Neural Network Predictive Control In The Application Of Reheating Furnace Furnace-temperature Control And Optimization

Reheating furnace is one of the most important facilities in steel rolling, and consumes great amount of energy. So it is essential to improve heating efficiency and reduce energy consumption for steel industry. Modern rolling mills are developing to be continuous, large-scale, fast produce, high quality, which deliver a higher demand for the research of reheating furnace modeling and optimal control. However, reheating furnace is a typical complicated industry system, complex thermodynamic, chemical and physical change makes it a system with multi-variable, time variety, non-linearity, strong coupling, pure time delay and a lot of disturbance factors. So the optimal control of the reheating furnace is complex problem of control and optimum, traditional control strategy is difficult to achieve good control effect.According to the product practice of the reheating furnace and complexity of modern industry, the thesis elaborates the application and research situation of optimal control of the reheating furnace and points out the problems in the reheating furnace’s optimal control.With regards to the characteristics of nonlinear, large inertia and large time-delay system, this thesis puts forward a predictive control strategy based on the neural network to control the temperature of the reheating furnace based on a lot of related literature, combining with product practice.First of all, the reheating furnace furnace-temperature predictive model should be established. Because the reheating furnace is a complicated nonlinear system which is hard to establish an exact mathematic model, the thesis sets up predictive model through BP neural network based on the data from field to predict output value; Feedback correcting is used to reduce the model predictive error resulted from uncertain factors of the system in order to get the relatively precise predictive value. Based on these, furnace-temperature optimal controller is established according to the quadratic cost function to adopted moving optimization, so that the future control sequence is got. MATLAB simulation results indicate that the control system has better following performance for temperature varying, less adjusting time, greater anti-interference ability and greater robustness, which lays the foundation for the application in production.