Fuzzy-Neural Network Predictive Control Of Calcination Temperature In Rotary Kiln

As one of the main equipment in the calcination of active lime, rotary kiln has been wildly used in our country. With the quick development of the colored metallurgy and architectural industry, there is an increasing demand for rotary kiln. Calcination temperature in the rotary kiln is the key factor that determines whether the clinker can be produced with high quality, high production and low consumption. It is difficult to achieve desired effects by using traditional control methods, so it is of great significance to do the research on the control of calcination temperature in the rotary kiln. Predictive control is a kind of practical control algorithm which developed from the production practice. It can eliminate the effects caused by modeling error and the uncertainty in the structure, parameters and the environment of systems, so it has good control effects. Fuzzy-neural network can be used as an universal approximator, which can approximate any nonlinear function. Further more, it has strong learning ability. Thus, it has been applied in many fields. From the above, this paper researches on the fuzzy-neural network predictive control of calcination temperature in rotary kiln.Firstly, the technological process of lime calcining using rotary kiln is introduced in this paper. Calcination temperature is chosen as the control objective, and gas flow is chosen as the manipulated variable to control the calcination temperature.Secondly, according to the high inertial, large time delay, multi-disturbance, and time-varying characteristics of rotary kiln, fuzzy-neural network is adopted to construct the one-step prediction model of calcination temperature in rotary kiln. A kind of fuzzy-neural network is used to build the model, and it is trained by a hybrid learning method. And a exhaustive method is used to identify the time delay of the system.Finally, based on the expounding of the basic principle of predictive control, the method of predictive control based on intelligent prediction model is analyzed. Then a recursive fuzzy-neural network prediction model is derived from the fuzzy-neural network one-step prediction model of rotary kiln. And then a kind of reference control vector rolling optimization strategy is adopted to control the calcination temperature in rotary kiln. In this method which is relatively accurate, the optimal control quantity can be obtained by solving a series of QP problems. Further, a rapid one-step experiential boundary reference control vector rolling optimization strategy is derived to reduce the computational burden. Then the hybrid learning method is used for the on-line training of the fuzzy-neural network one-step prediction model as the part of feedback compensation. And by using this method both the output disturbance and model disturbance can be eliminated. Finally, the simulation results show that, in both situations, with and without constraint, the presented fuzzy-neural network predictive control has better performance than traditional control.