Research Of Predictive Control On The Temperature Of Coke Oven Based On Neural Network

Coke is widely used in metallurgy, machinery and chemical industry as the main raw material and fuel, and our country is the biggest coke producer and exporter in the world. The stability of Coke oven temperature is very important for reducing energy consumption, advancing output and quality, prolonging the life-span of coke oven body, and etc. However, just as most of the plants systems in the industry field, coke oven heating combustion process is nonlinear and simple control method will not perform perfectly. Orthogonal neural network can overcome the defects of feedforward neural networks such as local minimum, slow convergence speed, determination of initial weights and the number of hidden layer, etc. Thus orthogonal neural network is a powerful tool for nonlinear system identification. Generalized predictive control is one part of predictive control which has been widely used in industrial process control areas for its nice control performance. From the above, predictive control on the temperature of coke oven based on neural network has been studied in this paper.Firstly, the technological process of principle and process for coke-oven is introduced in this paper. Flue temperature of coke oven is chosen as the control objective, and gas flow is chosen as the manipulated variable to control the flue temperature.Secondly, this paper emphatically discussing the orthogonal polynomials neural network and how to implement in Matlab based on the theory Legendre polynomials. Then coke oven flue temperature model is established by using field data, and the Matlab simulation results verify the effectiveness of this method.Finally, on the based of expounding basic principle of predictive control, step predictive control upon orthogonal Neural Network was given for coke oven heating control system. In the process of system operation, the model is linearized at each sampling instant, and the step control strategy algorithm is employed to implement the controlled plant. The simulation results show the effectiveness of the presented algorithm and this paper also analyze the main parameters in the algorithm. At last, simulation was conducted according to different actual conditions, and the results prove the validity of the algorithm.