Loss Reduction of Power Distribution NetworkUsing Optimum Size and Location of Distributed

GenerationAdnan Anwar, Student Member, IEEE, and H. R. Pota, Member, IEEE

Abstract—Distributed generation (DG) units reduce electricpower losses and hence improve reliability and voltage pro-file. Determination of appropriate size and location of DG isimportant to maximize overall system efficiency. In this paper,an optimization method has been presented to determine theappropriate size and proper allocation of DG in a distributionnetwork. Results obtained from this method have been comparedwith using the repeated load flow method. A new approach toperform repeated load flow by using simulation engine ‘openDSS’COM server through Matlab programming is also introducedhere. Both optimized method and repeated load flow basedmethod have been compared for three IEEE distribution testsystems. This analysis shows that using appropriate size andlocation of DG, total power loss in primary distribution systemcan be reduced significantly.

Index Terms—Distributed generation, Optimum size, Optimumlocation, Power loss, Sensitivity analysis.

I. INTRODUCTION

The growth of electricity demand is increasing rapidly. DGis one of the better alternatives to fulfil this ever growingenergy demand. Moreover, it reduces system energy loss,alleviates transmission congestion, improves voltage profile,enhances reliability and provides lower operating cost. Be-cause of small size compared with conventional generationunits, DG is more flexible to install in terms of investment andtime. As a result, integration of Distributed Energy Resources(DER) with distribution network offers a promising solution;therefore, an intensive level of research is needed to understandthe impacts of distributed resources on Distribution System.Before operating distributed and dispersed generation in powersystem, different technical, environmental, commercial andregulatory issues should be analysed properly. Most significanttechnical barriers are protection, power quality, stability andislanding operation. However, there are some other issueswhich should be analysed before to maximize these technicalbenefits. From previous studies, it has been seen that differentpenetration level and various placement of DER will impactthe distribution system differently [1]. Moreover, improperDG size and inappropriate allocation of DER may lead tohigher power loss than when there is no dispersed generationin the system at all [2]. Therefore, detail and exact analysismethod is required to determine the proper location and sizeof DG more accurately and precisely. In distribution system,

Adnan Anwar and H. R. Pota are with the School of Engineering and Infor-mation Technology (SEIT), The University of New South Wales at AustralianDefence Force Academy (UNSW@ADFA), Canberra, ACT 2600, Australia.E-mail:Adnan.Anwar@student.adfa.edu.au and h.pota@adfa.edu.au.

DG should be allocated in an optimal way such that it willreduce system losses and hence improve voltage profile [3].In our study, we will try to focus on optimum location andsize of DG to decrease total system power loss. In mostof the previous researches of DG sizing and allocation, DGhas been connected with grid directly. Significant risks areassociated in connecting such equipment directly to utilitydistribution system. The insulation level of the machines maynot synchronize with the system. Therefore, direct connectionof DG is often discouraged [4]. In our analysis, we haveconnected DG through a three-phase step down transformer.

The organization of this paper is as follows; Definitionof DG is discussed in section II. Section III describesdistribution losses with the variation of DG size and location.Previous analysis methodology is discussed in section IV.Objective of our analysis is given in Section V. Our analysisMethodology is shown in Section VI. In Section VII, testsystem and analytical tools have been discussed. Simulationresults and discussion have been presented in section VIII.Evaluation of estimation performances have been analyzed andcompared in Section IX. Finally, the paper is concluded bybrief remarks in Section X.

II. DEFINITION OF DISTRIBUTED GENERATION ANDRATING OF DG (SIZE)

Generally, Distributed generation means the electric powergeneration within distributed network to fulfil the rapid energydemand of consumers. However, distributed generation can bedefined in a variety of ways.

1) The Electric Power Research Institute (EPRI) definesdistributed generation as generation from ‘a few kilo-watts up to 50 MW’ [5].

2) International Energy Agency (IEA) defines distributedgeneration as generating plant serving a customer on-siteor providing support to a distribution network, connectedto the grid at distributed level voltages [6].

3) The International Conference on large High VoltageElectric Systems (CIGRE) defines DG as ‘smaller than50-100 MW’ [5].

Although there are variations in definitions, however, theconcept is almost same. DG can be treated as small scalepower generation to mitigate the consumer energy demand.Distributed Generation can come from a variety of sourcesand technology. To analysis the DER impacts, different typesof ‘generator groups’ can be considered [7]. Here, we will

consider induction generator as distributed generation sourcefor our analysis purpose.

As the technical design of each distribution network isunique, therefore, it can not be answered what should bethe optimum generation capacity or rating of DG [5]. Themaximum size or rating of DG which can be connected toa distribution network depends on numerous factors, such asvoltage level within the distribution system, power loss profileand other technical, environmental, commercial and regulatoryissues. In our paper, we will focus on the technical issues only.

As DG offers lots of benefit, the penetration of DG indistribution system is increasing rapidly. Therefore, DG shouldbe allocated in an optimal way to maximize the systemefficiency.

III. DISTRIBUTION LOSSES WITH THE VARIATION OF DGSIZE AND LOCATION

In a distribution system power loss varies with numerousfactors. Real power losses of a distribution system depend onthe resistance of distribution lines, core losses of transformersand motors. As dielectric and rotational losses are so smallcompared with line losses, therefore, only line losses areconsidered in this analysis. The complex power Sij from nodei to j and Sji from node j to i are

Sij = ViI�ij (1)

Sji = VjI�ji (2)

Where, Vi and Vj are the voltages at node i and j respec-tively. The line current Iij which is measured at bus i in thepositive direction of i to j and Iji which is measured at bus jin the positive direction of j to i. Therefore, power loss in anyline between node i and j can be written as the algebraic sumof power flows determined from (1) and (2) [8].

SLij = Sij + Sji (3)

After any converged loadflow, power loss in any line canbe calculated using (3) and taking the summation of all linelosses, total power loss of the network can be calculated usingequation (4) where n is the number of lines.

Loss =

n∑

k=1

SL(k) (4)

For any distribution system, placement of any DG unit willchange the power loss profile of that system. Actually, ina distribution network, power loss curve with the variationof power generation at a particular location is approximatelyquadratic function because Line Losses ∝ I2R and I ∝ Sconsidering I is the line current, R is the resistance, and S isthe apparent power flowing through the line [9]. Therefore, asthe DG size is increased in any location of a power distributionnetwork, the total system losses are reduced to a minimumvalue. With further increasing of DG, losses again start toincrease. This trend of losses with DG size variation is givenin Fig. 1 for a test case to demonstrate the sizing and locationissues of DG. Here, for DG size PDG2 we get the minimumpower loss which is called optimum DG size for that bus.

Size of DG (kW)

Pow

er L

oss

(kW

)

PL3

PL1

PL2

PDG1 PDG2 PDG3

Fig. 1. Power loss characteristic of distribution system with DG size variation

Actually, the structure of distribution system is such thatpower should flow from the substation to the consumer end andconductor sizes are also decreased gradually [3]. When a DGis placed in the network, it is desirable that power should beconsumed within the distribution network and thus improvespower profile. Any size of DG more than the optimum size willcreate reverse flow of power towards distribution substation.Therefore, excessive power flow through small sized conduc-tors towards the transmission area will increase the power lossin distribution network.

IV. PREVIOUS METHODOLOGIES

A good number of research work is going on DG integrationwith grid and its safe and reliable operation [3], [10-17].However, only a few studies have been done on DG sizingand allocation issue. Different methodologies to determineoptimum location and size have been discussed in differentliteratures. The 2/3 rule is often used in capacitor allocationstudies in power distribution network. Similar approach can beperformed in DG allocation to reduce system power loss [10].In the paper [10] authors have used this analytical method andrule of thumb for analysing the distribution system which isradial and has uniformly distributed loads. Rule is simple andeasy to use but it cannot provide the proper solution whenthe load distribution type is changed. Moreover, it can not beapplied in meshed network.

In [11], analytical approaches for both radial and networkeddistribution systems with different types of load configurationare given. Here, separate algorithms have been used for radialand meshed networks. To simplify the analysis, authors haveconsidered only overhead lines for which uniformly distributedparameters like R and L per unit length are same along thefeeder. Results obtained from the analysis are very quick; how-ever, one generalized algorithm is expected for both radial andmeshed networks. Besides, in practical distribution system,conductor sizes are gradually decreased from substation toload centre, therefore, this analysis procedure would be verycomplex when line parameters are not uniformly distributed.One major limitation of this approach is, they have only solvedthe location problem for a fixed size of generator but they havenot considered DG sizing issue in their analysis.

Another analytical approach has been proposed based onexact loss formula in [3]. Authors have considered the losscoefficients constant. Here they have considered both sizingand location issues. This process takes only two load flows todetermine the location and size of DG. Although the techniqueis very fast; however, this methodology can be applied only ifDG delivers real power [12]. This is one major limitations ofthis approach. For load flow, authors have considered Newton-Raphson algorithm. Although Newton-Raphson approach hasan excellent convergence characteristics but in distributionsystem because of smaller X/R ratio it can not be decoupled.Moreover, in distribution network, multi-phase, unbalancedoperation, unbalanced distributed load and dispersed gener-ation makes the Newton-Raphson approach unattractive [13].

For selection optimum size and location of DG, severalgenetic algorithms (GA) and fuzzy logic based methods havebeen discussed in [14], [15], [16], [17]. Although GA providesalmost near optimum output but they are computationally verydemanding and have a slow convergence [3].

As load flow represents the system states, therefore it canbe used for planning the future expansion of power systems.We can calculate the system loss from the load flow resultand doing the load flow repeatedly, we can easily tell thelocation and size of DG for which we get the minimumpower loss of the system. This method is known as exhaustiveload flow (ELF) method. Although this ELF method gives theexact answer; however, it needs lots of load flow computation.Therefore, ELF method needs to be optimized to get accurateanswer and less computational time.

In the previous literatures, researchers have consideredradial distribution system but they have not considered threephase unbalanced system. As, distribution system is threephase and unbalanced, therefore, more detailed analysis isneeded based on these type of distribution networks.

V. PROBLEM FORMULATION

The main goal of our analysis is to determine optimumsize and location of dispersed generation so that it can reducethe real power loss in a distribution system. This analysis isimportant for efficient power system planning and operation.Distributed generation not only reduces the power loss ofa system but also improves the voltage profile. However,inappropriate size and allocation of DG can cause low or over-voltage in the distribution system [18]. Therefore, another goalof our analysis is to check whether the voltage profile remainswithin permissible limit. So, voltage constraint becomes,

Vmin ≤ V ≤ Vmax (5)

Distribution step-type voltage regulators are used to maintainline voltages within predetermined limits; as a result enduser can get a constant voltage output [19]. Impact of DGon voltage regulation with the presence of voltage regulatoris discussed in [20]. The test systems, which we used forour analysis, comes with voltage regulators which maintain apermissible voltage limit throughout the analysis. During ouranalysis, we considered Vmin=0.94 pu and Vmax=1.06 pu Insection IX-B, we will show how optimum size and locationof DG impacts on voltage level of the interconnecting bus.

VI. PROPOSED ANALYSIS METHOD

In our analysis, repeated load flow for loss reductionhas been performed in a different approach by integratingopenDSS with Matlab which makes the analysis faster, moreaccurate and efficient. Base on sensitivity, a new methodologyhas been proposed to calculate optimum size and locationof DG. Finally, these two methods have been analysed andcompared.

A. Optimized Algorithm using Sensitivity

For any distribution system, if DG size is varied from PDG1

to PDG2 and their corresponding change in power loss isrespectively PL1 to PL2, then the sensitivity factor becomes,

dPL

dPi=

PL1 − PL2

PDG1 − PDG2(6)

In our analysis, Sensitivity factors are evaluated for eachbus using equation (6) and the bus with maximum sensitivityis identified. Only those buses which have sensitivity factorsclose to the maximum value have been considered in ouranalysis. Thus solution space is reduced to only a few buses.After that, for each of these buses, power loss has beendetermined using large step size of DG variation and thena quadratic curve has been formed using these few samples.The minimum value of the curve represents the optimum sizefor that bus and corresponding generation is the optimum DGsize. The bus which is responsible for minimum loss of thesystem is the appropriate location for DG allocation. The flowchart of our analysis methodology is given in Fig. 2.

VII. TEST SYSTEM AND ANALYTICAL TOOLS

Our analysis method has been implemented on three IEEEdistribution test systems. These are IEEE 34 node test feeder,13 Node test feeder and 123 Node test feeder. The first one,IEEE 34 Node test feeder is an actual feeder located in Arizonaand its nominal voltage is 24.9 kV [21]. The second one,IEEE 13 Node test feeder is small but good for test cases. Thethird one, IEEE 123 Node test feeder operates at a nominalvoltage of 4.16 kV and contains overhead, underground linesegments with various phasing, unbalanced loading with allconfiguration of loads [21]. Single line diagrams of the testfeeders are shown in the appendix. The basic data for thesethree test feeders can be obtained from [22]. For analysis pur-pose, we have developed several Matlab programs to performthe tasks 3 to 12 shown in flow chart. Only for three phaseunbalanced load flow we have used Open Distribution SystemSimulator (OpenDSS). OpenDSS has an in-process COMserver DLL through which the user can design and executecustom solution modes and features externally [23]. We haveinterfaced openDSS COM server with our Matlab programand load flow results have been imported from openDSS. ThenMatlab programs have been used to find out optimum size andlocation of DG.

Start

1. Input basic data of Distribution system

2. Base case load flow

3. Calculate sensitivity of each buswith a small penetration of DG

4. Make a list of most sensitive buses

5. Select a bus from the list and calculatepower loss for large variation of DG size

6. Continue until loss starts toincrease and record each sample

7. Fit those samples into a quadratic curve

8. Find the min point of the curve whichrepresents the min loss of that bus

9. Checkwhether all

sensitive buseshave beenanalysed?

10. Find the bus which has min power loss

11. Find corresponding DG size

12. Print optimum DG size and location

End

yes

no

Fig. 2. Flowchart to identify optimum size and location of DG

VIII. SIMULATION RESULTS AND DISCUSSION

Fig. 3 shows the trend of power loss with the variation ofDG size for each bus of 34 Node test system. This figurehas been obtained using repeated load flow method. In thatanalysis procedure, DG which is responsible for minimumloss at each bus has been identified first and that is optimumDG size for that bus; therefore, the bus which has minimumloss among the whole network is considered as the optimumlocation for that network. Although repeated load flow isinefficient due to its large number of load flow computationsand thus called ‘exhaustive’ approach but it gives us accurateresults. Using this methodology, optimum location of DG for123 Node test system is bus 76 and optimum DG size is1.32 MW. Now, using our methodology, for same test system,we have found the optimum size of DG is 1.32 MW andoptimum location is bus 76 which is similar with the generated

0 200 400 600 800 1000 1200 1400 1600 1800120

140

160

180

200

220

240

260

280

300

DG Size (kW)

Pow

er L

oss

(kW

)

Fig. 3. Power loss profile with the variation of DG size (for 34 Node testsystem)

0 200 400 600 800 1000 1200 1400 1600 180065

70

75

80

85

90

95

Variation of DG Size (kW)

Pow

er L

oss

(kW

)

ActualApproximate

Fig. 4. Comparison of power loss curve using both methodology at busnumber 76 for 123 bus test system

result obtained from repeated load flow. For the other two testsystems, IEEE 13 Node test system and IEEE 34 Node testsystem, we have obtained optimum DG size and location usingexhaustive search and proposed method. Results obtained fromsimulation are shown in table I. For all of the three testsystems, results from our proposed method are quite similarwith repeated load flow approach but our method is quitefaster.

TABLE ICOMPARISON OF THE RESULTS OF TWO METHODS

Test Repeated load flow Proposed Method

System Optimum Optimum Optimum OptimumSize (MW) Location Size (MW) Location

13 Bus 1.28 Bus 675 1.29 Bus 67534 Bus 1.10 Bus 844 1.15 Bus 844123 Bus 1.32 Bus 76 1.32 Bus 76

IX. EVALUATION OF ESTIMATION PERFORMANCES

For 123 Node test system, at bus number 76, power losscurves using both methods have been shown in Fig. 4. Weconsidered the output of repeated load flow as the actual caseand output using our method is the approximate one. From thefigure we can say, our methodology has a very good matchwith the actual case; moreover, it takes less no of solutionspace and time.

A. Active power loss reduction by DG

Active power loss reduction (PLR) by DG may be definedas below:

PLR =PLoss − PDG

Loss

PLoss100% (7)

where,PLoss =Power loss of the system before introducing DGPDGLoss =Power loss of the system after adding DG

For 123 bus test system, before adding any DG, total systempower loss was 95.3 kW. Adding an additional DG of optimumsize on the optimum location, this loss becomes 65.3 kW.Therefore, active power loss is reduced by 31.5%. Therefore,overall system efficiency improves. Using optimum size andlocation, active power loss reduction for the three systems aresummarized in table II.

TABLE IIACTIVE POWER LOSS REDUCTION IN DIFFERENT TEST SYSTEMS

Test Power Loss (kW)

System Before After PLRadding DG adding DG

13 Bus 110.4 85.4 22.6%34 Bus 272.8 125.5 53.9%123 Bus 95.3 65.3 31.5%

B. Impact on Voltage profile

Voltage range for static voltage stability varies from countryto country and utility to utility. However, ANSI standardshould be maintained to supply a steady-state voltage onconsumer side of a distribution network [19]. The acceptablesteady state voltage range of systems between 1kV and 132kV is considered within ±6% of the nominal voltage [24]. For123 bus test system, nominal voltage is 4.16 kV and voltagebase is 4.16 kV, therefore, acceptable voltage should be within0.94 pu to 1.06 pu For the other two test systems, we get thesame voltage range for steady state voltage stability. In ouranalysis, using voltage regulator and allocating optimum sizeof DG in optimum location voltages remain within acceptablelimit which is shown in Fig. [5].

X. CONCLUSION

The impact of proper allocation and sizing of DG isvery significant. Power Loss of distribution system increasesoverall system cost and has a major impact in power systemmanagement. From our analysis, we can come to know that,improper DG size and inappropriate DG allocation may causea greater system loss than the loss without DG. DG shouldbe allocated in those locations where they provide higherreduction of losses. Although DG is usually consumer’sproperty, nevertheless, it is the interest of utilities andengineers to determine the appropriate size and location forsafe, reliable and stable operation of the distribution system.Our analysis method identifies this sizing and location issuesaccurately and precisely in a faster manner. To understand theimpact of proper allocation and size of DG properly, more

0.8

0.85

0.9

0.95

1

1.05

1.1

Voltage variation at the phases of bus 675

Vol

atag

e M

agni

tude

( P

.U.)

Voltage before adding DGVoltage after adding DG

Phase−3Phase−2Phase−1

(a) 13 Node test system

0.8

0.85

0.9

0.95

1

1.05

1.1

Voltage variation at three phases of bus 844V

olta

ge M

agni

tude

( P

.U.)

Voltage before adding DGVoltage after adding DG

Phase−1 Phase−2 Phase−3

(b) 34 Node test system

0.8

0.85

0.9

0.95

1

1.05

1.1

Voltage variation at three phases of bus 76

Vol

age

Mag

nitu

de (

P.U

. )

Voltage before adding DGVolatge after adding DG

Phase−1 Phase−2 Phase−3

(c) 123 Node test system

Fig. 5. Voltage profile of different test systems

detail dynamic analysis is needed which will be our futurework.

APPENDIX

Fig. 6. IEEE 13 Node Test Feeder

Fig. 7. IEEE 34 Node Test Feeder

Fig. 8. IEEE 123 Node Test Feeder

REFERENCES

[1] D. Rizy, F. Li, H. Li, S. Adhikari, and J. Kueck, “Properly understandingthe impacts of distributed resources on distribution systems,” in Powerand Energy Society General Meeting, IEEE, 2010.

[2] O. T. Mithulananthan, Nadarajah and L. V. Phu, “Distributed generatorplacement in power distribution system using genetic algorithm to reducelosses,” Thammasat International Journal of Science and Technology,vol. 9, pp. 55–62.

[3] N. Acharya, P. Mahat, and N. Mithulananthan, “An analytical approachfor DG allocation in primary distribution network,” International Journalof Electrical Power & Energy Systems, vol. 28, no. 10, pp. 669–678,2006.

[4] R. Arritt and R. Dugan, “Distributed generation interconnection trans-former and grounding selection,” in Power and Energy Society GeneralMeeting, IEEE, July 2008.

[5] T. Ackermann, G. Andersson, and L. Sder, “Distributed generation: adefinition,” Electric Power Systems Research, vol. 57, no. 3, pp. 195–204, 2001.

[6] “IEA publication, distributed generation in liberalizedelectricity market, 2002. (page-19).” [Online]. Available:http://www.iea.org/textbase/nppdf/free/2000/distributed2002.pdf

[7] T. McDermott, “Behavioral models of DR technologies for feeder-levelanalysis,” in Power Energy Society General Meeting, IEEE, July 2009.

[8] H. Saadat, Power System Analysis. Tata McGraw-Hill, India, 2002, p.212.

[9] J. O’Donnell, “Voltage management of networks with distributed gen-eration,” Ph.D. dissertation, The University of Edinburg, 2007.

[10] H. Willis, “Analytical methods and rules of thumb for modeling DG-distribution interaction,” vol. 3, pp. 1643–1644, 2000.

[11] C. Wang and M. Nehrir, “Analytical approaches for optimal placement

of distributed generation sources in power systems,” IEEE Transactionson Power Systems, vol. 19, no. 4, pp. 2068–2076, nov. 2004.

[12] D. Q. Hung, N. Mithulananthan, and R. Bansal, “Analytical expressionsfor DG allocation in primary distribution networks,” Energy Conversion,IEEE Transactions on, vol. 25, no. 3, pp. 814 –820, sept 2010.

[13] M. Srinivas, “Distribution load flows: a brief review,” in Power Engi-neering Society Winter Meeting, IEEE, 2000.

[14] G. Celli and F. Pilo, “Optimal distributed generation allocation in MVdistribution networks,” in IEEE Power Engineering Society InternationalConference on Power Industry Computer Applications, 2001, pp. 81–86.

[15] M. Ameli, V. Shokri, and S. Shokri, “Using fuzzy logic amp; full searchfor distributed generation allocation to reduce losses and improve voltageprofile,” in International Conference on Computer Information Systemsand Industrial Management Applications (CISIM), oct 2010, pp. 626–630.

[16] L. de Queiroz and C. Lyra, “A genetic approach for loss reduction inpower distribution systems under variable demands,” in IEEE Congresson Evolutionary Computation, 2006, pp. 2691–2698.

[17] J. Savier and D. Das, “Impact of network reconfiguration on lossallocation of radial distribution systems,” IEEE Transactions on PowerDelivery, vol. 22, no. 4, pp. 2473–2480, oct 2007.

[18] C. Borges and D. Falcao, “Impact of distributed generation allocationand sizing on reliability, losses and voltage profile,” in Power TechConference Proceedings, 2003 IEEE Bologna, vol. 2, june 2003, p. 5.

[19] W. Kersting, “Distribution feeder voltage regulation control,” IEEETransactions on Industry Applications, vol. 46, no. 2, pp. 620–626, 2010.

[20] L. Kojovic, “Impact of DG on voltage regulation,” in Power EngineeringSociety Summer Meeting, 2002 IEEE, vol. 1, july 2002, pp. 97–102.

[21] “Distribution system analysis subcommittee report.” [Online]. Available:http://ewh.ieee.org/soc/pes/dsacom/testfeeders/testfeeders.pdf

[22] “Distribution system analysis subcommittee ra-dial test feeders [online].” [Online]. Available:http://ewh.ieee.org/soc/pes/dsacom/testfeeders/index.html

[23] “OpenDSS manual and reference guide.” [Online]. Available:http://sourceforge.net/projects/electricdss

[24] C. Masters, “Voltage rise: the big issue when connecting embeddedgeneration to long 11 kv overhead lines,” Power Engineering Journal,vol. 16, no. 1, pp. 5–12, feb 2002.

Adnan Anwar was born in Bangladesh in 1987.He received B.Sc in Electrical and Electronic En-gineering from Islamic University of Technology(IUT), Bangladesh in 2008. In 2009, he joined atthe University of Asia Pacific (UAP) as a lecturer.He is currently pursuing his Masters by Researchat the University of New South Wales, AustralianDefence Forces Academy, Canberra, Australia. Hisresearch interests include computer aided analysis ofpower distribution network, optimization of distri-bution system, power system dynamics and control,

integration of renewable energy sources in distribution system and smart grids.

Hemanshu R. Pota received the B.E. degree fromSVRCET, Surat, India, in 1979, the M.E. degreefrom the IISc, Bangalore, India, in 1981, and thePh.D. degree from the University of Newcastle,NSW, Australia, in 1985, all in electrical engineer-ing. He is currently an Associate Professor at theUniversity of New South Wales, Australian DefenceForce Academy, Canberra, Australia. He has heldvisiting appointments at the University of Delaware;Iowa State University; Kansas State University; OldDominion University; the University of California,

San Diego; and the Centre for AI and Robotics, Bangalore. He has acontinuing interest in the area of power system dynamics and control, flexiblestructures, and UAVs.