The Research On Harmonic Prediction And Control Method Of Active Power Filter

With the development of modern industrial and power electronic equipment's extensive using, non-linear loads in electric power system largely increased. The grid system has being polluted, and power quality has become worse and worse. Using the APF to suppress the harmonic and compensate the reactive power is an effective way to resolving the harmonic contaminator in grid system. By comparison to conventional passive LC-filter, APF has the advantage of quick response, good output wave shape and dynamic performance.Firstly, the paper briefly presents the current situation and trends of active power filter. After introduction the basic structure and principle of APF, it makes comparison among some conventional harmonics detections and current control methods. Based on the idea mentioned above, the influence of time delay inherent in the digital control system on the compensation effect of APF is analyzed in detail.Secondly, to solve the drawbacks of time delay inherent in the digital control system on the compensation effect of APF, harmonic prediction algorithms based on adaptive predictive filter and ANN for active power filter were researched in the paper. In prediction algorithms based on adaptive finite impulse response (FIR) predictive filter, the least mean square algorithm is used to predict the signals to be detected, and compensated the harmonic currents according the tracing and predicting of the harmonic currents. This algorithm is simple and easy to come true with digital controller. As to the second method, by means of dynamic identification of APF system, training off line and modifying on line, the ANN model of system is acquired. Considering the robustness of predictive control, then, Artificial Neural Network(ANN)based predictive control for APF is achieved. In order to make the RBF network much simpler and tighter, an adaptive learning algorithm that adjusted the structure and parameters of the network dynamically is proposed.Finally, to validate the validity of the algorithm, it gives harmonic ratios' datum by means of math model and computer simulation. Simulation test with variable non-linear load is implemented. The results obtained show that, compared with the popular non-forecasting control method, proposed predictive control methods have preferable dynamic responses and controlling precision in both steady-state and transient operations, and the validity to alleviate the detrimental influence of time delay inherent in the digital control system on the compensation effect of APF.