller consists of

65% from bus 2 and 35% from bus 4, then

TS 0

0.65

0.35

T
15

Bus Voltage and Angle Sensitivities

to a Transfer

Then solve for the voltage and angle

sensitivities by solving
S
1 TS
V J 0
S

B
1 TB
V J 0
B

, , V , and V

are the sensitivities of

the Buyer and Seller sending power to the
slack bus
S

16

What about Losses?

If we assume the total sensitivity to the transfer is

the seller minus the buyer sensitivity, then
S B

V VS VB

Implicitly, this assumes that ALL the change in

losses shows up at the slack bus.
PowerWorld Simulator assigns the change to the
BUYER instead by defining
k

Slack S Change in slack bus generation for seller sending power to slack

Slack B Change in slack bus generation for buyer sending power to slack

Then

B
S
V V k V

S
B

17

Lossless DC Voltage and Angle

Sensitivities

Use the DC Power Flow Equations

B

' 1

Then determine angle sensitivities

S

' 1

TS

' 1

TB

The DC Power Flow ignores losses, thus

S B
18

Lossless DC Sensitivities with

Phase Shifters Included
DC Power Flow equations B P
'

Augmented to include an equation that

describes the change in flow on a phaseshifter controlled branch as being zero.

Line Flow Change B B 0

Thus instead of DC power flow equations

we use

B
'

P
0

Otherwise process is the same.

19

Why Include Phase Shifters?

230 kV Phase Shifter
Canada

MD N50 0

DMR 500

ING 500

RO S360

WA H36 0

CU STE R

CB K50 0

SEL 500

NL Y23 0

S NOH OMS 3

SN OHO MS4

CH IEF

&1

MO NRO E
CH IEF
CHI EF

&2

&1

1 000
1 99

C HIEF

C HIEF

E CH OLA KE

MW
MVR

J 5

J O

6 53
10 5

CH IEF

J3

C HI EF
MA PLE

M W
MV R

CO ULEE 19

6 53

MW

1 05

MVR

CO ULEE 20

J4

VL

C OU LEE

S IC KLE R
MAP LE

&1

COU LEE &1

RO CKY

R H

CO ULE E&2

CO VING TN

COU LEE &3

CO VIN GT2
CO ULE E&4

115 kV Phase Shifter

R AVE R

SC HUL TZ

T ACOM A

OL YMP IA

MW

1 77

MVR

30

MW

19

MV R

BP A

B EL L

B &1

S AT SOP

CEN TR

Weak Low
Voltage Tie
To Canada
HO T

B EL L

VA NTA GE

6 70

G1

PAU L

S PR

TA FT

DW OR

D WO R

9 5 MW
-4 5 M VR

D WO R

DW ORS HAK

769 MW
26 MVR

H AN FOR D

H ATW AI
H AN FOR &1

ASH E

A SHE

L OW
2

L OW

MO N

MO N

L IT

GO OS

1155 MW
101 MVR

L IT

LO W

G RAN

LO W

G RAN

GO OS
76 9
2 6

MW
MV R

GA RRI S&4
76 9
2 6

MW
MV R

MW
MVR

G AR RIS& 2

G ARR IS& 3
A LLST ON

GA RRI S&1
S AC JAW EA

BPA
T ROU TDA L

K EEL ER

JO HN

C OY O

D 1

C OY OTE

DAY

JO HN

B IG

SA CJW A

MC NAR Y

DAY

E DD Y

SL ATT

MC LOU GL N

C EL ILO

AS HE

& 1

PE ARL

BU CKL E& 2

O ST RNDE R

B UCKL EY

JO HN

D& 1

A SHE

M AR ION

& 2

BOA RD

30 MW
15 MVR

GR IZZ L& 2

S ANTI AM

GRI ZZL &1

A LVE Y

BO ARD

540 MW
114 MVR

BU CKL E& 1

& 2

RO UND

BU

G RI ZZL Y

L ANE

PO NDRO SA
AL VEY

PON DRO &1

CA PTJ A& 5
GRI ZZL &3

AL VEY

& 1

C AP TJA& 4
G RIZZ L& 4
P ON DRO &2

BU RNS

BUR NS

S UMME R

&2

B URN S

& 1

DI XON VLE

C APTJ A& 3

G RIZ ZL& 5

M AL IN

&2

CAP TJA &2

G RI ZZL &6

DI XONV &1

C AP TJA& 1

GR IZZ L&7

MA LIN

&1

M ER IDI NP

MID POI NT

B ORA H

MI DPO INT

MA LIN

CA PTJ ACK

ADE LAI DE

OL IND A& 1

R OU ND

& 1

115 kV Phase Shifter

California
ROU ND

R D

MT

&2

RO UND

&4

1M

HU MBO LDT

CO YOT ECR

OL IND A&2

R OUN D

M T

V AL MY

VA LMY

G2

IDA HO- NV

AD EL

TAP

Phase Shifters are often on

lower voltage paths (230
kV or less) with relatively
small limits
They are put there in order
to manage the flow on a
path that would otherwise
commonly see overloads
Without including them in
the sensitivity calculation,
they constantly show up as
overloaded when using
Linear ATC tools 20

Power Transfer Distribution

Factors (PTDFs)

PTDF: measuresthesensitivityoflineMW
flowstoaMWtransfer.

Lineflowsaresimplyafunctionofthevoltages
andanglesatitsterminalbuses
UsingtheChainRule,thePTDFissimplya
functionofthesevoltageandanglesensitivities.

Pkm is the flow from bus k to bus m

Pkm
Pkm
Pkm
Pkm
PTDF Pkm
VK
Vm
K
m
Vk
Vm
k
m
Voltage and Angle Sensitivities that were just discussed
21

Pkm Derivative Calculations

Full AC equations

Pkm

VkVm g km sin k m bkm cos k m

Pkm

Vk g km cos k m bkm sin k m

Vm

Pkm

VkVm g km sin k m bkm cos k m

Pkm

2Vk g kk Vm g km cos k m bkm sin k m

Vk

Lossless DC Approximations yield

Pkm

bkm
m

Pkm

bkm
k

Pkm

0
Vk

Pkm

0
Vm
22

Line Outage Distribution Factors

(LODFs)

LODFl,k: percentofthepreoutageflowon
LineKwillshowuponLineLafterthe
outageofLineK
Change in flow on Line L
LODFl ,k

Pl , k
Pk

after the outage of Line K

Pre-outage flow on Line K

Linear impact of an outage is determined by

modeling the outage as a transfer between
the terminals of the line
23

Modeling an LODF as a Transfer

Other Line l

The Rest of System

Create a transfer
defined by
Pn and Pm

Switches
Line k

Pn

~
Pk

m
Pm

~
Assume Pk Pn Pm
Then the flow on the Switches is ZERO, thus
Opening Line K is equivalent to the transfer
24

Modeling an LODF as a Transfer

~
Thus, setting up a transfer of Pk MW
from Bus n to Bus m is linearly equivalent
to outaging the transmission line
~
Lets assume we know what Pk is equal to,
then we can calculate the values relevant to
the LODF

Calculate the relevant values by using PTDFs

for a transfer from Bus n to Bus m.

25

Calculation of LODF
flow on Line L
Estimate of post-outage
~
Pl , k PTDFl * Pk

Estimate of flow on Line K after transfer

~
Pk

~
~
Pk Pk PTDFk * Pk

Thus we can write

~
Pl ,k PTDFl * Pk
LODFl ,k

Pk
Pk

LODFl ,k

Pk
1 PTDFk

Pk
1 PTDFk
Pk

PTDFl *

PTDFl
1 PTDFk

We have a simple function of PTDF values

26

Line Closure Distribution Factors

(LCDFs)

LCDFl,k: percentofthepostclosureflowon
LineKwillshowuponLineLafterthe
closureofLineK
LCDFl , k

Pl , k
~
Pk

Linear impact of an closure is determined

by modeling the closure as a transfer
between the terminals of the line
27

Modeling the LCDF as a Transfer

Other Line l

The Rest of System

Net flow from rest
of the system
Pn
m
Create a transfer
n Line k
~
Pk
defined by
~
Pn and Pm
Assume Pk Pn Pm

Net flow to rest of

the system
Pm

Then the net flow to and from the rest of the system are
both zero, thus closing line k is equivalent the transfer
28

Modeling an LCDF as a Transfer

~
Thus, setting up a transfer of Pk MW
from Bus n to Bus m is linearly equivalent
to outaging the transmission line
~
Lets assume we know what Pk is equal
to, then we can calculate the values relevant
to the LODF.

Note: The negative sign is used so that the

notation is consistent with the LODF
transfer direction.

29

Calculation of LCDF

Estimate of post-closure flow on Line L

~
Pl ,k PTDFl * ( Pk )

Thus we can write

~
Pl ,k PTDFl * Pk
LCDFl ,k ~
PTDFl
~
Pk
Pk

LCDFl ,k PTDFl

Thus the LCDF, is exactly equal to the

PTDF for a transfer between the terminals
of the line
30

Modeling Linear Impact of a

Contingency
1

......

nc

Contingent Lines 1 through nc

Monitored Line M

Outage Transfer Distribution Factors (OTDFs)

The percent of a transfer that will flow on a
branch M after the contingency occurs

Outage Flows (OMWs)

The estimated flow on a branch M after the

contingency occurs
31

OTDFs and OMWs

Single Line Outage

OTDFM ,1 PTDFM LODFM ,1 * PTDF1
OMWM ,1 MWM LODFM ,1 * MW1

Multiple Line Outage

nC

OTDFM ,C PTDFM LODFMK * NetPTDFK

K 1

nC

OMWM ,C MWM LODFMK * NetMWK

What are NetPTDF

K 1

and

NetMWK ?

32

Determining NetPTDFK
and NetMWK

Each NetPTDFK is a function of all the other

NetPTDFs because the change in status of a line
effects all other lines (including other outages).
Assume we know all NetPTDFs except for the
first one, NetPTDF1. Then we can write:
NetPTDF1 PTDF1 LODF12 NetPTDF2 ... LODF1nC NetPTDFnC
nC

PTDF1 LODF1K NetPTDFK

K 2

In general for each Contingent Line N, write

1.0 * NetPTDFN

nC

LODF

K 1
K N

NK

NetPTDFK PTDFN

33

Determining NetPTDFK
and NetMWK

Thus we have a set of nc equations and nc

unknowns (nc= number of contingent lines)
Known Values

1
LODF
21

LODF31

LODF12
1
LODF32

LODF13
LODF23
1

LODFn 1
C

LODFnC 2

LODFnC 3

LODF1nC
LODF2 nC
LODF3nC

NetPTDF1
NetPTDF
2

NetPTDF3

NetPTDFn
C

PTDF1
PTDF
2

PTDF3

PTDFn
C

NetPTDFC LODFCC PTDFC

1

Thus
Same type of derivation shows

NetMWC LODFCC MWC

1

34

Fast ATC Analysis Goal =

Avoid Power Flow Solutions

When completely solving ATC, the number of power

flow solutions required is equal to the product of

The number of contingencies

The number of iterations required to determine the ATC

(this is normally smaller than the number of contingencies)

We will look at three methods (2 are linearized)

Single Linear Step (fully linearized)

Perform a single power flow, then all linear (extremely fast)

Iterated Linear Step (mostly linear, Contingencies Linear)

Requires iterations of power flow to ramp out to the maximum
transfer level, but no power flows for contingencies.

(IL) then Full AC

Requires iterations of power flow and full solution of contingencies

35

Single Linear Step ATC

Foreachlineinthesystemdeterminea
TransferLimiterValueT
Limit M MWM

PTDFM

TM

(infinite)
Limit M MWM

PTDFM

; PTDFM 0
; PTDFM 0
; PTDFM 0

36

Single Linear Step ATC

Then, for each line during each contingency

determine another Transfer Limiter Value
Limit M OMWM ,C

OTDFM ,C

TM ,C

(infinite)
Limit M OMWM ,C

OTDFM ,C

; OTDFM ,C 0
; OTDFM ,C 0
; OTDFM ,C 0

37

Important Sources of Error in

Linear ATC Numbers

Linear estimates of OTDF and OMW are quite

accurate (usually within 2 %)
But, this can lead to big errors in ATC estimates
Assume a lines present flow is 47 MW and its limit is 100 MW.
Assume OTDF = 0.5%; Assume OMW = 95 MW
Then ATC = (100 - 95) / 0.005 = 1000 MW
Assume 2% error in OMW (1 MW out of 50 MW change estimate)
Actual OMW is 96 MW
Assume 0% error in OTDF
Actual ATC is then (100-96)/0.005 = 800 MW
2% error in OMW estimate results in a 25% overestimate of the ATC
38

Single Linear Step ATC

Thetransferlimitcanthenbecalculatedtobethe
TM ,C
minimumvalueof TM or
foralllinesand
contingencies.
Simulator saves several values with each Transfer
Limiters
TM or TM ,C[Transfer Limit]
Line being monitored [Limiting Element]
Contingency [Limiting Contingency]
OTDF or PTDF value [%PTDF_OTDF]
Good for OMW or MW value [Pre-Transfer Flow Estimate]
filtering
Limit Used (negative Limit if PTDF_OTDF < 0)
out errors MW value initially [Initial Value]
39

Pros and Cons of the

Linear Step ATC

Single Linear Step ATC is extremely fast

Linearization is quite accurate in modeling the impact

of contingencies and transfers

However, it only uses derivatives around the

present operating point. Thus,

Control changes as you ramp out to the transfer limit

are NOT modeled

Exception: We made special arrangements for Phase Shifters

The possibility of generators participating in the

transfer hitting limits is NOT modeled

The, Iterated Linear Step ATC takes into account

these control changes.
40

Iterated Linear (IL) Step ATC

Performs the following

1.
2.
3.
4.
5.
6.

Stepsize = ATC using Single Linear Step

If [abs(stepsize) < Tolerance] then stop
Ramp transfer out an additional amount of Stepsize
Resolve Power Flow (slow part, but takes into account all
controls)
At new operating point, Stepsize = ATC using Single Linear Step
Go to step 2

Reasonably fast

Takes into account all control changes because a

full AC Power Flow is solved to ramp the transfer

On the order of 10 times slower than Single Linear Step

41

Including OPF constraints in (IL)

to enforce Interface Flows
When ramping out the transfer, Simulator can be

set to enforce a specified flow on an interface.

This introduces a radical change in control
variables that is best modeled by completely
resolving using the OPF

The objective of the OPF is to minimize the total

controller changes (sum of generator output changes)

Why would you do this?

Represent a normal operating guideline that is obeyed

when transfers are changed.
42

Example: Bonneville Power

Administration (BPA)
Operating procedures for BPA
require them to maintain interface
flows into Seattle in specific ranges
(These are stability constraints!)

1000 MW
199 MVR

653 MW
105 MVR

653 MW
10 5 MVR

67 0 MW
17 7 MVR

30 M W
19 M VR

Seattle
95 MW
-45 MVR

769 MW
26 MVR

1155 MW
101 MVR

769 MW
26 MVR

769 MW
26 MVR

5 MW
2 MVR

Interface
Flow

Chief Jo
2000 MW

Grande
Coulee
6800 MW

1000 MW
199 MVR
1000 MW
199 MVR
653 MW
105 MVR
653 MW
105 MVR

A Lot of
Generation

43

653 MW
105 MVR
653 MW
105 MVR

(IL) then Full AC Method

Performs the following

1.
2.
3.
4.
5.
6.
7.
8.

9.

Run Iterated Linear Step and ramp transfer out ATC Value found
StepSize = 10% of the initial Linear Step Size saved during the (IL) method, or 50
MW whichever is larger.
Run Full Contingency Analysis on the ramped transfer state
If there are violations then change the sign of Stepsize
if [abs(stepsize) < Tolerance] then Stop
Ramp transfer out an additional amount of Stepsize and resolve Power Flow
At new operating point, Run Full Contingency Analysis
if [ (Stepsize > 0) and (There are Violation)] OR
[ (Stepsize < 0) and (There are NO Violations)] THEN
StepSize := -StepSize/2
Go to step 5

Extremely slow.

Number of Contingencies times slower than the

iterated linear. If you have 100 contingencies, then this

is 100 times slower. (1 hour becomes 4 days!)
44

Recommendations from
PowerWorlds Experience

Single Linear Step

Use for all preliminary analysis, and most analysis in

general.

Iterated Linear Step

Only use if you know that important controls change as

you ramp out to the limit

(IL) then Full AC

Never use this method. Its just too slow.

The marginal gain in accuracy compared to (IL) (less

than 2%) doesnt justify the time requirements

Remember that ATC numbers probably arent any more
than 2% accurate anyway! (what limits did you choose,
what generation participates in the transfer, etc)
45