Study On Multivariable Intelligent Control And Its Application To Ball Mill Coal Pulverized System

Ball mill coal pulverized system is a multivariable coupled nonlinear time-varying system which is extensively applied in coal fired power plants and is difficult to be controlled by conventional PID regulating algorithm. No good control method for BMPS has been gotten by now. Energy consumption in this system is very enormous and it's important for power plant so that it is necessary to realize the automation and optimal operation for BMPS. In view of its multivariable and strong coupling, This paper presents a DMC and GPC decoupling design, they decompose a multi-input multi-output system into multi-input single output systems, then use DMC or GPC algorithm to design controller. By means of solving matrix equation group, predictive decoupling controller is realized, in order to get good control performance, multiple model based MPC is used and fuzzy self-tuning technology is proposed for designing parameters of MPC Future tracking errors weighting is introduced in generalized predictive control, based on further research of weighting factors, general stability results are obtained. Dynamic mathematical model of ball mill pulverizing system with storage bunker is obtained by mechanism analysis. In view of the operation modes(i.e. normal mode and fault mode) of pulverizing system, an inverse control strategy based on distributed NN(neural network) is put forward, and PID regulator is introduced in pseudo linear system including NN inverse controller. The new method can overcoming nonlinear and strong coupling features of the system in a wide range. Coal load in ball tube of pulverizing system is affected by many factors, this paper analyzes these factors and their relations. A method of load measuring based on NN(neural network) is proposed, a varying structure NN measuring scheme is presented. Computer simulation results demonstrate the efficiency of the strategy proposed, which is helpful to the automatic control and optimal operation of the ball mill pulverizing system. Neural networks are used for modeling a nonlinear system, and two kinds of scheme are proposed for designing of neural network predictive controller The method overcomes the large amount of computation load in traditional nonlinear system predictive control, it utilizes nonlinear optimization technology to get optimal predictive control output witch is used to train neural network predictive controller in order to win a well trained predictive controller. The new method has only a little on-line computation load and is wieldy when application in industry processes. Conventional method which use electric energy consumed per ton coal for assessing coal pulverizer system has some shortcomings, so we put forward a new objective effective methoduzzy Comprehensive Assessment Method. Ph.D Candidate: Wang Dong-feng Directed by: Prof Song Zhi-ping and Prof Li Zun-ji...