Elektrotehniski vestnik 78(5): 304-311, 2011 English edition Effective Controller Design for the Dynamic Voltage Restorer (DVR) for Voltage Sag Mitigation in Distribution Utilities Lin Xu1, Yang Han2 1Sichuan Electric Power Research Institute, No.24, QingHua Road, QingYang District, 610072 Chengdu, China E-mail: xulin198431 @hotmail.com 2School of Mechatronics Engineering, University of Electronics Science and Technology of China, No.2006 XiYuan Road, West Park of Chengdu High-Tech Zone, 611731 Chengdu, China E-mail: hanyang_facts @ hotmail. com Abstract. The voltage sag detection algorithm and control algorithm synthesis of the dynamic voltage restorer (DVR) are proposed. The double synchronous reference frame phase-locked loop (DSRF-PLL) is applied for the voltage sag detection, which is able to calculate the phase angle of the positive-sequence fundamental component of grid voltages even under unbalanced and distorted grid voltages. The effective control strategy of the DvR is proposed, which includes the reference compensation voltage generation, the control scheme for the shun-connected voltage source converter (VSC) and the control algorithm for the series-connected voltage source inverter (VSI). Finally, extensive simulation results are presented to verify the validity and effectiveness of the proposed control strategy. Keywords: Dynamic voltage restorer, phase locked loop (PLL), voltage sag, sequence decoupling, proportional resonant controller. 1 Introduction The modern industrial plant is subjected to abnormal shutdown or malfunction due to the power quality problems. one of the most concerning disturbances affecting power quality are voltage sags. Their major source are short-circuits on the utility lines. Faults many kilometers from the disturbed process will generate a momentary voltage sag in the electrical environment to the end user. A reduction of the short-circuit current magnitude may lead to a substantial power quality improvement because the majority of the sensitive industrial processes (including computer system, power electronics and variable speed drives) are capable of riding through a sag of a very limited amplitude [1-5,8]. in retrospect, many researchers have discussed the various algorithms to calculate voltage sag phenomena. In [6], a simple approximated method for calculation of voltage sags in an electrical network was presented. The paper illustrates the construction of a square matrix that, under acceptable assumptions, directly provides voltage drops at each node of the network and for each faulted node. The voltage sags matrix of the network, gives a representation of sags produced by symmetrical faults for the overall network. in [7], a three-phase precision measurement system for the electronic electricity meter was presented, which is based on a digital signal processor (DsP) in combination with a multichannel analog-to-digital (A/D) converter. Besides, the analyzer functions are implemented in compliance with the European EN5160 standard, and they include detection of short- and long-term voltage sags, swells and outages, harmonic analysis of current and voltage signals as well as assessment of voltage flicker. The solutions to mitigate the consequences of voltage sag have also attracted attention of the academic community in recent years. In [8], it was reported that, independently on its magnitude and shape, each voltage sag lasts as long as the protection equipment allows fault current to flow. A reduction of the clearing time of the lines protection equipment may lead to a substantial power quality improvement because the majority of sensitive industrial processes are capable to ride through the sag of very limited duration. In [9], the effects of limiting reactors (LR) connected at the beginning of the medium-voltage (MV) feeders are investigated, which allow for effective mitigation of voltage dips and are considered as a trivial solution, attractive for the reason of a very low cost, simplicity and reliability. With the development of modern power electronics, digital controllers, such as digital signal processors (DSPs), the dynamic voltage restorer (DVR) has been recognized as the best choice to protect the industrial facilities from voltage sag and other voltage disturbances. Fig.1 shows the typical circuit diagram of the DVR in a distribution system. It demonstrates that voltage sag may be incurred by the fault from the adjacent feeder or the fault from the transmission network. Therefore, the DVRs can be applied to protect the sensitive loads of high-tech industries with adjustable speed drives and other power electronic loads. For the industries with a high penetration of the induction motors, the energy storage might be used and a sophisticated controller must be adopted due to the inherit inertia of induction motors and their capability to withstand short duration, shallow sags and phase jumps [10]. Figure 1. The circuit diagram of the dynamic voltage restorer (DVR) for voltage sag mitigation in the distribution utilities. A schematic diagram of DVR is shown in Fig.2. The DVR inverters are based on three single-phase full-bridge voltage-source inverters (VSI) using common dc-link.These inverters are series coupled into the utility grid using three single-phase transformers to ensure galvanic isolation and to step-up the injected voltage. A capacitive filter is located at the line side to filter out the switching harmonics generated by the inverter. The dc link can be charged with voltage source rectifier (VSR) which is connected to the utility grid side [10]. VSabc tranforiner n: 1 •4u Shunt converter Series converter Figure 2.Single-phase representation the DVR system. In order to generate a fast and accurate reference voltage, the most important aspect to consider in the control of the DVR system is the proper synchronization with the utility voltages. Specifically, in case of voltage sag or harmonics, the detection of the positive-sequence voltage components at fundamental frequency is essential for the control of DVR. The detection of the amplitude and the phase angle of the positive-sequence component must be fast and accurate, even if the utility voltages undergo sag, distortion and/or unbalanced. In this paper, an effective controller design method for the DVR is proposed, which is based on the double synchronous reference frame phase-locked-loop (DSRF-PLL) for grid-synchronization, the reference injection voltage generation, the control schemes for the PWM voltage source converter (VSC) and series-connected PWM voltage source inverter (VSI). To ensure accurate compensation for the grid distortion, the proportional-resonant current controller is adopted in the inner current loop, which provides infinite controller gain at the selected harmonic frequencies. Hence, the DVR is capable to mitigate voltage sag, grid unbalance and distortion. Finally, the simulation results are presented to verify the validity and effectiveness of the proposed control algorithms. 2 Grid-Synchronization Algorithm using Double Synchronous Reference Frame PLL Considering a three-phase three-wire system, the zero-sequence component is null. To introduce the decoupling network for grid-synchronization, one supposes the voltage vector consisting of two generic components rotating with nm and mm frequencies respectively, where n and m can be either positive or negative with the fundamental frequency m. Hence, the voltage vectors in the a-ft plane can be denoted as [10]: S(aß) : ' VS (1) cos(nOt + (n) m cos(mOt + (m) vsß J sin(n«f + (n) S sin(mot + (m) Additionally, two rotating reference frames are considered, dqn and dqm, whose angular positions are n8 and m8 respectively, where 8 is the phase angle detected by the PLL. Hence the voltage vector vs can be expressed in the dqn and dqm synchronous frame as [11]: S(dq') ' VSq' iT* J' ■ Vn ■ VS cos(nO + (n - iff) sin(not + (n - iff) cos iff sin iff - sin iff cos iff +vm vs (aß) cos(matt + (m - iff) sin(max + (m - iff) (2) , i = n, m . where [Tq ] When the grid synchronization using the proposed PLL is achieved, that means 8=mt, hence we get: vs w)=vn -vn - vs cos((n) sin((n) cos((n) sin((n) +vsm cos(mO + ( - not) sin(mO + (m - not) +Vsm cos((m) cos((n - m)ot) - sin((n - m)ot) +vm sin((m) sin((n - m)ot) cos((n - m)ot) (3) v v v L v L c The same expression in dqm axes may be deduced by just interchanging m and n. The amplitude of the signal oscillation in the dqn axes depends on the mean value of the signal in the dqm axes, and vice versa. Therefore, in order to cancel the oscillations in dqn axes, the decoupling algorithm between the d-axis and q-axis quantities is proposed, as shown in Fig.3. The same cell may be used in dqm axes by just interchanging m and n in Fig.3. v r,- d + q+j-d d VSd LPF V q q LPF e vsdn, vsqn, vsdm, vsqm can be derived in the frequency domain as in the following equations: - af - - \ VSdn (s) = s + ( (VSdn (s) - u1(s) * VSdm (s) - ^(s) * Vgqm (s)) (4a) --(Of --- vSq' (s) = , . (vq (s) - Ui(s) * vSqm (s) + u2(s) * vSdm (s)) (4b) s + rnf Sq (Of Sqm VSdm (s) =-(VSdm (s) - % (s) * VSdn (s) + ^(s) * V n (s)) (4c s + (f ) --(Of --- VSqm (s) = (VSq.m (s) - Ul(s) * VSqn (s) - ^(s) * V^, (s)) (4d) Note that the asterisk '*' represents the convolution product of the signals in the frequency domain (s-domain). These equations can be transformed back to the time domain as: VSd• = (f (VSd" - VSd• - UlVSdm - U2 VSqm ) (5) Figure 3. Decoupling cell for canceling the effect of VSm on the dqn frame signals. In order to calculate the averaged quantities (dc values) of Vsdn, Vsqn, Vsdm, Vsqm, the cross term decoupling network is proposed, as shown in Fig .4. Notably, the low-pass filer (LPF) is introduced herein to calculate the dc components of these quantities in the synchronous frame (d-q frame) with the transfer function LPF(s)=Wf /(s+mf), where Wf denotes the cut-off frequency and ®f=35Hz is selected herein. VSd, = V, cos(^n) + u{Vsm cos(^m) + U2VSm sin(^m) (6) Moreover, Equation (6) can be rewritten: VSdn=(f [vSn cos(^n)+u^m cos(^m) +U2VSm sinif1) - VSd, - u1VSdm - VSqm ] (7) Followed by the same procedure, the mathematical equations for the as shown in Eq.(4) can also be derived. After manipulation of these equations, the following state-space model is obtained: \X (t ) = A(t).X (t) + B(t)U (t) IF (t ) = C.X (t ) (8) where Y^ = X t = [v V" cos(/)" V" sin(f) , , VSdm , V Sd"' Sq' v c r Sq ■] , C (t ) = 14 U(t) = vm vm sinT) A(t) = -B(t) = af -1 0 -« -U2 0 -1 -U1 « «2 -1 0 -U2 -U1 0 -1 In the case of n=1, m=-1, the positive and negative fundamental frequency components in the dq+ and dq~ axes are decoupled. Meanwhile, the DSRF-PLL should cancels the double frequency oscillation terms at 2® in the voltage V*sq+l. With the aim of V*sq+1=0, a proportional-integral (PI) control loop is added to manipulate V*sq+l where the phase error is [as shown in Fig.4]: — * = (kp + -L) X (V*q+i - 0) s Sq+' (9) Figure 4. Block diagram of the double synchronous reference frame PLL. The mathematical model of the double synchronous frame PLL can be derived by manipulation in the frequency domain. Referring to Fig.3, the unit sinusoidal functions are defined as: ui=cos[(n-m)mf], u2=sin[(n-m)mt], the averaged quantities (dc values) of a>c = 2nx 25 rad/s m2 k where kt = m+T , kp = 2-vcb _ ref -W I ^ v ^ vcc _ ref Figure 8. Block diagram of the reference voltage generation Fig.8 shows the simplified block diagram for the reference injection voltage generation algorithm. The quantities uDyR.d=Vsd+-Vsd-ref and uDVR.q=Vsq+ are compensated for the d-axis and q-axis, respectively. With the reference value, the magnitudes of the output terminal voltages are regulated to the constant rated line voltage Vsd-ref. The control variables in the d-q coordinates uDVR-d, uDVR-q are transformed back to the a-b-c coordinates by using inverse Park's transformation. Hence, the reference compensating voltages for the DVR in the a-b-c coordinates vca_rei, vcb.ref, vcc-ref are calculated. Next, the control scheme for the series connected voltage source inverter and shunt connected voltage source converter would be explained. 3.2 Control Strategy for the Series-Connected Voltage Source Inverter To achieve a good transient and steady-state performance, a multi-loop proportional-integral (PI) controller is proposed to control series converter of the DVR. As shown in Fig.9, an inner current control loop (feedback filter capacitor current iC) is adopted for ensuring fast response as well as attenuation of the filter LC resonance. The outer voltage loop is used for reference signal tracking. Normally, the control + bandwidth of the outer voltage control loop is designed to be five or ten times lower than that of the inner current control loop. The proposed controller achieves the response time of about 500lis to sudden voltage sag or swell in the grid without any oscillation, the performance of the proposed control strategy is validated by simulation results. Figure 9. The simplified diagram for the multi-loop control strategy ofVSI 3.3 Control Strategy for the Shunt-Connected, Voltage Source Converter As it is known, DVR operates to maintain the load supply voltage at its rated value, and during a voltage distribution, the DVR exchanges active and reactive power with the grid. If active power is supplied to the load from the DVR, it needs a source for this energy. In this paper, the dc link voltage is charged with the voltage source converter shunt connected to the grid. The dc voltage can be controlled to almost constant. Fig. 10 shows the complete control diagram for the DVR, which is capable of mitigating voltage sag disturbances even under unbalanced and distorted grid scenario. pu^ q» ^1 ,'l,'l / aß it — ^1 ,'I,'l aß PI P+resonant controller PI VSI ^' en _ ref - (n = a,b,c) Figure 10. A complete control strategy of the shunt-connected VSR and the series-connected VSI. P+resonant controller VSR k/^ VSß Both positive and negative sequence currents are controlled, the current command Idf is determined by the dc-link voltage controller, the required power P0 is obtained by multiplying vdc* to the output of the voltage controller if, i.e., p*=vdc*idf, the current commands are obtained as: tp* 1d jp* ? = —vdcx(PI)x(vdc-vdc) (10) where D= Kf-