Supply

Active power filters have been widely used for harmonic elimination. The performance of a conventional three-phase shunt Active Power Filter (APF) using Synchronous Detection Method (SDM) has been compared with Nonlinear Autoregressive-Moving Average (NARMA)-L2 based APF. The novelty of this study lies in the application of NARMA-L2 control to generate the amplitude of the reference supply current required by the APF circuit and the successful implementation of the APF system for harmonic elimination. The entire system has been modeled using MATLAB 6.1 toolbox. Simulation results demonstrate the applicability of NARMA-L2 controller for the control of APF.


INTRODUCTION
Power Quality (PQ) is an important measure of an electrical power system. The term PQ means to maintain purely sinusoidal current wave form in phase with a purely sinusoidal voltage wave form. The power generated at the generating station is purely sinusoidal in nature. The deteriorating quality of electric power is mainly because of current and voltage harmonics due to wide spread application of static power electronics converters, zero and negative sequence components originated by the use of single phase and unbalanced loads, reactive power, voltage sag, voltage swell, flicker, voltage interruption etc.
To improve the power quality traditional compensation methods such as passive filters, synchronous capacitors, phase advancers, etc. were employed. However traditional controllers include many disadvantages such as fixed compensation, bulkiness, electromagnetic interference, possible resonance etc.. These disadvantages urged power system and power electronic engineers to develop adjustable and dynamic solutions using custom power devices. Custom power devices are power conditioning equipments using static power electronic converters to improve the power quality of distribution system customers. These include APF [1][2][3][4][5][6][7][8][9][10][11] , dynamic voltage restorer (DVR) [12][13][14][15][16] and Unified Power Quality Conditioner (UPQC) [17][18][19][20][21] . APF is a shunt compensator used to eliminate the disturbances in current, whereas DVR is a series compensator used to eliminate the disturbances in voltage. Recently UPQC, which consists of both shunt and series compensators is proposed as a one shot solution for power quality problems.
This study proposes the effectiveness of NARMA-L2 controller for the control of APF system. The novelty of this paper lies in the application of NARMA-L2 controller for the determination of amplitude of reference supply current required in an APF system. This paper also discusses the control of APF system using SDM. The control strategies of APF system are detailed in the second part of this paper. Simulation results in the third part demonstrate the effectiveness of NARMA-L2 controller for the control of APF system.

CONTROL TECHNIQUES OF APF SYSTEM
In this section, control scheme of APF system using SDM is discussed and compared with that using NARMA-L2 controller. Principle of operation: A three-phase system has been selected to study the performance of the APF system. Comparison of Fig. 1 [22] and 2 [22] shows the compensation principle of a shunt active power filter.
APF injects a current equal in magnitude but in phase opposition to harmonic current. Figure 3 shows the block diagram of a three-phase APF system. The heart of the APF system is the IGBT based Voltage  Figure 4 shows the control circuit of an APF system using SDM. The control circuit consists of an outer voltage control loop and two inner current control loops. The outer control loop is used to maintain the capacitor voltage constant and to determine the amplitude of the mains currtents required in an APF system. SDM method is basically used for the determination of amplitude of the source currents. In this algorithm, the three-phase mains currents are assumed to be balanced after compensation. The real power P(t) consumed by the load could be calculated from the instantaneous voltages and load currents as:

Control of APF system using SDM:
(1) where v sa (t), v sb (t), v sc (t) are the instantaneous values of supply voltages and i la (t), i lb (t), i lc (t) are the instantaneous values of load currents. The average la sa sb sc lb value P dc is determined by applying P(t) to a low pass filter. The real power is then split into the three phases as follows: Thus for purely sinusoidal balanced supply voltages, where, V sma , V smb , V smc are the amplitudes of the supply voltages. With the objective of achieving Unity Power Factor (UPF), the desired mains currents are ca sa la obtained by equations 4a-c. The compensation currents are calculated using equations 5a-5c. Since the PWM VSI is assumed to be instantaneous and infinitely fast to track the compensation currents, it is modeled as a current amplifier with unity gain [23] .

Control of APF system sing NARMA-L2 controller:
NARMA-L2 controller [24,25] is successfully used for the control of UPQC [24] and the speed control of separately excited DC motor [25] . The simulink model of a NARMA-L2 controlled APF system is shown in Fig. 5. The simulink plant model shown in Fig. 6 is used to train the NARMA-L2 controller. As shown in Fig. 5, the control input of the plant is the amplitude of the desired mains current and the capacitor voltage is the plant output. NARMA-L2 controller checks the desired capacitor voltage and the actual capacitor voltage and the control input is adjusted to achieve the reference value. Table 1 shows the specifications used to train the controller. It has been observed that the complicated equations in SDM could be eliminated by the use of NARMA-L2 controller.
Sample performance graph and training data obtained from a NARMA-L2 controller are illustrated in Fig. 7 and Fig. 8 respectively.

RESULTS AND DISCUSSION
NARMA-L2 controller has been successfully modeled and tested for the control of APF system. MATLAB 6.1 toolbox is used to model the system. The effectiveness of the system has been tested for various firing angles (α) in the range of 0° and 90°. The performance of the developed system is illustrated with the one using SDM for two different firing angles as tabulated in Table 2.
Simulation results for α = 0° shown in Fig. 9 and Fig. 15 demonstrate the effectiveness of the developed system for the control of APF.

CONCLUSION
A NARMA-L2 controlled APF system has been modeled and successfully tested for the control of APF. The novelty of this paper lies in the application of NARMA-L2 controller to determine the amplitude of the reference source current required in an APF system. This paper also discusses modeling and control of APF system using SDM. The performance of the different system has been compared. It has been observed that the complicated calculations used in SDM could be eliminated by the use of NARMA-L2 controller.