You are on page 1of 8

Electrical Power Quality and Utilisation, Journal Vol. XI, No.

1, 2005

Vu Van THONG1, Johan DRIESEN1, Ronnie BELMANS1, Bart De MOOR2
(1)K.U. Leuven, ESAT - ELECTA; (2)K.U. Leuven, ESAT - SCD - SISTA

Summary: In this article the authors describe how genetic optimization algorithms can be used Keywords:
to find the optimal size and location of distributed generation units in a residential distribution Distributed generation
grid. Power losses are minimized while the voltage profile is kept at an acceptable level. The Optimization methods
method is applied on a system based on an existing grid topology with production and residential Microgrid
load data based on measurements. Different scenarios are chosen to run the algorithm. The
obtained optimal location and size prove to depend strongly on the given conditions.

1. INTRODUCTION wer injection will affect the voltage profile throughout the
grid. This way the location where active power is injected
Decentralized generation will become more and more im- will be of importance.
portant in the future electricity distribution system. This ten- All these different aspects of distributed generation sour-
dency is increased by the commercial availability of small- ces in a low- or medium-voltage grid will influence the impor-
scale production units (e.g. fuel cells, micro-CHPs, photo- tant question of where to place the production units. Diffe-
voltaic panels) and the liberalization of the energy market, rent objectives can be put forward, such as highest effciency
putting more pressure on the system. Also the support for (i.e. lowest power losses), minimal cost (installation and ope-
sustainable development using renewable energy sources ration), highest reliability, etc. This optimization problem can
plays a key role. be solved in different ways like exhaustive searches [14],
As such, distributed generation is defined as local genera- Lagrangian based approaches [12] or tabu searches [11].
tion of heat and electricity in the distribution grid. A group of In this article, the search for the optimal placement and
DG units can form a virtual power plant, being centrally con- power level of the different types of generation units (in such
trolled and behaving as a single power plant towards the grid. a way that the power losses are minimized) is a non-convex
The extreme case is an energy island, in which production and optimization problem that requires exhaustive search. The
consumption of energy are locally matched. Energy is also use of Genetic Algorithms is proposed, in order to implement
kept in balance by local storage and by an optional connec- the search of the optimal solution. Advantages of genetic
tion to the main grid across which power is transferred. Howe- algorithms are the ability to avoid being trapped in local opti-
ver, energy islands should have the ability to move from an ma, and also the expected number of function evaluations
uncontrolled power mode (when connected to the main grid) before reaching the optimum is significantly reduced compa-
to a load tracking mode (while in island mode) [8, 9]. red with exhaustive search methods. This paper is structured
In this new environment characteristics will be fundamen- as follows. Section 2 describes the general framework of the
tally different compared to a situation in which generation is problem (the grid architecture, the load and production pro-
centralized. There is almost no inertia in the energy sources files). Section 3 describes the methodology based on Gene-
which is needed for stability reasons. Although storage of tic Algorithms. The results are presented on Section 4.
energy can have a stabilizing role. The load pattern is more
varying in time since the averaging of consumption vani-
shes due to the small amount of users on such a small-scale 2. GENERAL FRAMEWORK
low voltage grid. Compared to conventional power plants
distributed generation units such as PV cells (depending on In this section the general configuration of the problem is
solar illumination) or CHPs (depending on heat demand) are addressed, describing the grid topology, power production
undispatchable. and load profiles for the different scenarios.
Another fundamental problem is the linking between acti-
2.1. Grid Topology and Power Production
ve power injection and the voltage profile in the grid. In high-
voltage (thus mainly inductive) grids an active power injec- For this analysis the topology of an existing grid at me-
tion imbalance will be the cause of a frequency deviation dium-voltage is used. It consists of 3 lines and 20 nodes,
which is equal in the entire grid. The sum of all produced with one aggregated load at each node, as shown in Figure 1.
power must be adjusted to keep the grid frequency at nomi- The distributed energy sources used are PV panels and CHP
nal level. In low-voltage (more resistive) grids the active po- units. Real measurements are used to provide data for the

Edwin HAESEN et al.: Optimal Placement and Sizing of Distributed Generator Units Using Genetic ... '%

 V  Q=  Qnom (2b) Medium (WM) and Winter-High (WH). This scenario contains the α most important peaks of each season. Medium Load Scenario (M): From the sets of available voltage-dependent load. defined by its rated active and reac- daily load profiles. Winter-Low (WL). voltages are denoted as Vnode. each one of these sets conta- ining 4 load profiles is repeated 5 times in each scenario. α dium (SM). Problem Definition grid. while node and line each season is selected. As the grid contains 20 load profiles. This scena. in order to current generated or consumed at a node) as minimize the power loss along the grid lines over a period of 24 hours. Winter. 1.  V  3. This scenario only contains 1 important the line/node number. Another 4 daily profiles are ran. 3. On each row two non-zero elements are present building for a period of one year on a 15 minute basis. thus leading to Summer. Fig.g.1. No 1. High Load Scenario (H): The load profiles with highest P=  Pnom (2a) daily averages are selected on each season. For each node a load is considered that is modeled as a 2. Fig.  Vnom  rio presents a heavy requirement to the generators. those with the highest daily peaks on tive power consumption Pnom. allows the conversion of line currents (cur- The objective of this analysis is to find the optimal place. the profile with the highest peak during Line impedances are denoted as Zi. 2. which implies that m = n + 1. Power Line Equations A network with m nodes and n power lines is considered. load profiles. Production Profiles for CHP (left) and PV (right) generation units for a period of one day '& Power Quality and Utilization. Summer-Me. Summer-High (SH). 3. Each one of these scenarios are computed for Summer and Winter. Qnom: each season are selected. 2005 .i with i denoting domly selected. The different load  Vnom  profiles for each scenario are depicted in Figure 3. In order to assess the effect of Winter and Summer Inode = T TIline (1) variations on different load situations. node where the line ’ends’ (1). PROBLEM FORMULATION AND METHODOLOGY In this section. the period of 24 hours over which the minimization takes place is defined in the with Inode Î ¡ (n+1)×1 a vector denoting the current leaving following scenarios: each node and Ilines Î ¡ n×1 the current flowing through 1. load peak over each season. rents flowing through a power line) to node currents (total ment and size of CHP and PV generation units. This e. The convention is used that –1 is assigned to the node that lies the closest to the power 2. the formalization of the problem and the coding of the Genetic Algorithm is described. where each row represents a po- based on heat demand. wer line and each column represents a node of the power sumption are based on measurements taken in a residential network.i and Vline. XI.Low (SL). The profiles of active power con.2. Grid Topology with 20 nodes The network is assumed to be radial. Journal • Vol. Low Load Scenario (L): From the sets of available daily each line. Figure indicating the node where the power line ’starts’ (–1) and the 2 shows the production profiles for each generator type. The topology of the network is defined by a topo- illumination on PV panels and the CHP production which is logy matrix T Î Z n×(n+1).

Qi.. (a) (b) (c) (d) (e) (f) Fig.i. '' . Pnom. The nominal active and initialization phase. Both generators An iterative solution strategy is chosen. Sets of 4 Load Profiles for the different scenarios where a = 2 in case of a resistive load [7]. given a network topo- Edwin HAESEN et al. node attached to the power grid) Vgrid. followed by a fixed number of forward reactive power of a load attached to node i are denoted as and backward iterations.: Optimal Placement and Sizing of Distributed Generator Units Using Genetic ..e. consisting of an and loads are modeled using (2).i. Qnom.. while the actual active and reactive power are Given the voltage of the power grid (i. 3. voltage of the denoted as Pi.

Each candidate solution is known as a 1. string T. G2) represents a PV (resp. The methodology can be summarized as follows: vance in the search. As an example. — Sort the individuals S according to their fitness the fitness of each chromosome is defined in such a way that — Perform genetic operators to produce a new generation the chromosome with the highest fitness represents the opti.  Power Quality and Utilization. It is assumed. ward and backward iterations are performed several times in As each string T represents the generators (and size) to order to obtain a converged solution. the total energy lost f (S) = Ploss + C1d1 + C2d2 (4) over an entire day can be calculated for the given configura- tion.0035 for each bit. As any string S describes a valid placement The resulting total power loss Ploss over the lines is calcu. therefore lated as: the string S is the chromosome used within the GA. which based on the hypothesis of natural selection [6]. strings. discretized between 1/16 and 100%. and this generator sho- at the given node are updated in order to satisfy the uld be a PV working at full capacity. Implementation Summary relative fitness and other genetic operators that serve to ad. .I line. The  i  fitness of each individual is given by the objective function. This iteration is used to update the G (thus T contains 10 bits). n and nominal loads thin the power systems framework [13. let T = G1G2. This iteration is used to update all G = [11111] represents a generator working at full capacity. 4. The implementation of the GA is done with generations consisting of 40 individuals (each one a different string S). the representation of the general rents are small compared to the line impedances (and there. voltage Vline lies outside the limits. The GA is are selected with a probability that depends on their fitness. Repeat for 300 generations: (crossover. Vline – Vmax). string T is defined consisting of the concatenation of 2 strings — Backward iteration. Journal • Vol. "i nerator working at minimum capacity. node voltages in order to take into account the voltage In order to represent the type of each generation. gins with a very large set of initial candidate solutions. . an evolutionary population-based search process that be. In this way.3. The individual S with the best fitness after the 300 genera- generation to the next [5].1 := Vgrid power level of the generator. where G1 power generation and consumption based on the node (resp. d1 = max(0. the string G = [1000] represents a ge- – Set Iline. etc. For example. . mutation. plus another penalization if the num- By solving the power line equations for all time instants of ber of generators exceeds 10. location of the generators over the grid is straightforward. that a maximum of one generator of each The initialization step is performed once. GAs have been used recently wi. No 1. so are the line currents.i. N – Nmax) takes a penalty C2 if the number of generators N on solution 3. G = [00000] represents a generator which is not operating (or nonexisting). . be placed at a given node. Under certain conditions. It is this total power loss that is being minimized in the where Ploss is computed from equation (3). a new drop over the power lines. this representation. 1]. at a given node the string T = ted. The initialization consists of several steps 5 binary bits. The fitness value f for any a set of production and generation profiles for all involved individual S is: devices connected to the local grid. it — Compute the fitness of the individuals from (4) has been proven that the average fitness improves from one 4. voltages calculated in the previous forward iteration step the type of each generator is given by the position on the and using equations (2). Each generator is represented by a string G of — Initialization. the line currents of the power lines starting rator should be placed on that node. In any generation. 2005 . chromosome. The representa- Pnom. based on power needs at their ending nodes. Genetic Algorithm Implementation S is larger than the maximum allowed number of generators A Genetic Algorithm (GA) is a search algorithm that is Nmax. type can be placed on any given node.). XI. of individuals S mal point in the search space. takes a penalty C1 if the lines power generators. the node and line voltages and currents are tion and implementation of the GA for the problem is propo- computed by executing the following steps: sed as follows. The remaining 4 bits represent the – Set Vnode. Compute the fitness of the individuals from (4) the previous set through the so-called genetic operators 3. The- se solutions are subjected to selection pressure based on 3. by modifying the placement of the local – Vline) + max(0. and the set of all chromosomes is created from 2. CHP) generator. and d2 = max(0. * i (3) The maximum number of generations is set to be 300. for all nodes [1111100000] represents the situation where only one gene- i = n. tournament. .logy T. Mutation takes place with probability 0. The first bit represents the state of the genera- tor (1 for on. and it also considers a penalization if the voltage goes outsi- de the allowed interval. and size configuration of generators over the grid. Generate a set of 40 individuals S. while the for. This sequence S contains 20 (nodes) × 10 (bits per ge to the correct solution.i. i = 1. Crossover is performed between individuals. line impedances Zi. . In case the line cur. Vmin following sections. Therefore. tions is the final solution.   Ploss = Re  ∑ Vline. A fore the line losses are sufficiently small compared to the string S is defined consisting on the concatenations of 20 T nominal voltage) the above method can be shown to conver. Subsequently. the string — Forward iteration.i .i := 0. . node) = 200 bits. 1.2. 0 for off). Qnom. . As power injections are upda.

Placement and sizing depend strongly on the It is also clear that the selected load scenarios translate into given base case scenario. The solution is stable as it remains unchanged for a large number of generations. 5.. and using a In this article this method is extended to placement and si- different one will increase the power loss. WM) to the solution of using 10 CHP units only (WH). The final column in Table 3 shows the total loss when solutions are obtained for each season. it is possible to compute the power loss when the 5. Evolution of the total power loss along the lines over 300 more CHPs are placed compared to the other scenarios wi- generations for the SL scenario thin the same season. zing of generators based on known generation and load pro- ching the configurations for within Summer scenarios has a files. Sometimes it is assumed that a different power losses that are related to the magnitude of CHP and a PV panel are best combined on one location be- the load requirement on each case. In particular. rators placed in the Summer scenarios. In Summer. logy. Table 2 shows a summa- rized description of the different solutions obtained. It is clear that the opti. type and size of generators in the grid.. Table 3 shows all the losses for the possible combi. swit. mal solutions are dependent of each scenario. Each solution represents the location. However. The use of genetic algorithms in distribution networks is nations of configurations/scenarios. the solution moves from 7 PV generators and 3 CHP units (SL.: Optimal Placement and Sizing of Distributed Generator Units Using Genetic . In the tested grid An interesting question is to check what would be the this occurs 5 times over all the possible solutions. power loss along the lines if a generator configuration diffe- rent than the optimal one is used within each scenario. 4. Simulations are run on a typical distribution grid topo- less dramatic effect than switching within Winter configura. cause of complementary power outputs. Detailed results are presented in Table 1 and Figures 4–5 for all scenarios. 6. using load profiles with seasonal variation. As could be expected. Results for the different Summer scenarios Fig. It is also clear that for the high demand scenarios. in the Summer scenarios more PV cells are chosen by the algorithm as compared to Winter when heat demand (and thus power production by CHP) is higher. Different tions. with more PV gene- there are no DG units. In Winter the situation also moves from 7 CHP and 2 PV Rules of thumb to place generators are not found in the (WL. known for feeder switching [2] or capacitor placement [10]. an optimal final chromosome S is fo- und. CONCLUSION optimal configuration from one scenario is used in a different scenario. given example. Fig. SM) to 6 PV and 4 CHP (SH).Fig. and more CHP gene- Edwin HAESEN et al.  . Results for the different Winter scenarios 4. Figure 6 shows the evolution of the total power loss for the best chromosome of each generation for the SL scenario. RESULTS For each scenario.

25 — — 6.50 56.75 — 31.50 18.25 — — — 25.25 — 14 — — — — — 62.00 — — — — 18.75 18.75 — 25.50 — 18.25 31.25 — — — — — — — — 5 — 6.25 — 4 — 25.25 25.25 — 31.25 12.75 — 6.25 — — 13 12.50 — — — 2 — — — 18.00 — — — — 31.25 — 12. No 1.00 — 25.25 — 6.00 — 25.75 — 12 — 37.50 — 12.75 — — — — 6.75 — — 19 — 31.50 — — — Table 2.25 — — — 31.25 — 12.00 — 43. Type and Size (in percentage) of the Generators for the Different Scenarios Node SL SM SH WL WM WH CHP PV CHP PV CHP PV CHP PV CHP PV CHP PV 1 — — 6.50 — — 6.25 — — — — — 18.25 — — 18. XI.75 — 3 6.25 — — — 6.25 — 12.75 — — 6. Journal • Vol.25 — 16 — 25.25 — 9 — — — — — — 18.00 — 20 — — — 18. Results from the Genetic Algorithm for all scenarios Placement.25 — — — — 6.50 — — — — — — 15 — — — — — — 18.25 — — — 17 — — — — 6.75 — 18 — — — — — — 6.25 — — 12. 2005 .25 — — — — 6.00 — — — — 6.00 6.75 — — — 11 — — — — — — — — — — 18.75 — — — 6. Summary of the Results and Total Power Loss (kWh) over a period of 24 hours Table 3.25 — 8 — 25.75 18. Cross-Comparison of optimal configurations used in different scenarios  Power Quality and Utilization.Table 1.25 — — — — — — 31.50 — 6 — — — — — — — — — — — — 7 6.25 — — — 10 — — — — — 25.00 — 31.

T r i s t i u I . F a l c a o . AWI.02. H a y a s h i Y. The minimization of the power loss is only a part of mance.03. Leuven. replacement and Sizing of power certificates which can be traded. Edwin Haesen was born in 1981 in Tongeren. neering Department of the K. several PhD/Postdocs & fellow grants). . . 9. in market research and copper sales management. New York. L a s s e t e r R . 1507. chine Learning. G o l d b e r g D . a n d P a h w a A . LMS. F u r o n g L . . Octo- ber 2004. IEEE Power Engineering So- installation costs there are also production costs.A. IPCOS. the EU(FP5. CIGRE Symposium. flywheels. G. Greece. IEEE Transactions on Power Systems.Belmans and B. D a b e e d i n C . 4. . April 2005. To enlarge the number of feasible of Capacitors for Radial Distributions Systems using Genetic Algorithms. S u n d h a r a r a j a n S . 1989.Soft4s). .U.Sc. de Chile. T . May 1993. . optimization algorithm which uses the solution of the gene- tic optimization as a start and searches for small deviations which lead to a more optimal situation. projects G. Athens. IEEE Transactions on Power Systems. Sevilla. 14th Power Systems Computation Con- ference (PSCC). the Flemish Marcelo Espinoza Government (FWO: PhD/Postdocs grants. in 1973. De Moor are full professors using econometrics and least-squares support vector machines.Leuven. Load Representation for Dynamic Performance Analy- sis. Athens. Data4s. A . R o s e h a r t W . IEEE Transactions on Power Delivery.D. 19. L a d j e v a r d i M . identification. F a c h o u r i d i s S . 7. IEEE Task Force on Load Representation for Dynamic Perfor- rations. 20. gineering Dept. G. 1994. L . The total energy loss in the optimal place. 2002. 4. CIGRE Symposium. Eureka Intelligence from the K. using a different placement for any given scenario (different M a n o u s a r i d i s K . De Moor and Prof. 1499– production sizes the discretization step can be decreased. IEEE Transactions on Power Delivery. the economic cost. : Adaptation in Natural and Artificial Systems. . N a r a K . 1. : energy resources are used.U. C h e e b o files being used. B u l a c C . : Artificial Intelli. D j a l m a M . 2002 he obtained the degree of Master in Artificial PODO-II (CP/01/40). Leuven. Feb. ! . R. Leuven under the 2. 6. Suykens. 9. I k e d a K . D a g o u m a s A . This can easily be widened to Application of Tabu Search to Optimal Placement of Distributed Generators. .ERNSI. This will call up other issues.: Optimal Placement and Sizing of Distributed Generator Units Using Genetic . and it is ciety Winter Meeting. This leads to longer chromosomes and eventually a more 14. H .0211. : Optimal Selection nominal power outputs.0407. A h i z a w a T . Is the energy price fixed lution. E r e m i a M . He is currently pursuing a PhD at the SI- gence and Advanced Mathematical Tools for Power Quality Ap. Besides 8.Leuven. April 2005. P a p a g i a n n i s G . L e t t a s N . University of Michigan Press. His rese- Operation of Distribution Systems in the Presence of Dispersed Generation.U.U. Sizing of generators is done by discretizing the possible 13. Another possibility is an algorithm in Optimal Location of Dispersed Generators for Reliability Impro- vement of Distribution Networks. . the Belgian Federal Go. G. with interest in load modelling Mastercard). 6. .. He worked for 4 years in Codelco-Chile. ANMMM. 2004. Bert Pluymers was born in Geel on August 24th. This stresses the impor. ACKNOWLEDGMENTS at the KU Leuven as a research assistent at the division ESAT / ELECTA. 1. Athens. . a n d A g g a r w a l R . He is currently pursuing a Ph.K. .etc.Quprodis. 2002. 918–923. GOA-Am. 2.02. Z a f i r a k i s A . 2. CIGRE Symposium. respectively. arch interests include Model based Predictive Control. : MicroGrid: A Conceptual So- not clear how can they be included. P a i g i P. . biorics. He rece- ived his M. April 2005. In the given simulations two common types of distributed 11. N o w i c k i E . tomation) research division of the Electrical Engi- 2. : Optimal Placement.. Ann Arbor. : Genetic Algorithms in Optimization and Ma- power is taken from the main grid. STA (signals. IWT: M. G. In vernment (Belgian Federal Science Policy Office: IUAP V-22. 2001 interesting issue is the use of storage like e.0452. B. Greece. Elia. In 1998 he G. K. a n d M o r c o s M . It is also shown that 3. 8.0491.0141. production of energy using renewables is coupled to green 10.: Genetic Algorithms for Optimal ment is in each scenario considerably lower (up to a factor 4 Reactive Power Compensation on the National Grid System.D with the Electrical En- Agreements (ISMC/IPCOS. P i l g r i m J . June 2002.Sc in Applied Economics from the Universidad PhD grants. An extension to this work is to include economic conside.04. Belgium 1. : Optimal Placement of batteries. TML. B o r g e s C . Another 2001. suggesting that the placement is dependent on the load pro. 1979 and received his Master Degree in Electrotechnical andMechani- REFERENCES cal Engineering (datamining and automation) at the Katholieke Universiteit Leuven. G. M a s o u m M .Eureka 2419-FLiTE) and Contracts Research/ currently pursuing his Ph.03. 20–25. . Belgium. 493–500. systems theory and au- plications: A Survey. at the K. or 5) as compared to the case in which there is no DG and all 5. : complex algorithm. ICCoS. IEEE Power Engineering Society Winter Meeting. Chile.g. IEEE 35th Annual Power Electronics Specialists Confe- or does a local trading market exist? Also in some countries rence (PESC).0240. H .04.0499. L a s s e t e r R . Belgium. 3.0120. T o m a L . He is 2063-Impact. (ESAT). in 2002. Addison Wesley. 17. superconductive coils. Robust Control and Convex Optimization. : Maximizing distributed generation ca- than the optimal solution) increases the power losses. neration. 1975. degree in electrical engineering at the KU Leuven in 2004.03. 6. . tance of DG and its optimal placement and sizing. 2005. J. which the genetic optimization is followed by continuous Greece. 668–673. 12. more types like wind power and small hydro power.05. Edwin HAESEN et al. Belgium. H o l l a n d J . . : Planning and supervision of Prof. : MicroGrids. G. was born in Santiago. J a f a r i a n A .Leuven (GOA-Mefisto 666. S . F u c h s E . D o k o p o u l o s P.0197. . IEEE Transactions on Power Systems. . Distributed Generation. thus pacity of small hydro plants in a distribution network. 27–31. All these issues have Capacitor Banks in Distorted Distribution Networks by Genetic to be well considered in order to keep assumptions realistic. His research interests are This work was supported by grants and projects for the in the domain of power systems and distributed ge- Research Council K. Algorithms. GBOU(McKnow. . . A n i s I b r a h i m W . . A .99.U.rators placed in the Winter scenarios. received the degrees of Civil Industrial Engineer and G.

con- trol theory. Bart De Moor gy. power qu- tion. Leuven. microsys. Before coming obtained his Master Degree in electrical engineering to Thailand for ME. Guillemin-Cauer best paper Award degree in 1996 as Electrotechnical Engineer from of the IEEE Transaction on Circuits and Systems (1990). dynamic study. teaching electric power and versiteit Leuven. He received the M. and the Habilitierung in 1993.U. 1996). London. No 1.E. Johan Driesen His work has won him several scientific awards (Leybold-Heraeus Pri- was born in 1973 in Belgium. Power Systems at Hanoi University of Technology. He obtained received the M. Leslie Fox Prize (1989).S. " Power Quality and Utilization.D. the K. Currently he is a post- doctoral research fellow of the Belgian ¨Fonds voor Wetenschappelijk Onderzoek . Department ering (http://www. Journal • Vol. Currently he conducts research on distributed generation. Since 2004 he is mal-electromagnetic problems and related applica. the Belgian Royal Academy of Sciences (1992). quantum information theory. SISTA research group of the Department of Elec. Currently.D. In 2000-2001 he was a visiting researcher in the Imperial College of Science. He received the Ph. In 2002 he was working at the University of California.U. inclu- ding renewable energy systems. Belgium. and the M. he had worked as an electrical in 1983 and the Ph.U. His main research interests are in energy systems. ze (1986). Belgium. USA. in the both from the RWTH.Leuven.kuleuven.D. several scientific journals. Currently.Leuven. 2005 . Leuven.U. data-mining.Leuven in Award (1994). both from the the Katholieke Universiteit Leuven.D. voltage linear algebra and optimization.D. Belgium. ality and distributed generation.U. bi-annual Siemens degree in Electrical Engineering at K. degree in 1998. he is a full professor trical Engineering (ESAT) from the Katholieke Uni. XI. best paper award of Automatica (IFAC. Thailand in 2001 respectively. Since June Vu Van Thong 2002 he is chairman of the board of directors of ELIA. for instance in drives and power quality. with the in Belgium. the Special Doctorate in 1989 rently he is working towards a Ph.Leuven. power electronics and its applications. Signal Processing Society Best Paper Award (1999). Belgium. Berkeley. degree in Laureate of the K. no-economic aspects of power systems. He is also guest professor at Imperial College of Science. both from engineer at Electricity of Vietnam for nearly 3 years. UK. in professor at the Department of Electrical Engine- the Electrical Energy research group. Vietnam in 1997 and Asian Institute of Technolo. Germany. Technology and Medicine. Aachen. IEEE 2000 on the finite element solution of coupled ther. he is a full Since 2001 he has been working towards a Ph. Belgium.S. de Belgian received the B. system theory and stability and optimal power flow. His research interests include tech- the fields of linear and non-linear system identifica.Vlaanderen¨and teaches power electronics and drives at the of the of Electrical Engineering of the K.U.Sc. a fellow of the IEEE (www. tems and power quality issues. identification. degrees in Electrical transmission grid operator. K. degree in Nuclear Physics in 2000 from 1979 and the Ph.esat. Medicine and Technology. His special K. Cur. Ivan Goethals Ronnie Belmans was born in Wilrijk. He is an associate editor of tions in electrical machines and drives.U. London-UK. information retrieval and bio-informatics. His research interests are in numerical fields of interest include distributed generation. degree in electrical engineering in the M. in 1978.