Active Filter Based On Neural Networks

The harmonic pollution of power system is becoming more and more serious with the using of power electronic equipments in power system. The harmonics should be compensated because it can do much harm to the power system. The study on Active Power Filter (APF), which is a novel device compensating harmonics and reactive power current simultaneously, is the focus in the field of power quality control.The harmonic current detection is one of the key elements for the APF to keep a well dynamic performance. This paper, firstly, introduces the basic structure and principle of APF, and then makes comparison among some conventional harmonics detections methods. Based on the idea mentioned above, one harmonics detection method based on Hopfield neural network concerning. Compared with other approaches, this one process the sample data rapidly, need less training in advance and can detect accurately the amplitude and phase angle of the fundamental wave as well as the harmonics.Compensating current is another one of the key element for the APF to keep a well dynamic performance. This paper advance one three-phase four-wire shunt APF system which adopts the hysteretic control of compensating current method. The hysteretic control of compensating current method is simple and has good real time. According to requirement of instant current tracking, the estimating formula of output inductance is proposed and the critical value of dc bus voltage is derived. The compensation effect will be also influenced by the output filter of the power circuit Therefore, a LCL filter is designed and harmonic of switch frequency is eliminated effectively.Simulation models of three-phase four-wire shunt active power filter system based on this method are built in environment of MATLAB software. The simulation results show that harmonics detection method based on Hopfield neural network is superior to traditional control method, it can not only eliminate harmonic current effectively, but also can improve the improve the integral performance of the APF effectively.