Study On Optimal Control Of Active Power Filter Based On RBF Neural Network And LCL Output Filter

The rapid development of the modern power electronic technology, make the power electronic equipment in modern society become industrial and agricultural production and people daily necessities of life. At the same time all kinds of nonlinear load produce of harmonic and reactive current for the influence of the power network is more and more serious. So, the power system harmonic processing and reactive compensation has become all users must be thinking of problem, and active power filter can realize the harmonic control and reactive power compensation double function, so that in the very great degree to make up for the traditional L, LC, LCR power filter can only achieve a single function of the defect, become the research focus.This paper firstly introduces the working principle of active power filter and technology development present situation, the summary now hysteresis current tracking control of the advantages and disadvantages, in view of the current tracking control of the shortage, proposes the use of neural network to solve these problems. In this paper, the principle of the artificial neural network learning methods and training algorithm is analyzed in detail. According to the traditional hysteresis control based on the principle of, through the training a radiate basis function (RBF) neural network which traditional hysteresis controller is the compensation current control and dynamic tracking function, make output control frequency controllable.The active power filter’ output filter greatly influences the performance of active power filter, the active power filter’ output filter can be roughly divided into LC type, type and LCL,LCR type. Due to solve the type filter LCL,LC type, LCR type filter vulnerable to power grid impedance and the uncertainty of the changeable influence to filter the effect not beautiful shortcomings, become the active power filter output filter the best choice. In the past but LCL power filter parameters design mainly by experienced engineers with experience set, like this whole DingZe will attend, the effect is not good. So this paper adopts multi-objective ant colony algorithm was introduced to optimize LCL filter, obtain good effect.In this paper the proposed based on radiate basis function (RBF) neural network hysteresis current tracking controller and a multiobjective optimization design of ant colony algorithm of LCL filter, are established in the MATLAB software environment simulation model, through the simulation resuls prove the correctness and feasibilty.