Based On Parzen Window Estimate Dynamic Random Renyi-'s Entropy Of The System Output Distribution Control Study

Dynamic control of stochastic system has long been recognized as one of the important and fundamental subjects in control and practice. Traditional control techniques are mainly concerned with mean and variance of the system variables, which are mainly subjected to Gaussian process. In the recent years, the efforts have been made to the control of the probability density function (PDF). The shape control of the output PDF for general stochastic system is named as stochastic distribution control (SDC), which provides a general framework for the stochastic systems including Gaussian and non-Gaussian process.In this thesis, the establishment of index performance function, the achievement of control algorithm and applications have been studied, which include origin, background, significance, modeling, control design and system analysis. The main contributions can be summarized as follows:(1) The Parzen Window Estimate function has been put into the index performance function which is based of the Renyi's entropy, and a new index performance has been made. The partial derivative of index performance has been obtained using the Partial Differential Equations. According to the gradient amendment theory, the parameters of PID have been amended and the entropy is minimum, and a new control algorithm has been produced. For testing the feasibility of this new algorithm, the simulation have been proposed. The simulation have been divided into both cases:existing Gaussian disturb and non-Gaussian disturb. When there is little range Gaussian disturb, the control purpose can be achieved by the traditional PID controller. The parameters can be optimized by the control algorithm of this thesis, and the control effect is better. When there is big range Gaussian disturb, the control effect is worse by the traditional PID controller, evenly, can not be achieved. Also, the control purpose can be achieved by the algorithm of this thesis. When there is non-Gaussian disturb in system, the control purpose can also be achieved. For fully explaining the simulation effect, the output signal tracking given signal curve, the entropy curve, the parameter correction curve and the output curve have been made.(2) The control algorithm proposed in this thesis has been applied to furnace temperature control system of industrial furnace for the purpose of testing the practicality of this algorithm. The furnace temperature distribution is mainly researched. When there is non-Gaussian disturb in system, the furnace temperature distribution can be expressed by the PDF of non-Gaussian, and the given signal is tracked by the output signal. In the same way, by analyzing the output signal tracking given signal curve, entropy curve, the parameter correction curve and the output curve, we know, the control algorithm in this thesis is practicable.