Research And Application On Stochastic Distribution Control Theory

The random phenomenon is widespread and unavoidable in various types of industrial systems. Different from the previous stochastic control criterion which only emphasizes the mean and variance of the key variables, stochastic distribution control (SDC) algorithm takes higher order moments of the variable information into consideration, and the goal is to control the shape of output probability density function of the key output variables.In this paper, the stochastic distribution control algorithm is combined with neuron control algorithm, and based on the data driven and the entropy criterion in information theory knowledge, the adaptive stochastic control algorithm is investigated for univariate and multivariate stochastic control systems, then the convergence condition is analyzed, respectively. Compared with PID controller, the control algorithm is applied to the numerical and the industrial process, the results illustrate the proposed controller can reduce the system’s volatility effectively and make the system output maintain at the set point with less fluctuation. In addition, the limitations of the the minimum variance based performance evaluation method is considered, a reliable and general performance based on minimum entropy index is put forward for the first time, which can be used to evaluate feedback loop control performance with any unknown disturbance. The quadratic Reyi’s entropy of output under the minimum entropy controller is proposed as the performance evaluation benchmark. At the same time, a filter based on improved entropy criterion is designed to identify the estimated perturbation sequences by using the processing data of closed loop. The analysis in terms of numerical simulation and practical process data verifies the efficiency of the new control system performance evaluation method. Compared with the MVC evaluation methodology, the new method has better versatility andaccuracy.