EPE20100400007_12298577

Energy and Power Engineering, 2010, 2, 262-270 doi:10.4236/epe.2010.24038 Published Online November 2010 (http://www.SciRP.org/journal/epe) Copyright © 2010 SciRes. EPE Novel Control Strategy for Multi-Level Active Power Filter without Phase-Locked-Loop Guojun Tan, Xuanqin Wu, Hao Li, Meng Liu School of Information and Electrical Engineering, China University of Mining and Technology, Xuzhou, China E-mail: gjtan@cumt.edu.cn,cumt_wuxuanqin@163.com Received June 13, 2010; revised August 9, 2010; accepted September 21, 2010 Abstract Active power filter (APF) using novel virtual line-flux-linkage oriented control strategy can not only realizes no phase-locked-loop (PLL) control, but also achieves a good inhibitory effect to interfere. However, there are some problems in the conventional method, such as the error of amplitude, the shift of phase angle and the non-determinacy of initial oriented angle. In this paper, two one-order low-pass filters are adopted in- stead of the pure integrator in the virtual line-flux-linkage observer, which can steady the phase and ampli- tude. Furthermore, an original scheme of harmonics detection under the rotating coordinate is advanced based on the simplified space vector pulse width modulation (SVPWM) strategy. Meanwhile, by using the new SVPWM algorithm, the voltage space vector diagram of the three-level inverter can be simplified and applied into that of two-level inverter, and this makes the control for Neutral Point potential easier. Keywords: Active Power Filter, Harmonics Detection, Virtual Line-Flux-Linkage Observer, Active Power Filter Control without Phase-Locked-Loop, Space Vector Pulse Width Modulation 1. Introduction The use of nonlinear loads such as power electronic de- vices leads to serious harmonics pollution and lager voltage fluctuation. Moreover, lager unbalance voltage and current distortions in the power system are harmful to the electrical equipment and power systems. Shunt active power filter (SAPF) can well compensate the har- monic whose frequency and amplitude both changes, and it is also recognized as an effective way to manage the grid harmonic, to reactive power pollution and to im- prove the quality of the power. Active power filter (APF) is a new power electronics device for dynamic harmonics restriction and reactive compensation without influenced by system inductance [1,2], besides it has such functions: adaptive ability for the parametric variation of system and load, automatic tracing and compensation for the varying harmonics [3,4]. There are some remarkable advantages in nature of three-level APF, such as lower distortion of output waveform, lower endure voltage and less switch loss, high efficiency and low electromagnetic interference (EMI), so it helps to raise the installed ca- pacity and improves the harmonics compensation effect as well as system reliability. On the one hand, the traditional harmonic current de- tection which is based on instantaneous reactive power adopts phase-lock-loop (PLL) to acquire the voltage vector angle [5-7], when the grid voltage fluctuation is more serious, PLL will be in unlocked condition because of the larger frequency offsets which can not accurately track the phase position. To solve the problem, this paper uses the virtual line-flux-linkage orientation to observe the vector angle, which converts the observation of vec- tor angle to the flux. By using the voltage integral on the AC side of the active filter to estimate the grid flux, PLL can be omitted; meanwhile, the power grid interference can be inhibited well. On the other hand, the traditional harmonic current detection requires that the integrated vector combined by the sine and cosine function should be synchronous and phase coincidence to the integrated vector of the three-phase positive sequence fundamental voltage, otherwise, the detection accuracy of the funda- mental positive sequence reactive component will be affected by phase difference [8]. Therefore, the harmonic detection principle based on the rotating coordinates is proposed in this paper. The application of virtual flux in active filter system also includes controlling generation of compensation current, so it has a very good control to harmonic current detection and compensation current ge- G. J. TAN ET AL. 263 neration, and the voltage space vector modulation strat- egy SVPWM (Space Vector Pulse Width Modulation) can be easily applied to active filter control. In this paper, the simplified three-level SVPWM with neutral point potential adaptive control is applied to con- trol multi-level active filter, based on this, a novel multi- level voltage active filter phase reconstruction algorithm is proposed. 2. Mathematical Model for Three-Level APF and the Rotating Coordinate Based Harmonics Detection Diagram Figure 1 is the main circuit of NPC three-level APF, while vector diagram of the virtual line-flux-linkage ori- ented system is shown in Figure 2. When the virtual line-flux-linkage is at axis, then mathematical model of three-level APF is in reference frame [9,10]: d dq Z X AX Be    (1) where:  0s s d dZ diag L L C C (2) 1 2 1 2 1 1 2 2 0 0 0 0 0 0 1 0 0 1 0 0 1 1 0 S d d S q q d q d q L S S L S S A S S S S               (3) 1 2 0 T d q dc dcX i i V V    (4)  1 1 0 0 0B diag (5) 0d q m e e E          (6) In (2) to (6), sL S S : AC side inductance of APF; d : capacitance of DC side; m : peak voltage of power system; : the D-axis and Q-axis AC current of APF; 1 2dc dcV V : the upper and down DC voltage relevant to neutral point; 1 1 2 2d q d q : the D-axis and Q-axis components of switching function. C E ,S ,d qi i , , , S According the virtual line-flux-linkage oriented strat- egy showed in Figure 2, the reference compensation current can be obtained: ~ * ~ * dh Ld qh Lq i i i i    (7) ~ ~ Ld Ld Ld Lq Lq Lq i i i i i i        (8) where: ,Ld Lqi i : The D-axis and Q-axis current of nonlinear load; ,Ld Lqi i   : The D-axis and Q-axis fundamental current of nonlinear load through low pass filter from ,Ld Lqi i ; ~ ~ ,Ld Lqi i : The D-axis and Q-axis component harmonics current of nonlinear load. 3. The diagram of Novel Virtual Line-Flux-Linkage Observer The essence of virtual line-flux-linkage oriented method is to gain accurate space angle ( ) estimation of the ori- ented vector  , as shown in Figure 2, the angle can be obtained: 2 2 2 2 sin cos                    (9) Then the estimation of  is converted to the evalua- tion of  and  components of the virtual line-flux- linkage. die L v dt di e L v dt            (10) In (10), v and v are the  ,  voltage quantity at AC side of three-phase APF, respectively. Integrated both sides of (10), v dt Li v dt Li                (11)  ,  :  ,  value of virtual line-flux-linkage. It can be seen from (11) that the flux linkage can be evaluated by the voltage integral at AC side of APF, while it brings a problem of the integral initial value which cause an error of the flux linkage, furthermore, the method of pure integrator doesn’t retrain the DC com- ponent of the input signal, even a little DC component can make the integrator saturation. In this manner, the flux linkage in  axes can be a circular trajectory with DC offset corresponding to the centre of the circle, meanwhile, it cause the inaccuracy of virtual line-flux- linkage oriented angle and impact both the authenticity of current feedback and the veracity of voltage space Copyright © 2010 SciRes. EPE G. J. TAN ET AL. Copyright © 2010 SciRes. EPE 264 Figure 1. Topology of a diode NPC three-level shunt APF. 1 s   d  q      ei e i E  V  I  IL  qi di Ldi Lqi Figure 2. Vector diagram of the virtual line-flux-linkage oriented system. nd even aggravate the performance of PF seriously. As a result, it brings large current shock blems of pure integrator, it usually ter (use the first order inertia fil ) instead of the f pure integrator ( vector supply, a A in the process of starting, and even can’t start. Therefore, in order to achieve an accurate flux estimate, it’s neces- sary to make some measures to eliminate influence of the integral initial value. In the traditional virtual flux estimation, in order to solve initial value pro 1 ) in flux estimation. It can be seen s from the formula of the first order that the first or- filter der inertia is similar to the pure integrator when the fre- quency  of the inusoidal input is far greater than s f . Howev r, it is also required that the first order i a e nerti should have a certain decay to the direct flow, that means f should be a certain value, and not too small, other- wise the ecay will be very slow .So the first order low d pa mea ss filter has contradictions between approximating pure points and decaying of the DC component, that ns f must maintain a certain value to decay, while it also should be far less than  to approximate the integral. Know from the above analysis. Low-pass filter is adopted by conventional virtual line-flu nkage estimate to substitute for the purely integrator to least the DC drift, howev , it would lead to errors of amplitude and phase angle. As desc x-li er ribed by the equation ( s c c N NW s s    ), the novel virtual line- flux-linkage observer and the comparison of three ob- servers are showed in Figure 3 and Figure 4 respec- tively. Acc , it is quite easy to found ording to Figure 4 that the proposed algorithm respond faster than that of conventional way. It can be seen from the bode figure G. J. TAN ET AL. 265 v i  v i   cc s N s N   cc s N s N   Figure 3. The novel virtual line-flux-linkage observer. Figure 4. The comparison of the three observers. Bode Diagram Frequency (rad/sec) 10 1 10 2 10 3 10 4 -180 -135 -90 -45 0 System: sys Frequency (rad/sec): 314 Phase (deg): -90 Ph as e (d eg ) -120 -100 -80 -60 -40 System: sys Frequency (rad/sec): 314 Magnitude (dB): -49.9 M ag ni tu de (d B) Figure 5. Bode diagram of the novel virtual line-flux-link- age observer. that it can replace the pure integrator by making the am- plitude decay 49.9db and phase shift 90° at the same time while 314 /Rad s  . 4. The Simplified SVPWM Algorithm and rit sp tor diagram of the three-level inverter can be garded as a hexagon composed by six small two-level se neutral-point voltage de Novel Voltage Estimation for Three-Level APF In this paper, the simplified three-level SVPWM algo- hm is adapted, as shown in Figure 6. The voltage ace vec re space vectors, and all the hexes are centered by vertexes of inner one. Hence, two-level SVPWM algorithm can be applied to calculate the duration-time and the switch sequence of voltage vector [11]. Because of inherent problem in the topology of the diode-clamping three-level converter, the various switch states have different impacts on the neutral-point poten- tial. The middle vectors can cau viation because of asymmetric parameter in practice. Small vectors will cause the fluctuation of neutral-point potential. There is a striking contrast effect between the two different middle vectors corresponding to different switch status vector [11]. There exist the regions that are overlapped by adjacent small hexagons as shown in Figure 6. So if the reference voltage vector stays at those regions, S can have any values that are possible. 1_s refV is the corrected refer- ence voltage vector when the index S has the value of 1, and 2 _s refV is the corrected reference voltage vector when the index S has the value of 2. If the two-level plane of S = 1 is selected switching sequence de- cided by reference voltage vector is given as follows: 1 2 8 1(0 ) (0 0 1) (1 0 1) (0 1 1)N N PV V V V , the 1 1         1NV , 2NV are respectively voltage space vectors of the corresponding switching sequence (0-1-1) and (0 0-1), both negative short vectors; (1 0 0) is voltage space and they a 8 , T re a positive short vector are PV1 vector. es o 1N 2N 1P dwelling times of the corresponding volt- age vectors. The voltage vector 1NV and 1PV have the same output voltage 1V , and dwelling tim f the 1NV , 1 T , T , T PV are equal. If the above switching se nc p the dwelling time of negative short vectors is longer than the positive ones; neutral-point potential will decline. If the two-level plane of S = 2 is selected, the ching sequence is given as follows: (0 0 1) (1 0 1) (1 0 0) (11 0) que e is ad s o te wit d,      , the dwelling time of negative short vectors is shorter than the positive ones, neutral-point potential will rise. So the neutral-point po- tential can be controlled by changing th As can be seen from (11), on the basis of the proposed algorithm, the voltage at the terminals of three-level rec- e corresponding the value of index S. tifier is estimated, which is shown in Figure 7. When the reference voltage is in the first sector, esti- Copyright © 2010 SciRes. EPE G. J. TAN ET AL. Copyright © 2010 SciRes. EPE 266 Figure 6. The simplified three-level SVPWM algorithm. Figure 7. The novel three-level voltage estimation algorithm. mated three-phase voltages are given by: 1 1 2 1 1( )acalc dc dc aon bon conu u u t t t        3 3 3 3 3 1 1 ) 3 3 3 3 3 1 1 2 1 1cos(60 ) ( ) 3 3 3 3 3 aon con ccalc dc dc con aon bon t t u u u t t t             (15) where, are the estimated voltage of DC bus voltage of three- are the duration-time of three-phase for two-level algorithm, respectively. Design and Experimental Results Analysis To verify the viability of proposed non PLL control scheme for multi-level APF and evaluate performance of this method, the test platform is established and operated, and its control sketch is showed in Figure 9. As shown in Figure 10, the structure of full-digital controller is composed of DSP (TMS320F2812 of TI) and FPGA. The system can implement DC bus voltage contro 5. The Main Circuit l, har- 1 1 2cos(60 ) (bcalc dc dc bonu u u t          acalcu , three-level rectifier. level rectifier. bcalcu , aont , ccalcu dcu is bont , tcon Figure 8. Experiments waveform of the novel three-level voltage estimation algorithm. G. J. TAN ET AL. 267 * qi * dhi di qi * dcu ai bi dcu* du * qu * u * u as bs cs sin cos sin cos * qhi Lai Lbi IPARK PARK aL bL cL Load Linear N on LdiLqi HPF AlphaCalcu BetaCalcu Figure 9. Novel diagram of the virtual-flux oriented vector control diagram of APF. Figure 10. Structure of full-digital controller. monics detection, close osis and etc. 5.1. The Principles of the Main Circuit Parameter Calculation 1) The design of inlet inductance on AC side The design of AC filter inductor has two principles: for one thing, it is the power of active power filter to track and control compensation current, and make sure it can still produce a corresponding compensation current while the load current has larger current rate of change; for another, it is to satisfy the requirements of trackin the size of compensation current ripple. It can be known from [12] that: d-loop current control, fault diag- n g c max * 4 9 (10 ~ 20) 2.3 d a U L f I    (16) The maximum inductance value can be acquired by the above equation, while the minimum inductor value is determined by the size of allowable ripple current, the size of the ripple current should be limited within the prescribed range while choosing the inductor value. min max 2 9 dc sU TL i   (17) Combined with (16) (17), It can be obtained: c 9 9 (10 ~ 20) d s dcL i f     *max 2 4 2.3 a U T U I (18) where, cdU is the DC bus voltage; sT is swi cy; maxi tching equenfr  is the maximum allowable value that compensation current deviate from the reference current; f is the current frequency of the fundamental wave; * aI is the effective reference current. osses cause current in the APF mpensation current and the energy pulse caused by the apacitance is difficult to maintain a constant. The larger the capaci- tor is, the smaller such fluctuations would be, but the 2) The design of the DC bus capacitor Owing to both the energy pulse and switching l d by harmonic and reactive co filter inductor energy storage on AC side, the c Copyright © 2010 SciRes. EPE G. J. TAN ET AL. 268 capacitance value is not unlimited, so the design princi- ples of DC bus capacitor is to work as a minimum ca- pacitance while APF can be under normal operative con- dition. It can be known from [12] that: min 2 cS TC  (19) c(1 ) dU  maxdc dc U U   w (20) here, cS is the compensating capacity of APF, T is the control cycle of DC bus voltage, cdU is the DC voltage,  is the voltage wave pace, maxdcU is the maximum allowable value of the voltage wave. 3) The controllable range of DC bus voltage DC bus voltage selection should be considered as fol- lows: it is not only necessary to function, but also should not lead to too fierce current ch ase APF should be greater than the peak value of AC power grid line voltage, that is achieve a certain current anges. The DC voltage threshold of three-ph c 3d mU E , otherwise, can not be tracked. Take grid fluctuations and linear con- e a er be the current trol rang nd other factors into consid ation, the DC voltage can c 3 2d mE . Where, mE is the peak value of the phase voltage on AC side. According to the above design principles, the main cuit parameters of the APF test platf U cir orm are as follow: ansformer: 380 V d-side transformer is used to buck-boost in the hi e effect caused by the leakage reactance of transformer can the non-linear load is of a large ca portion harmonic. As can be seen in Figure 11, the voltage distortion caused by tra makes the phase-lock-loop in use not n ual flux orie ux, the virtual -lin er compensation, th tio low-pass fil n ux-linkage observer with two first-order low-pass fil no 3, the rientation angle and amplitude errors of the virtual flux the novel gorithm responses faster than traditional methods which a) Power line voltage: 380 V/50 Hz, tr /120 V, and system impedance is neglected; b) Three-phase uncontrolled rectifier bridge is adopted as the nonlinear load, its R = 8 inductance L = 1.5 mH; c) Incoming inductance of APF: L = 1.7 mH, DC bus voltage: dU = 320 V, capacitance: C = 2300 uF. 5.2. The Experimental Results Analysis The gri gh-power applications system, th not be ignored if pacity or in high pro- nsformer leakage