Research On Active Power Filter Based On RBF Neural Network Sliding Mode Control

With the rapid development of power electronic technology,a large number of power equipment and facilities are electrified.The high power electronic equipment and facilities improve the performance of the equipment,but at the same time,it also aggravates the harmonic pollution of the power supply system.Active power filter has received widespread attention because it can track and compensate harmonics generated by nonlinear equipment in real time,and is the main technical means to resolve harmonic pollution.As two important parts of APF,harmonic current signal detection and compensation signal tracking control are always the focus of research.Firstly,this paper summarizes the research status and technical difficulties of APF,summarizes and analyzes the existing problems in APF harmonic detection and control strategy,and determines the research direction of the subject.In this paper,the parallel APF is taken as the research object.In the harmonic detection,this paper proposes an improved variable step size adaptive harmonic detection method.to solve the problem that the traditional fixed step size adaptive detection algorithm cannot take into account the steadystate accuracy and the convergence speed.The algorithm establishes a new nonlinear function and designs a new step size iterative formula to realize real-time dynamic adjustment of the algorithm step size.In the current tracking control,this paper proposes an RBF neural network sliding mode control strategy to solve the problems that the chattering and control accuracy in the application of sliding mode control in APF.By using the strong nonlinear approximation ability and self-learning ability of RBF neural network,the adaptive approximation of unknown functional in sliding mode control law and the real-time training optimization of high-frequency switching control coefficient are realized.The stability and convergence of the whole closed-loop system are guaranteed by adjusting the adaptive weight.On this basis,through further research,it is found that although the combination of RBF neural network weakens chattering,it does not really solve the chattering problem caused by the discontinuity of the control variable,and the control structure is relatively complex.Therefore,this paper proposes a second-order sliding mode control based on the supertwisting algorithm.The discretized high-frequency chattering output signal from the traditional sliding mode control is transferred to a higher-order one on the basis of maintaining the advantages of the traditional sliding mode control.As a result,the control quantity is continuous in time,thus effectively suppressing chattering.In order to verify the feasibility and effectiveness of the harmonic detection algorithm and control strategy proposed in this paper,the whole APF system simulation model is built in Matlab/Simulink simulation software,and the improved algorithm and control method are verified by simulation experiments under different working conditions.The simulation results show that the harmonic detection algorithm proposed in this paper reflects better detection performances and overcomes the shortcomings of traditional detection method.For the control part,both RBF neural network sliding mode control and super-twisting second-order sliding mode control can effectively suppress sliding mode chattering.However,in comparison,the latter is better than the former in terms of harmonic compensation accuracy and tracking control performance,and the effect of chattering suppression is better,which effectively improves the dynamic and static characteristics of APF.