Research And Application On Stochastic Distribution Control Theory In Non-Gaussian Systems

Random disturbance is widespread and unavoidable in industrial systems.The previous design methods only describe the characteristics of system by means of the mean value and variance,and assume that the disturbance obeys Gaussian distribution.But in the actual industrial process,the disturbance doesn't obey Gaussian distribution necessarily,the original design methods are not applicable to the nonlinear non-Gaussian systems.To solve this problem,stochastic distribution control(SDC)algorithm provides a new idea which takes higher order moments of the variable information into consideration.In this paper,subsystem models of wind energy conversion system(WECS)are set up,and the whole model of WECS could be built by connecting these subsystems.The WECS operates in different regimes according to wind speed,and then the whole models of WECS are built under low wind speed and under high wind speed respectively.Meanwhile,a combined wind speed model is given which is convenient for simulation analysis.Secondly,the WECS model under low wind speed may be affected by the randomness of wind speed,in order to achieve maximum wind energy capture under low wind speed,a SIP-based controller is designed within stochastic framework to control the rotating speed.And some results about PID controller are given for comparison.Because the PID control effect is not good,this paper designed a PID self-tuning control algorithm based on RBF neural network,considering the influence of non-Gaussian noise,the mean value and variance can't describe the characteristics of systems.So SIP is used as the performance index in identification and controller design.The simulation results show that the algorithm can effectively improve the control effect of PID.Finally,this paper uses correntropy to describe the uncertainty of the non-Gaussian systems.A filter is designed based on the maximum correntropy criterion,and then sensor fault diagnosis method is proposed based on a bank of maximum correntropy filters,i.e.sensor malfunction is determined by residual which is produced by the proposed filter,then the probabilistic neural network is used to locate faults.This method is applied to WECS model under high wind speed affected by random wind speed and random measurement noises,simulation results show that this method can effectively diagnose sensor faults.