IEEE T r a n s a c t i o n s on PowerApparatusandSystems,

VOl.

PAS-Q6, no. 6 , November/Dece&er 1*7?

CLOSED-LOOP COMPUTER CONTROL O F ASYSTEM OF RADIALLOAD BUSSES, USING TRANSFORMERS, CAPACITORS ANDREACTORS S . A. Arafeh R. E. Kilmer J. H. Rumbaugh Potomac Electric Power Company Rockwell International Corporation Anaheim, California Washington, D. C.

ABSTRACT A new closed-loop control scheme is proposed for the control of a radial network by an on-line computer. The control objectives are the substation bus power factor and voltage regulation, and the balanced loading of transformers operating in parallel. The overall control strategy is decomposed independent into control objectives which are ordered to meet a hierarchy of priority requirements. To reduce the computational effort, the radial load bus network is decomposed into subnetworks.Thecontrol criteriaand mathematicalmodelforeach subnetwork are formulated for the purpose of derivation and simulation of control actions. The procedure which tends to optimize the total process is also included with numerical results. INTRODUCTION Theproblemof providing electricpower to loadcentersat a scheduled voltage in a secure operational environment is one of direct concern to electric power utilities[ 1 I . In the new Potomac Electric Power Company (PEPC0)Consolidated Control System, the automatic control of radial load busses* is t o be provided. The control will be accomplished by using transformer tap position changes and capacitor or reactor switching t o insure adequatepowerfactorand voltage regulation while maintaining essentiallybalancedloading conditions of parallel transformers. This control scheme is intended to monitor and process the real-time power system variables. and t o then transmitthe derived feasible controlactions directly tothesubstation thishas not controlequipment. To the knowledgeoftheauthors, been implemented before. Hkrarchial decomposition of the plysical and conccptual structures of real world systems can be achieved in terms of space, time and bcr'n priority[2,31. This approach power to system problems has recognized[4 0 1 , and is particularlydesirable duetotheir size and complexity. I n t!lis system. the overall network is viewed i n t e r m of its spatial (geographical) characteristics due to the nature of the localized control capability i n each radial load bus subnetwork. Tile :,vera11 c0I:trol strategy is also viewed in terms of its priority o f action of priority requirements ci~aracteristics t l ~ ~ e inherent to the order associated with the control objectives. For example, the control of a !ransforlner leading powerfactorcondltion which could cause the p.rotectire relay to trip is of higher priority than the substation bus vtiltage controi. The contentsthis of paper include the statementthe of problem.the radial loadbusnetworkdecomposition andmodeling,

the required control strategy decomposition criteria, and and a method solution of using heuristic programming. The derivation ofthecomplex radial loadbusmodel, along withtest cases which compare model the solution to the load flow solution, is also provided. The control procedure numerical results are included. STATEMENT OF THE PROBLEM A typical example of a radial load bus networkshown is in Figure 1. This network is composedofinterconnectedsubnetworks,whichconsist ofcombinationsof voltage ratio(VR) transformers, phase angle (PA) transformers**, capacitors reactors and feeding one radial load bus center.

M KVBUS

* * r * u ,

* * * TRANSFORMERS W13.8 KV
13.8 KV BUS

CAPACITORS4 REACTORS

LOAD CENTERS

Fig. 1. A typical radial load bus network. Theintendedcontrol program is t o be executed on-lineon a periodic basis (e.g., every ten minutes) to check all radial load busses in the system and t o formulate the necessary local corrective actions. Theformulated corrective actions arefirst verified by a simulation procedure and are then directly implemented on the physical system. The regulation scheme is devised such that the following control objectives, listed in their order priority, checked and of are for decision rules applied, if possible, to maintain satisfactory operation. within theoperational Priority 1: Leading powerfactor(PF) safety limit of eachtransformer's protective relay. Priority 2: Schedule voltage magnitude center. atthecustomer load

*A distributionsystem in which theload feeders radiate out from independent supply busses.

Priority 3: No MW feedback (circulating real power) in parallel Paper C 75 032-8, recommended andapprovedbythe PowerSystemEngineering Committee of the IEEE Power Engineering Society for transformers. presentation at the IEEEPESWinterMeeting,NewYork,NY,January 26-31, 1975. This paper was recommended for TRANSACTIONS status76 4424 for presentationby title only F OR, for written discussion at the IEEE PES Summer Meeting, Portland, July 18-23, Priority 4: No MVAR feedback (circulatingreactivepower) in 1976.ManuscriptsubmittedSeptember 3, 1974; madeavailable for preprinting parallel transformers. November 7, 1974. This paper has been published inIEEE 1975 Power Engineering from the WinterMeeting, copyright Society WinterMeetingConferencePapers 1975 by the IEEE. Discussion and closure for C 75 0328 have been published in **PEPCO system uses PA transformers which have both voltage ratio IEEE 1975 PESWinterMeetingDiscussions and Closures of Conference Papers, utilization and phase angle control capability to maximize the copyright1975bytheIEEE.Madeavailable for publication in TRANSACTIONS of the system's extensive underground cable network. May 4, 1976. 1731

Priority 5 :

Balanced MVA loading of parallel transformers in proportiontheir to MVA nameplate ratings. close to unityfor each

NETWORK DECOMPOSITION The procedure to solve for the control actions at all radial load busse$ when they are cast in one control problem is exceedingly complex. This is due to the high dimensionality of the problem, whose solution can be computationally prohibitive for real-time environment implementations. Instead, we attempt to solve this high dimensional controlproblem via asetof several low dimensional control subproblemswhich will seek the sameoverall control objectives. This approach is more practical and computationally tractable. Considering system the network spatial structure and local control objectives characteristics, the overall network can be viewed as composedofinterconnectedsetsofsubnetworks. Eachsubnetwork,providedwithindependentlocal controlcapability,consists ofset a of parallel transformers, capacitors reactors and feeding one radial load bus center. Due to the constrained requirement upon the number of localized control actions, and under the assumption* thattheloaddemand remains constantduringtheperiodofdata collection, processing and control implementation, each subnetwork This will allow is decoupled** the from rest of the network. independent and simplified treatment of each subnetwork in formulatingtherequiredcontrolactions.Theformulatedcontrolactions are then regarded as a set of non-interactive local controls with little effect on the overall system operation. The mathematical model for each subnetwork, which is used for the purpose of derivation and simulation of required control actions, is treated in the following section. SUBNETWORK MODELING The radial load bus model, which is necessary to determine the required control actions and simulate their effect, is derivedin this section. A list oftheadoptednotationsandtheirdefinitionsis given in AppendixA. Transformer Voltage Change Due to Tap Movements. The tap changingtransformerrepresentationforthecalculationofvoltage changesof the VRand PA transformers is depictedinFigure 3 as viewed from the transformer secondary with TO= Vo taken as the reference.

priority 6: Lagging powerfactor(PF) radialloadsubstationbus.

In reaching the above regulation conditions, following the physical and operational constraints must be satisfied:
0

The order of priority will be preserved such that the formulated control actions for a lower priority control requirement will not violate any higher priority control requirement. If capacitor orreactor basesapply: a. b. c. switching is needed, the following

Either capacitors or reactors may be switched on-line at a given time. If capacitorsreactors) (or are on-line, reactors capacitors) should not be switchedon line. One on-line switching perday. (or

of a given capacitor or reactor

d. capacitor The or reactor the with lowest number of switching history is chosen.
0

Ifcapacitor orreactor switching is notneeded,thetransformer tap movements operationally are limited to a predetermined maximum number of tap position changes unless the calculated changes exceed the equipments upper or lower physical limits.

In addition t o the above conditions and constraints, the control

program is required to provide the capability for the dispatcher,using a man-machine interface (CRT Display), to:
0

Establish a new systemwide percent voltage change whenever voltage reduction or restoration is required. Receive any alarm message associated with control operation. the system voltage

The overall control scheme structure which realizes the closedin loop radial bus load control the of PEFCO system is shown Figure 2.
PRIMARY DISPATCHER VOLTAGE SCHEDULE MAN-MACHINE INTERFACE

SECONDARY

MESSAGES

Fig. 3. Tap changing transformer representation for VR and PA transformers.

It
ACOUlSlTlON INTERFACE DATA ACTIONS)

RADIAL LOAD BUS CONTROL PROGRAM

SUPERVISORY CONTROL INTERFACE

*This assumption i s 3 s o necessary fortheproperperformanceof

the closed loop control. Its effect is to insure that control actions are applied to the same monitored operating conditions, The operatingconditionsremainessentiallyinvariantdue t o t h e fast control program execution. **The goal coordination concept in Hierarchical System Theory[2,3] was not enforced under assumption decoupling here the that radial subnetworks supplying instantaneousconstantloadshasa small effect on the systems overall control objectives.

OUTPUT (CONTROL (MEASUREMENTS)

Fig. 2. Cl~~sed-loop rudiul load bus control of the PEPCO system.

1732

The secondary internal voltage vector diagram is shown in divided into *I6 tap Figures 4 and 5. The range of tap position is + I O percentvoltagechange of theneutral movementsresultingin setting voltage Tn. The change due to a given tap positionsetting is added in phase with Tn for a VR transformer and 60' out of phase for a PA transformer. In terms of Tn, the voltagedifferencefrom the neutral setting ATn can be represented by For VR transformers ATn = For PA transformers and the voltage increment (decrement) (1)

formers,

In order to obtain a unified model for the VR and ii is defined by:

For VRtransformers

PA trans-

a=

[A 9) (3
+j

For PA transformers

Thus, F.q ( I ) and (2) are reduced to

ATn = GOTn

A by T
For VR transformers and the transformer secondary can be represented by the equivalent circuit diagram shown in Figure 7.

AT =

n-no160 vn

n-"o 160

(i+ jq) Vn

(2)

For PA transformers

SECONDARY

Fig. 4. Voltage vector diagram for a VR transformer: Fig. Z Secondary equivalent circuit for VR and PA transformers. Neutral Setting Tap Internal Voltage. To calculate the internal Tn from theactualmeasurements, voltage forneutraltapsetting consider the circuit diagram of Figure 7 with AV = 0. Since

V,

+ ATn - Vo
S*

TIT,

(4) (5)

= VoT

and by substituting for Avn from

p '

Eq (4), Vn is determinedfrom

Fig. 5. Voltage vector diagram f o r a PA transformer.

This change in voltage is used in the transformer model representation as shown in Figure 6. A tap increase or decrease is presented not by changing a, but by inserting the voltage change AT, given by Eq (2), in series with the high-low transformer admittance

v.

Substation Bus Voltage. To calculate the change in the substation bus tap consider the voltage ATo due to transformer movements, equivalent diagram circuit of the radial load subnetwork bus represented in Figure 8. Sinas the changestransformer in current flow A\ and the substatio~: currentflow Ais can be expressed by

AT = ATi

n
bY
Fig. 6. Transformer representationfor voItage change due to tap movementx 1733

(7)

and the substation bus admittance

Ys and ATs are

given

where
0 when capacitororreactor

not considered

sw tching i s
is considered

Since 3 = VoT* and by using Eq (7), the incremental change in the q transformer power flow can be obtained from E (1 3) as:

Q=

f C R when capacitororreactor

AS =

(Vo

AVOS + ATo) (AV - ATA* 'J* + "0

t o be switched on

i C R when capacitororreactor

t o be switchedoff the change i n the substation voltage output t a i ned from:

is considered
( 0)

Load Center Voltage. To calculatethe voltage attheloadcenter, consider the substation-feeder-load equivalent circuit diagram shown in Figure 10.

can obbe

CAPACITOR/ REACTOR

Fig. 10. Substation-feeder-load center equivalent circuit diagram.

Due to the voltage drop across the feeder impedance the voltage at the load center is calculated as follows:

v,

E,,

V,VL '

= 7JL

I
Fig, 8. Equivalentcircuit diagram of a radial load bus subnetwork.

IL=

v0vf l

and

It should be noted that the calculation of YS is valid under the assumption that the power load demand is not changing during the processing and implementation of the control actions. Transformer Power Flow. As a result of transformer tap changes, a change in power flows will occur. These changes can be calculated by using the parallel transformer model shown in Figure8, which treats a tap change as an added voltage source. The equivalent circuit diagram of one transformer is depicted in Figure 9.

Therefore

VL = V0(l

- YfYfl)

and

(18)

A :

- Avo

4 -

The loadcentervoltage

change ATL is calculated

by assuming a small perturbation i n Eq(17) such that

Fig. 9. Equivalent circuit diagram o f one transfomer.

and by usingEq(9) t o substitute for substitute for Vfl, we obtain

The power flow of one transformer to the substation bus due to changes in tap position is given by
-1

Vo and Eq(15) t o

(vo + A V ~ )(i + ai)*

(1 2)

1734

A summary of the derived mathematical equations for the radial is given in Table 1. The vector diagram load bus subnetwork model describing the voltagechanges as aresult of transformer tap movements and capacitor or reactor switching is represented in Figure 11. Simulation procedures which validate this model are presented in the next section. Table 1. Mathematical Formulation of a Radial Load Bus Subnetwork Model DEFINITION iOTATION
1 -

FORMULATION for 16-0-16 VR transformer

Transformer Type Constant

160

r (I + o 3

JT
'

")

for 16-0-16PA transformer Fig, 11. Vector diagram representing the voltage changes.
RADIAL LOAD BUS

Transformer Internal Voltage at Neutral Tap hnsformer tnternal Voltage Change per Ta Substation Load Admittance

GENERATION BUS

AV;

F2

iq
I
84

Substation Bus Voltage Incremental Change

AVO

F4

)Y
PA4

Transformer Power Flow Incremental Change

(Vo + ATo) AniAVi - AVO) Yi ASi

- *-*

Fig. 12. A radial load bus subnetwork one-line diagram used in the simulation procedures. four feeders, two VR transformers, two PA transformers a and capacitor bank feeding a radial load bus center. Table 2 represents the model'sparametersdataandtheinitialvaluesofthebasecaseas obtained from a load flow solution. Results Three simulation tests were conducted different for transW n g s . The valuesofthecomplexchangein formertapposition bus voltage. and transformer power f l o w are obtained from the radial load bus model solution and the load flow solution independently. To compare these values the absolute percent error, relative to the load flow solution, is calculated in each case. These results are summarized in Tables 3, 4, and 5 which show a range of error of 0.3 percent to 0.48 percent for the bus voltage flows. and 0.02 percent to 4.16 percent for the transformer power

Load Center Voltage

Load Center Voltage Incremental Change

AV~

SIMULATION Sample Problem

CONTROL DECOMPOSITION AND CRITERIA In order t o test the radial bus load mathematical model performanceincomparison to aloadflow solution,anadaptation of a PEFCO subnetwork is used. The one-line diagram of this subneta generation bus, work is shown in Figure 12 and consists of In solving for the radial load bus subnetwork control problem, i t may not be possible to s a t i s f y all control objectivesduring one control cycle. It is then desirable for the control program to provide

1735

Table 2. Radial Load Bus Data and Base Case

CONTROL FUNCTION 2: Load Center Voltage Control. For a specifiedvoltagescheduleVsch,toleranceVtol,andcalculatedload center voltage VL, transformers, capacitors, and reactors will be controlledsuch thatthe changeinload center voltage AvL satisfies:

CONTROL FUNCTION 3: Transformers MW Feedback Control. The ith transformer will be controlled to satisfy

Re ( 4 ) 2 0

apartialfeasible control, which will meethigherpriorityobjectives in their decremental order, rather to than attempt to satisfy all objectives which would not be feasible. To achieve this capability it is necessary to decompose the overall control strategy into independent control functions. Therefore, based on control the objectives and of their inherent order priority characteristics, the independent control functions with their criteria can be described as follows: CONTROL FUNCTION 1: TransformersLeadingPFControl. For a specifiedleadingpower factor anglelimit81im,transformers, capacitorsandreactors will be controlledsuchthattheithtransformer complex power flow satisfies:

trol.

CONTROL FUNCTION 4: Transformer MVAR Feedback ConThe ith transformer be controlled to satisfy will

g m (Si) Abs (Si)

fori = 1, 2,
*

-<

sin elim

*, m

CONTROL FUNCTION 5: Transformers MVA Balance Loading Control. For a specified MVA tolerance, MVAtol, the ith transformer will be controlled to satisfy

Table 3. Numerical Results of Case (1) [An, = 4, An2=An3 = An4= 01 VARIABLES Radial Load Bus Model Solution Load Flow Solution Absolute Error (%) 1.02 19

vo p.u.

s, p.u.
0.1 133

s2p.u.
1.01846

s 3 p.u.

s4p.u.

+ j 0.O430l

- j0.017453
~ ~ ~~~~~

- j 0.016850.09971 0.095268 - j 0.01276 0.095777 0:095 - j 0.017 0.099 -jO.O17 0.096 1.4 0.7 0.6

0.112 +jO.O44 0.7

- j 0.018

0.3

VARIABLES Radial Load Bus Model Solution Solution Flow Load Absolute Error (7%) 0.999

vo p.u.

s 1 p.u.

s 2 p.u.

H3

p.u.
~~~~~

s4p.u.
0.12311 +j0.01081

0.12967+j0.006407 0.12968 +jO.O0642 0.011758-jO.027602

0.994+j0.008 10.131 10.14 +j0.008 10.124 10.131 +jO.Ol -j0.028 0.48 1.1
1.oo

1I

0.072

4.16

Table 5. Numerical Results of Case (3) [An 1 = 4, An2 = 2 = An3 = -8, An4 = 51 VARIABLES RadialLoad Bus Model Solution Load Flow Solution Absolute Error (%)

vo p.u.
1.0033 1.0076 0.42

SI p.u.

s 2 p.u.

s 3 p.u.

s 4 p.u.
0.17455 + j 0.006167 +jO.O06 0.171 2.08

0.116309+j0.00782 0.123 +jO.O39

0.115
+jO.O08

0.12326+j0.0380 4.009124-jO.0565 -0.009 -j0.059

1.12 1736

0.02

4.0

I1 + mAbS
j= 1

- 1 < MVAtol

Abs ($1

programming approach, problem. In contrast tothe mathematical heuristicprogrammingdoes notguarantee an optimalsolutionbut to makes full use of the problemsphysicalcharacteristicsinorder yield a feasible solution. The derived which implemented rules, are programming, are described as follows: by the heuristic

fori=1,2;..,m where ki is the ith transformer distribution factorbased on the parallel transformers MVA ratings MVAgivenby

MVAl$.
ki= m

factor, switching of a For a lagging or a leading load power improves the substation bus capacitor on-line (or off-line) power factor and causes the voltage magnitude to be increased (or decreased). For a lagging or a leading load power factor, switching of a reactor off-line (or on-line) improves.the substation bus power factor and causesthevoltagemagnitude to beincreased (or decreased). Increasing (or decreasing)the tapmovementsof all parallel transformers increases (or decreases) the bus voltage magnitude.

MVA9 j= 1 fori=1,2;*-,m CONTROL FUNCTION 6 : Substation Bus Lagging PF Control. A capacitor or reactor will be switched on/off if there exists a capacitor or reactor with Q, given in Eq (1 l), satisfying

< C 9 m (si)
i= 1

MW and MVAR Feedback can be decreased by increasing the tap position on the feedback transformers.
PA transformerscontrolboth realandreactivepowerflow, while VR transformers can only control reactive power flow. MVA loading of a transformer is increased (or decreased) by increasing (or decreasing) the tap position. Tapchangesin the raisedirectionoffsettapchangesinthe of keeping the voltage lowerdirection.Thishastheeffect magnitude essentially constant at the bus bar while changing the power distribution in parallel transformers.

The above independent control functions, casted in a hierarchial on transstructure, are shown inFigure13.Respectiveconstraints former tap position changesandcapacitor or reactor switching, are assumed to be imbeddedwithineachcontrolfunction.Thecontrol procedure which implements these functions is described in the next section.

TRANSFORMERS LEADING

PF CONTROL

- ---

1ST PRIORITY

1
LOAD VOLTAGE CENTER CONTROL

---- 2ND PRIORITY

----3RD PRIORITY

1
TRANSFORMERS MW FEEDBACK CONTROL

5TH PRIORITY

The controlalgorithm logic using heuristic programmingis shown in Figure 14, which consists of controlmodules,andasimulator routine. The logic is set up such that the control modules will implement the control functions, described in the last section, by using the above rules to formulate the required control actions. The simulator routine uses the radial load bus subnetwork model, Table 1, to calculate the effect (power factor, voltage and power flow changes) due t o the formulated set of control actions (transformer tap position changes andcapacitor or reactor switching) as received from each control module one at a time. This simulation is required to determine the validity of the formulated control actions in correcting detected deviations and in preserving the order of priority requirements. Assuming valid data is obtainedthroughthedataacquisition interface control the algorithm is activated as follows. Starting at the fint control module, the control actions are formulated in the samesequenceastheirorderof priority. If afeasiblesolution is obtained satisfying the first order of priority objective, the second control module will be effected to seek an upgraded solution satisfying the second order of priority objective. The process is continued until a feasible solution cannot be found, due to control constraints, or until all control objectives have been satisfied. The result, whether a partial or a total feasible solution, is directly implemented on the substations equipment through the supervisory control interface. SUMMARY AND CONCLUSION A real-time closed-loop voltagecontrol scheme with the capability of controlling out-of-limit power factor, and MVAR feedback, and MW MVA imbalanceloading of paralleltransformers is presented. The

1
SUBSTATION BUS LAGGING PF CONTROL

---- 6TH PRIORITY

Fig. 13. Radial l w d bus overall control strategy hierarchy.

CONTROL ALGORITHM

In order to devisea control algorithmwhichmeets the radial load bus control objectives, a mathematical or heuristic programming is can be adopted. In this effort, a heuristic programming implemented which makes use of the physical properties of the problem to obtain a manageable procedure via rules of thumb derived from these.properties. That is, the search for a better solution conducted through a set is of feasible directions defined by the physical properties the of

1737

+
TRANSFORMERS LEADING PF CONTROL FEASIBLE SOLUTION

feasible solution should be provided to meet higher priority objectives rather than attempting a solution to meet all control objectives which cannot be realized. Thismethod uses a heuristic programming approach which manipulates a set of rules derived from the problem's physical properties to yield an improved solution. The basic radial load bus mathematical model, which is necessary for the formulation andsimulationofrequiredcontrolactions,has beenderived.Severaltestcaseswereconductedandcompared to a load flow solution. In these tests good results were obtained in support of the derived model. The advantage of the model is that it provides an easy to handle, explicit representation of the problem's variables within a closed-form (non-iterative) mathematical solution. Dynamic simulation test cases were also performed to measure the overall algorithm effectiveness in tracking various operating conditions. The results these tests are not included in this paper due to of space limitation. However, a sample test case is given in Appendix B which shows the excellent tracking capabilities the algorithm withof in the imposed limits.

FEASIBLE NO SOLUTION EXISTS lYES

ROUTINE 4

4YES TRANSFORMERS MVAR FEEDBACK CONTROL FEASIBLE SOLUTION

I II

APPENDIX A - RADIAL LOAD BUS MATHEMATICAL NOTATIONS

Number of transformers in parallel Transformer transformation ratio Transformer type constant Tap setting before tap movement

MVA TRANSFORMERS BALANCED CONTROL FEASIBLE SOLUTION EXISTS?

SUBSTATloNF C O M R O L LAGGING P Bus

I
I

Tap setting after tap movement Incremental (decremental) tap movement Measured substation bus output voltage at the secondary side (also considered the reference voltage) Transformerphase to neutralsecondaryinternal for the neutral tap setting voltage

EXECUTE FORMULATED CONTROL ACTIONS

Fig. 14. Control algorithm logic using heurktic programming.

Transformerphase to neutralsecondaryinternal for the nth tap setting. Internaltransformervoltagedeviationfromtheneutral tap setting

voltage

scheme, which is being implemented for the radial load bus network of the PEW0 system, is intended to have direct control of the substation control equipmentswhile interacting with the dispatchert o receive the required voltage reduction or restoration information. In the course of solution, the control problem has been decomposed to simplify and reduce the required computational efforts by taking full advantage of the posed problem characteristics. Onecharacteristicoftheproblem is thatthe period of data gathering, processing and control implementation is small relative to the normal load changes. This is necessary for the proper performance of the closed-loop control in applying control actions which match the monitored operating conditions. Further, due to the constraints imposed onthe localized controlactions,the overall network was decomposed sets interconnected into of subnetworks which were decoupled and treated independently simplify the problem solution. to
of Another Zharacteristic, which led to an additional form decomposition, is the priority associated with each control objective. Thus,the originalnetworkproblem was furtherdecomposed to its basic elements of control functions in accordance with its operational priority requirements. This is not only to simplify the control problem solution,but is alsonecessitated by therequirement that apartial

Transformer phase to neutral increment (decrement) due totap movementfromthe n$ tothenthtapsetting Load center bus voltage schedule Load center actual voltage Load center incremental (demmental) change Transformer high-low admittance Substation load admittance Feeder to the load impedance Transformer complex power flow Transformer current flow Change in transformer complex power flow Change in transformer current Substation current flow

1738

ATo
A is
C

Change in substation bus voltage

REFERENCES
[ 1 1 Ku, W.S., and Van Olinda, P. Security and Voltage Applications of the Public Service Dispatch Computer. IEEE, PICA Conference Proceedings, 1969.

Change in substation current flow

Capacitor or reactor reactive power MVAR rating

[2IMesarovic, M.D., Macko, D., and Takahara, Y . , Theory of Hierarchical Multilevel Systems. Academic Press, 1970.

Capacitor or reactor reactive power MVAR calculated value

[3] Arafeh, S.A., On the Multilevel Hierarchical Decomposition and Coordination Methodsin Large-Scale Systems, Ph.D. Dissertation, Chapter 11, SMU, 1973, also MSAC Proceedings, April, 1975.
[4]DyLiacco,T.E.,ControlofPowerSystems via Multilevel Concepts, Ph.D. Dissertation, Case Western Reserve University, 1968. [SISchweppe, F.C. andMitter, S.K., HierarchicalSystem Theory and Electrical Power System. in Hadschin, F. (Ed.), Real Time ControlofElectricPowerSystems, Elsevier, Amsterdam,1972. [ 6 ]Narita, S., and Hammam, M.S.A.A., Multicomputer Control of SystemVoltage and Reactive Powerona Real-Time Basis, 1971 PICA Conference Proceedings.

Transformer leading power factor angle limit

Transformer distribution factor

Bus voltage tolerance

MVA balance tolerance APPENDIX B

CONTROL ALGORITHM TRACKING RESULTS

MVA

l1

I 1

Control Cycles

3
1739

Discussion St. Johns, Newfoundland, Canada): Thisis an exceedingly interesting paper in that it addresses a problem which is security-reliability oriented in a systemtheoreticcontext. This discussor likes to make some commentsto which the authors response is appreciated. In virtually all themodelsreportedinthepapersecondorder terms in the voltage increments are retained. It is worthwhile to consider neglecting these terms in favour of a much easier problem formulation. Did the authors experiment with this? In case they did, what order of magnetidue of errors does this incur. The six controlfunctions considered by theauthorsdefine six distinctsetsinthespaceofthevariablesconcerned.Obviously,the authors are interested in finding an element which ideally lies in the intersection of these sets. In case of an empty intersection, they would drop the set of lowest priority, and so on. This is perhaps a logic which will work provided one finds only one solution. The authors indicate that their method does not attemptto find the so called best or optimal alternative in the feasible region. This is based on the dimensionality of the problem and more importantly on the nonlinearity of the chosen model. In the light of the previous paragraph, would linearization enable an algorithm which strives to find a feasible region in an efficient manner? The authors should be congratulated for a clear and well written paper. Manuscript received August 5, 1976.

M. E. El-Hawary(MemorialUniversityofNewfoundland,

actions are derived based on the uptodate measurement set which reflects the last control cycle response in its steadystate condition. 2. Regarding the capabilities (a) through (d), stated by Dembecki and Couch, we like to note that these capabilities were not, categorically, required by the PEPCO system. However, capability (a) can be conceptually implemented as shown in Figure 15. The other capabilities can be easily adopted by modifying the mathematical model, summarized in Table (I), to include the additional elements such as the shunt capacities of cables. Furthermore, the present control program could be modified to includeadditionalroutines,usingasimilarheuristicapproach, to enable the extended control capabilities of switching lines and transformers in the proper manner.

u
DISPATCHER MESSAGES REDUCTION RESTORATION

WN MCHINE

INTERFACE

H
__c

SYSTM UIDE VAR/VOLTASE COllTROL PROGRAM

DATA ACQUISITION

INTERFACE

RADIAL LOAD BUS

COHTROL PROGRAH

SUPERVISORY CWTROL INTERFACE

J. A. Dembecki and G . H. Couch (Electricity Commission of N.S.W., Australia): The authors have presented an elegant solution to the problemofinstalling an automatic control system for a distribution network. The method would be suitable for adaptation in other systems. The concept of hierarchical control objectives has a close parallel in more usual manual controlprocedures. Principally, in theshort term, three levels of objective may be considered to apply; first, prctection of plant; second, quality of supply; third, economy of operation. The priorities used by the authors is consistent with this policy. Much of the papers value lies in the authors having developed an heuristic,computationallysimplealgorithmforafrequentlyencountered type of distribution system. Comments regarding other systems would be of interest; for example ring main systems, as distinct from radial networks. Has any consideration been given regarding extension of the algorithm to include a capability for: (a) providing a scheduled Mvar draw on the transmission network? (b) switching lines or cables to reduce losses or for voltage control if necessary? (c) taking account of the shunt capacitance of cables? switching lines and transformers from standby in the event of main equipment failure?
Manuscript received February 6,1975.

PCUER SYSTEM

I
Fig. 15. Control Adaption of the Scheduled MVAR Transmission Network along with the RLB Network.

Samir A. Arafeh, Robert E. Kilmer, and Jefferey H.Rumbaugh: The authors wish to thank the discussors for their interesting remarks and the important suggestions they have made. To answer the points they have raised the following comments are given: 1. The same control procedure hasbeen tested for aring main system and acceptable results were obtained. These results were expected due to the physical nature of the process, that is, the small incremental changesintransformerstapmovementsandtherelativelysmall size capacitor switching. This has little effect on the total system which as a result damps any interaction between subsystems in the same ring. It should also be noted that at every new control cycle the new control

3. Second order terms in the voltageincrements,werenot neglected due to thefollowing reason. The trend toward using a rigorous (as practically as possible) model is to lessen the compounded errors, especially when applied to real-life systems. Such errors could be attributed to measurements and parameters inaccuracies in addition to not accounting for any interaction between adjacent radial substations. Thus introducing model approximations (errors) in favor of a much easier problem formulation was unnecessary forcomputerapplications.This is due to that the control actions were obtained via closed-form (non-iterative) solutions, which appreciable savings in computer overhead, if these second order terms were not retained, could nothave been realized. 4. In regard to finding an optimal solution from the set of feasible solutions, as raised by El-Hawary, the dimensionality of the problem was not viewed as the limiting factor. This is due to that each radial substation is solved separately which kept the dimensionality of each control problem relatively Also, low. thetendencyto linearize the problem formulation may further be complicated by the fact that the control.actions arediscreteinnature(transformertapmovement or capacitor switching). Instead,non-linear a integer programming ap proach, where the ultimate goalofunitypower factor could be the objective function, keeping in mind the order of priority if necessary, may be the answer to this question. This approach has been discussed in Reference [ 71 .
REFERENCE
[ 71 Arafeh, S. A,: Application of Hierarchical Control Concepts to a

Manuscript received March 30, 1977.

Large Power Distribution Francisco, June, 1977.

System. JACC Conference in San

1740