1032 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 21, NO. 4, JULY 2006 Model-Based Generation of Low Distortion Currents in Grid-Coupled PWM-Inverters Using an LCL Output Filter Bruno Bolsens, Student Member, IEEE, Karel De Brabandere, Student Member, IEEE, Jeroen Van den Keybus, Student Member, IEEE, Johan Driesen, Member, IEEE, and Ronnie Belmans, Senior Member, IEEE Abstract—In this paper, a single phase inductance–capacitance–inductance (LCL) output stage for grid coupled inverters is designed and built. An accurate model and observer of the output filter and the distorted grid voltage are implemented. The paper deals with the construction of a 14-state model, and the feedback control loop to obtain adequate closed loop response. Simulations indicate a good performance of the controller, with a total harmonic current distortion (THD) below 1%. Experimental results confirm simulations, and illustrate the correct operation of the Kalman observer to estimate the distorted grid voltage (THD 3%). The observer only uses the inverter current measurement as input. The output filter effectively reduces the pulsewidth modulation harmonics in the grid current. Index Terms—Active front end, active power filter, harmonics, inverter current control, inductance–capacitance–inductance (LCL) output stage design, low grid current distortion, power quality, voltage source inverter (VSI). I. INTRODUCTION N THE literature on shunt active power filters, many topologies are found incorporating only an inductor as an output filter between inverter and utility grid [1]–[4]. However, it has been shown that such an output filter may not suffice to meet standards on harmonics emission, such as IEEE 519 or IEC 61000-3-2 [5], mainly because of the high current ripple due to the switching mode inverter. To further reduce the effect of these pulsewidth modulation (PWM) harmonics, more complex output filters can be used, such as an inductance–capacitance (LC) or an inductance–capacitance–inductance (LCL) topology [6], [7]. In this paper, the hardware design and the construction of a state space control loop of an LCL output topology, such as in Fig. 1, is treated. A practical setup is made, using a generic single phase PWM inverter and a digital signal processor (DSP) prototyping platform. Comparisons are made between the LCL topology and others, such as L or LC topologies. The final evaluation criterion is the total harmonic distortion (THD) of the current drawn from the grid. I Manuscript received October 14, 2004; revised June 21, 2005. This work was supported by the Belgian “Fonds voor Wetenschappelijk Onderzoek Vlaanderen” and the Belgian IWT-GBOU Research Project on Embedded Generation. Recommended by Associate Editor H. du T. Mouton. The authors are with the Department of Electrical Engineering, Katholieke Universiteit Leuven, Leuven B-3001, Belgium (e-mail: bruno.bolsens@esat. kuleuven.ac.be). Digital Object Identifier 10.1109/TPEL.2006.876840 Fig. 1. Topology of the LCL output stage. II. EXPERIMENTAL DESIGN A. Hardware Design The hardware designer has to determine optimal values for the parameters L , C, and L , for a given PWM inverter with following constraints: ; • rated dc link voltage ; • rated IGBT current . • inverter switching frequency In grid coupled applications, the bandwidth of the reference signal can vary significantly, depending on the application. For instance, if the purpose is to build an active filter, the bandwidth requirements are higher than for an active front-end, where only the fundamental frequency is concerned. This work uses a superposition concept, where the influence of the grid and the inverter voltage is analysed separately, and subsequently summed to obtain a complete model. The influence of the inverter voltage is treated in following paragraphs, and allows to derive output stage design criteria. The influence of the grid voltage is covered by the grid oscillator model, that is used while constructing the control loop, in Section II-C. To study the effect of the choice of L L , and C on electric quantities, such as the grid current , the inverter current , the capacitor current , and the capacitor voltage , the following transfer functions can be calculated, from the inverter voltage to these respective states (note that the parallel sign // is used to describe the equivalent impedance of a parallel connection of two impedances) as (1)–(4), shown at the bottom of the 0 a single L output stage is modelled. As next page. When can be seen, the impedances L and L are added. Equations (1)–(4) also include the transfer functions when 0 . These solutions can be easily there is no damping split into poles and zeroes, while the more general third-order solution cannot. 0885-8993/$20.00 © 2006 IEEE BOLSENS et al.: MODEL-BASED GENERATION OF LOW DISTORTION CURRENTS Fig. 2. Bode diagram of the different transfer functions. Furthermore, these resistances are parasitic elements of the reactive components, and therefore small. This is beneficiary for the overall efficiency of the LCL output stage. Interpreting these transfer functions, it becomes clear that in the transfer to the grid current , the output stage behaves like a large inductance L L for low frequencies. The Bode diagrams of the current responses are plotted in Fig. 2. On the other hand, the LCL output stage acts as a third-order filter for high frequencies. The resonating equivalent elements are the parallel connection L L and the capacitor C. Their 1033 resonance frequency should therefore be chosen in such a way harmonics are adequately supthat switching frequency pressed, while still providing enough bandwidth for the reference signal [8]. Beyond the resonance frequency, the response decays steeply at a rate of 60 dB/decade. For the example at hand, this frequency was chosen to be 700 Hz. This choice establishes a first relation between the unknowns L L , and C. The next design step is to determine the total output inducL . To achieve this, a form of bandwidth criterion is tance L used. A high bandwidth requires a high voltage range of the inverter output. If the peak grid voltage is , the worst case of available voltage range occurs when the grid voltage is maximal (or – is availminimal). In these cases, only the difference able for control. This value immediately determines the worst rate-of-change (the steepness) of the reference current (5) If the reference current contains steeper parts, it may depend on their location in the phase of the grid whether the available voltage will be sufficient or not. This steepness condition can also be expressed in terms of harmonics (nth harmonic): L (6) (1) (2) (3) (4) 1034 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 21, NO. 4, JULY 2006 where is the grid pulsation and the peak value of the harmonic current. The phase of the harmonic versus the grid voltage fundamental is not considered here, neither are simultaneously occurring harmonics—the worst case is taken. Clearly for an increasing harmonic number, a given inverter can produce proportionally less current. If the reference current contains only a 50-Hz component, the depends on the phase angle of this required inverter voltage current. If the current is purely active, the voltage of the inverter must be equal to the grid voltage, but shifted in phase. If reactive currents are necessary, the inverter voltage is in phase with the grid voltage, and with • a larger amplitude than the grid for capacitive current; • a smaller amplitude than the grid for inductive current. This corresponds to the situation of active/reactive power transport on an overhead line, where the impedance is mainly reactive. Hence, if no capacitive reactive compensation is can be taken needed, the active front end’s dc link voltage smaller. With the bandwidth criterion in mind, a proper value for L L can be chosen respecting the rated values and of the inverter. This establishes a second relation between the design parameters of the LCL output, in the example, L L 10 mH. The only degree of freedom left is to distribute the 10 mH over L and L . This choice does not affect the transfer function from to grid current (see Fig. 2). All other inverter voltage transfer functions, however, are involved. In Fig. 2, and three cases are shown. 1 mH, L 9 mH. C is 57.4 F (dashed). • L 5 mH, L 5 mH. C is 20.7 F (dotted). • L 9 mH, L 1 mH. C is 57.4 F (full line). • L The value of C is adjusted to maintain the resonance frequency at 700 Hz. As seen, to obtain a minimal current in the inverter and the capacitor, the value of L must be large compared to L . Apparently, the optimum lies at an infinitely small value for L . However, the capacitor value C increases. Large capacitors in the output stage are a drawback: when the inverter is not operating, the capacitors cause a (large) reactive grid current, including harmonics due to the voltage distortion. When the inverter operates, it has to supply the same current by itself to the output stage to avoid unwanted grid currents. Therefore, large capacitors put a large current burden on the inverter, leaving less useful current for the application. The inverter rated current is to be compared to the reactive current of the output stage at no operation. This criterion defines L L . L , the inverter current Furthermore, in the case where L L present: behaves as if there were only an inductance of L the straight line of in Fig. 2 is only affected at the resonance frequency. This allows the use of a current ripple criterion to estimate the current ripple in the inverter. This relation between the design parameters concerns the amplitude of the saw tooth waveform due to the PWM effect. The effect is maximum when the effective PWM duty cycle is 50 %. The expected variation of the current during one PWM period is L (7) Fig. 3. Bode diagrams of laboratory output currents (I =U ). Fig. 4. Bode diagrams of input admittance (I =E). Again, this variation of the inverter current is to be compared to the rated inverter current. Usually, the switching frequency can be adjusted to meet this criterion. For a given inverter, with its current rating and current measurement range, the effect of the saw tooth current (7) and the necessary reactive current at no-load must be considered simultaneously. For instance, if an inverter has a rated current and measurement range of 15 A, the saw tooth current and the reactive current should not exceed 5 A. This way, the controller has a free control range of 10 A. Taking these criteria into consideration, the complete output stage can be designed to optimally fit the rating of the given inverter. B. Laboratory Output Stage In the laboratory, two output stages have been put to test. The first contains no resistive damping elements, apart from the parasitic resistance of the inductances. The second one is intentionally damped. The parameter values are (for the case without 4.1 mH, 0.95 C 20 F, L damping): L 1.5 mH and 1.5 . The output stage with damping has 31.5 . Figs. 3 and 4 show Bode diagrams of the influence (of, respecand the grid voltage ) on the tively, the inverter voltage current of primary interest, i.e., the grid current . The lower traces correspond to the damped output filter, while the upper traces concern the output filter with no damping. The arrows indicate the grid frequency of 50 Hz and the switching frequency of 7.2 kHz, to illustrate the suppression of the latter. BOLSENS et al.: MODEL-BASED GENERATION OF LOW DISTORTION CURRENTS 1035 Note the appearance in Fig. 4 of a transmission zero due to the parallel resonance of L and C near 550 Hz. Signals on the grid near 1 kHz may also excite the resonance frequency of the undamped filter. Taking a look at the aforementioned design criteria, and 450 V, 230 V, the undamped topology is taking estimated to produce currents with a maximal steepness of L (8) In practice, a rate limiter is inserted in the signal path of the reference current to obtain a rate limitation to 8 kA/s. The four-switch topology of the inverter, operating at 3.6 kHz, produces an effective switching frequency (as seen by the output stage) of 7.2 kHz. Following saw tooth current is hence estimated: (9) Finally, the expected reactive current of the capacitors of the output stage, can be estimated (only the fundamental is considered here) (10) These values can be summed and compared with the rated inverter current. This gives a no-load burden of 4.63 A of the inverter, rated at 15 A leaving about 10 A of grid current available for control. Fig. 5. Developing the 14-state discrete-time model. The inverter voltage output and its effect on the LCL output stage can efficiently be described in the discrete-time domain, since the assumption that the output voltage is constant during the PWM period is valid here (Fig. 5, left). 1) Grid Voltage Influence Model: A continuous-time oscillation at frequency is determined by following state space description: C. Control Loop Design (11) Once the hardware has been designed, it is possible to construct the control loop. In this work, the controller is implemented using a state space description. In a first step, a correct input/output model must be developed. Fig. 5 shows the approach used: the superposition principle is used to obtain the influence of grid and inverter voltage on the LCL stage separately, and subsequently, these influences can be summed to obtain the complete system description. This sum is indicated at the bottom of Fig. 5. The LCL output stage is connected to the inverter and the grid. The grid is modelled as a voltage oscillator. If modelled using a discrete-time oscillator, the output can be connected directly to the discrete-time model of the LCL output stage (Fig. 5, right). This approach assumes the grid voltage to be constant during one sample time. In reality, the grid voltage evolves continuously in time, and a description using a discrete time oscillator introduces a small phase and gain error. To avoid this, the grid is modelled using a continuous-time oscillator and the resulting voltage is applied to the continuous-time model of the LCL output stage. Subsequently, the combined model can be discretized (Fig. 5, middle). The states and are complementary, and leads e by exactly 90 . The complete grid voltage model, incorporating oscillators at 50, 100, 150, 250, and 350 Hz, hence becomes (12), shown at the bottom of the next page. As can be understood from the matrix product at the bottom line in (12), the output of the system is the sum of all states. The states are only internal oscillator states, that do not appear at the output. The input is noise, and may change amplitude and/or phase over time, corresponding to the real behaviour of the grid voltage, which is never exactly 50 Hz all the time. The model requires 10 states. The oscillator output voltage is subsequently connected to the corresponding input of the continuous-time model of the output stage. This state space model has been derived from the fundamental equations (Kirchhoff’s Laws applied to Fig. 1) [10], [11]. There are three reactive elements (LCL), and this yields a third-order system. The inputs, in vector , are the and the grid voltage . The state variables inverter voltage are the currents through L and L , as well as the capacitor 1036 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 21, NO. 4, JULY 2006 voltage. They are also the output of the system. Note that the sign conventions of Fig. 1 are respected (13) Fig. 6. Grid current I response on noise input impulse. The resulting continuous-time model is finally discretized (using zero-order hold), and represents an accurate model for the influence of the grid voltage e on the states , and u of the output stage. To illustrate the 13-state combined model, of the the following signal is applied to the noise input N model: (14) This impulse signal excites the grid oscillator at once, and the strength is such that the grid voltage output becomes 230 V at 50 Hz. The response of the grid current is indicated in Fig. 6, showing that the LCL output stage is underdamped. The inverter voltage is assumed to be zero, i.e., a short circuit, explaining the large current. The steady state value can also be calculated using complex numbers (15) 2) Inverter Voltage Influence Model: The effect of the inverter output voltage is entirely described in the discrete-time domain, and requires the conversion of the continuous-time model (13). Furthermore, to account for the delay effect of any digital control loop, a delay block is inserted before the input of the LCL output stage, providing the necessary delay in the model. This system also introduces a noise input w for Fig. 7. Combining the system matrices. the process noise of the inverter operation. Altogether, both and , systems take four states: the LCL output states . as well as the delayed control voltage 3) Combining the Influences in One Model: In previous sections, two models are derived: a 13-state model to describe the effect of the grid voltage on the output stage, and a four-state model for the influence of the inverter voltage on the output stage. Both discrete-time models are now combined, and the reand are added, acsponses for corresponding states of cording to the superposition principle. This can only be done if the system matrices of both separate systems contain a common part, indicated by the shaded square on Fig. 7. The two original system matrices are displayed in bold lines: the upper left matrix (12) BOLSENS et al.: MODEL-BASED GENERATION OF LOW DISTORTION CURRENTS is the four-state inverter influence, and the bottom right matrix is the 13-state grid voltage influence. The resulting system matrix is 14 14. 4) State Observer Design: To reduce the number of sensors only the measurement of the inverter current is used to estimate all 14 states. As this system is fully observable through the inverter current measurement (the observability matrix is of full rank), a stationary Kalman estimator can be built. The inputs to and the estimator are the reference inverter output voltage the current measurement of . The measurement noise of is obtained by monitoring samples of a dc current over a longer period, and calculating the 50 mA. variance. This yields The process noise inputs are w (inverter) and W, the ten grid is estimated at 0.5 V . voltage oscillator noise inputs. The ten “noise” inputs for the grid estimation have been determined by trial and error, as practiced by [9]. The reason is that if the convergence is too fast, the measurement noise results in a noisy signal of the oscillator output, and audible noise in the inductance L in the output stage. On the other hand, if it is too slow, the settling time is large, and large currents appear before the grid voltage is estimated properly. These noise parameters 40 V/s. are set at 5) Reduced State Vector—Disturbance Decoupling: If the system’s state is known, the state vector can be fed back to the input to obtain a good closed-loop performance. The system at hand, however, is not fully controllable, due to the internal grid voltage oscillators. To solve this problem, a reduced state vector is created, containing only the controllable states and . To eliminate the contribution of the grid voltage on the system’s states, a compensating voltage is fed to the inverter reference input. It is impossible however, to eliminate the grid voltage effect in all states at once. In this work, the most important state is the grid current , and the compensating voltage is calculated to obtain a grid current free of disturbance. This implies immediately that the three other states still contain grid frequency components. In fact, the has to track the grid voltage e perfectly to capacitor voltage 0, and charging of C is done through . obtain The calculation of the compensating voltage is done by considering following transfer functions from grid voltage e and to the states , and : inverter voltage 1037 the grid voltage effect in the grid current, following complex division is done for every frequency: (17) The resulting five complex numbers are split into their real and imaginary components and are the feedback gain from the to the state vector’s oscillator components inverter reference input to perform compensation. The second step is to erase the effect of the grid voltage and the compensating voltage in the remaining three states. To achieve this, following four complex numbers are calculated for every frequency (18) The same procedure applies: all complex numbers are split into real and imaginary part and make up a 4 10 gain matrix. are multiIf again the 10 oscillator states plied, the resulting 4 1 vector erases the grid voltage effect in the remaining states. This concludes the construction of a reduced state vector. 6) State Feedback: The reduced state vector represents as only input, and the state of a reduced system, having and as states and output. State feedback for this fourth-order system can be made by pole placement techniques or an other criterion. In this work, a linear-quadratic regulator is designed using Matlab. The cost function J minimized, puts the stress on the grid current : (19) 7) Reference Input: The final step is the insertion of the reference signal in the controller. In this work, the reference is fed to the voltage reference input of the inverter, through an appropriate static gain. To optimize the step response of the system, an additional preconditioning filter can be inserted. An overview of the Simulink control loop is given in Fig. 8. (16) D. Laboratory DSP System Description The control algorithm is converted from the data flow description in MATLAB/Simulink to “C”-code using MATLAB’s Real-Time Workshop feature. The algorithm is then executed in a software framework on a Texas Instruments C6711 150-MHz digital signal processor (DSP). The total processor load is 89% with the algorithm running at 7.2 kHz. An FPGA daughtercard on top of the C6711 DSK board (accommodating the C6711 DSP) generates the PWM signals and provides an interface to the IGBT PWM inverter and the current transducers. These eight transfer functions are evaluated in the complex domain (in continuous-time resp. discrete-time) for the five frequencies of the voltage oscillator. To obtain the cancellation of III. SIMULATIONS AND MEASUREMENTS Simulations are carried out to evaluate the system’s convergence after startup. The voltage observer must provide an ac- 1038 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 21, NO. 4, JULY 2006 Fig. 8. Overview of the control loop. Numbers indicate vector dimensions. Fig. 9. Grid voltage estimator convergence at startup. Fig. 11. Grid current for various output stages (PWM disabled). Fig. 10. Suppression of the grid influence on inverter and grid current. curate estimate of the grid voltage for the control loop to work properly. Figs. 9 and 10 show that the observer is capable of suppressing the influence on the grid current. The reference signal is a 5-A square wave, not synchronized to the grid. The lighter trace in Fig. 10 is the inverter current, clearly showing current saw tooth ripple due to the PWM switching. Also the grid influence is felt in the inverter current, but not in the grid current after the output stage filtering. Measurements are carried out on the laboratory LCL and LC output stage, to evaluate the effect of different topologies, and damping. In Fig. 11, the PWM inverter is disabled, and the grid current is the capacitive output stage current. The grid voltage is also shown. For clarity, the currents for the various cases are given an artificial offset. The upper current trace is for the undamped LCL topology, the middle trace for an LC topology, and the bottom trace for a well damped LCL topology. Clearly, the undamped LCL system shows resonance, though the amplitude is not large. If an LC topology is used, the small parasitic grid inductance resonates with the capacitor, explaining the higher resonance frequency. Only the well damped LCL system of the bottom trace shows no resonance. Note that all currents are highly distorted due to the voltage distortion of 2%–3%. The peak value of this current is to be compared to the expected value of 1.8 A when designing the output stage, and where only the fundamental is considered. In Fig. 12, the control loop is operational, and the reference grid current is zero. As in the remaining figures, the lighter trace BOLSENS et al.: MODEL-BASED GENERATION OF LOW DISTORTION CURRENTS 1039 Fig. 12. Inverter and grid current, various output stages (control operational). Fig. 14. Inverter and grid current for square wave reference. Fig. 13. Inverter and grid current for LCL (upper) versus LC (lower) output stage. shows the inverter current, while the black trace is the grid current, after passing the output stage. The grid voltage is also displayed in each figure. Both damped LCL and undamped LCL topologies are shown in Fig. 12, each on a different offset. The undamped topology (upper traces) still exhibits a little resonance, negligible on the full scale of the current. The small active current drawn from the grid covers the power losses inside the PWM inverter, and keeps the dc link voltage at 450 V. Also clearly visible is the amplitude of the saw tooth current in the inverter, expected to be about 2.79 A. Since the rated inverter current is about 15 A, positive and negative, the upper and lower boundaries of the figure are also the inverter’s boundaries. Even at no-load a part of the inverter’s current range is used. is given. In Fig. 13, a reactive reference current of 5 A The PWM harmonics are well suppressed, and the grid current has a THD of only 0.93%. The current is 90 out of phase with respect to the grid voltage. As can be seen, the reference current consumes about the entire inverter current range. The figure also shows the inferior performance of the LC output stage, compared to the undamped LCL topology (shown on a different offset). To assess the performance of the controller on higher bandwidth applications, a square wave reference current of 5 A is taken (not synchronized to the grid frequency). The slope of this signal is also limited to 8 kA/s. Figs. 14 and 15 show a global and detailed picture of the various currents. Note that the saw tooth inverter current is minimal when the grid voltage crosses zero. Finally, Fig. 16 shows the current transients when the inverter suddenly switches off. Clearly, the output stage resonance Fig. 15. Detail of Fig. 14. Fig. 16. Transient grid current when the inverter switches off. slowly fades away. The influence of the oscillatory current on the grid voltage is also noticeable. IV. CONCLUSION In this paper, the hardware design and control aspects of a single phase LCL output stage are treated. By designing a control loop that observes the grid voltage and distortion, a very good grid current performance is achieved. Total harmonic distortion reaches values below 1%, even in the presence of a voltage distortion of 2%–3%. The LCL output stage also 1040 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 21, NO. 4, JULY 2006 efficiently suppresses the switching frequency currents in the grid. The control loop requires only the measurement of the inverter current. Future research is focusing on optimizing the control loop parameters, and reducing the DSP load, so that more DSP computation time is available for more complex algorithms, such as coupled to the grid. Also the exUPS-devices, active filters, tension to a three-phase application is studied. Karel De Brabandere (S’05) received the M.Sc. degree in electrotechnical engineering from the Catholic University of Leuven (K.U. Leuven), Belgium, in 2000 where he is currently pursuing the Ph.D. degree in the control of grid-connected power electronic inverters for the integration of distributed energy resources in island grids and the utility grid. His research interests include power electronics, digital control, distributed power systems, and microgrids. REFERENCES [1] H. Akagi, “New trends in active filters for power conditioning,” IEEE Trans. Ind. Appl., vol. 32, no. 2, pp. 1312–1322, Mar./Apr. 1996. [2] S. Ponnaluri and A. Brickwedde, “Generalized system design of active filters,” in Proc. IEEE PESC’01, 2001, p. 6. [3] F. Pöttker de Souza and I. Barbi, “Single-phase active power filters for distributed power factor correction,” in Proc. IEEE PESC’00 , Galway, U.K., 2000, pp. 500–505. [4] N. Mendalek, K. Al-Haddad, L. Dessaint, and F. Fnaiech, “Nonlinear control strategy applied to a shunt active power filter,” in Proc. IEEE PESC’01, 2001, p. 6. [5] T. C. Wang, Z. Ye, G. Sinha, and X. Yuan, “Output filter design for a grid-interconnected three-phase inverter,” in Proc. IEEE PESC’03, Acapulco, Mexico, 2003, p. 6. [6] M. Lindgren, “Control of a voltage-source converter connected to the grid through an lcl filter—Application to active filtering,” in Proc. IEEE PESC’98, 1998, pp. 229–235. [7] B. Bolsens, J. V. D. Keybus, J. Driesen, and R. J. M. Belmans, “Single phase active filter using digital PWM,” in Proc. IEE PEMD’02, Bath, U.K., 2002, p. 6. [8] M. Prodanovic and T. C. Green, “Control and filter design of threephase inverters for high power quality grid connection,” IEEE Trans. Power Electron., vol. 18, no. 1, pp. 373–380, Jan. 2003. [9] A. A. Girgis, W. B. Chang, and E. B. Makram, “A digital recursive measurement scheme for on-line tracking of power system harmonics,” IEEE Trans. Power Delivery, vol. 6, no. 3, pp. 1153–1160, Aug. 1991. [10] H. Meyr, Regelungstechnik und Systemtheorie II: Zeitdiskrete Systeme, Darstellung im Zustandsraum. Aachen, Germany: RWTH, 1992. [11] G. F. Franklin, J. D. Powell, and M. Workman, Digital Control of Dynamic Systems, 3rd ed. Reading, MA: Addison-Wesley, 1998. Bruno Bolsens (S’05) received the M.Sc. degree in electrotechnical engineering from the Catholic University of Leuven (K.U. Leuven), Belgium, in 2000 where he is currently pursuing the Ph.D. degree in power electronic circuits with limited electromagnetic interference properties in electric distribution networks. His research interests include power electronics, digital control, inverter topologies, power quality, and EMC issues. Jeroen Van den Keybus (S’05) was born in Belgium in 1975. He received the M.S. and Ph.D. degrees in electrical engineering from the Catholic University of Leuven (K.U. Leuven), Belgium, in 1998 and 2003, respectively. He is currently a Research Assistant with the Electrical Engineering Department, K.U. Leuven, where he performs research in the area of power electronics, and has a wide working knowledge and expertise in designing complex digital systems using reconfigurable and dedicated signal processing components. Johan Driesen (S’93–M’97) was born in Belgium in 1973. He received the M.Sc. and Ph.D. degrees from the Catholic University of Leuven (K.U. Leuven), Belgium, in 1996 and 2000, respectively. He is currently an Associate Professor at K.U. Leuven and teaches power electronics and drives. From 2000 to 2001, he was a Visiting Researcher in the Imperial College of Science, Technology and Medicine, London, U.K. In 2002, he was with the University of California, Berkeley. Currently, he conducts research on distributed generation, including renewable energy systems, power electronics and its applications, for instance in drives and power quality. Ronnie Belmans (S’77–M’84–SM’89) received the M.Sc. and Ph.D. degrees in electrical engineering from the Catholic University of Leuven (K.U. Leuven), Belgium, in 1979 and 1984, respectively, and the Special Doctorate and Habilitierung degrees from the Rheinisch-Westfdlische Technische Hochschule Aachen (RWTH), Aachen, Germany, in 1989 and 1993, respectively. Currently, he is a full Professor with K.U. Leuven, teaching electric power and energy systems. He is also a Guest Professor at Imperial College of Science, Medicine and Technology, London, U.K. His research interests include technoeconomic aspects of power systems, power quality, and distributed generation. Dr. Belmans Chairman of the Board of Directors of ELIA (the Belgian transmission grid operator).