Research On Coke Oven Flue Temperature Control Based On Intelligent Predictive Control Algorithms

As an important fuel and raw material for metallurgical and other industries,coke is very important in the industrial background.In the coke oven vertical fire way,it is of great significance to ensure the stability of the flue temperature,which has a direct impact on reducing the energy consumption of coke oven production,improving the quality of coke,prolonging the service life of the furnace,and even protecting the environment.Coke oven heating process has strong nonlinearity and large hysteresis.It is difficult to achieve satisfactory control effects simply relying on traditional control.Neural networks can be well fitted and identified for nonlinear systems.Generalized predictive control has outstanding control ability,good anti-interference and robustness.On this basis,this paper studies the predictive control of coke oven flue temperature based on neural network.Firstly,based on the mass production data collected by the 6# coke oven production site of Baotou Steel,the data is preprocessed to obtain good data,and the key parameters affecting the flue temperature are confirmed by correlation analysis,and the relevant data is used for predictive modeling of the flue temperature.Through expert experience and the analysis of the main factors affecting the flue temperature,it is determined that the gas heat value and gas flow fluctuation are the main influencing factors,and the flue temperature is controlled by the adjustment of the gas flow.Historical data analysis of gas flow and flue temperature shows that the flue temperature fluctuations are large and the gas consumption is unstable.Therefore,it is necessary to control it and determine the overall research plan.Secondly,in view of the current situation that it is difficult to measure the flue temperature on-line in real time,using the neural network to fit the nonlinear system well,based on the excellent data collected and processed in the field,the prediction model of flue temperature is established by shallow BP neural network.Simulation results show that BP neural network can better predict the flue temperature.However,in the process of data calculation,there are hidden dangers such as data over-fitting.The Long Short-Term Memory Network(LSTM)can avoid data calculation problems in shallow neural networksand has better processing ability for time series data.Therefore,under the tensorflow framework,the LSTM program is written in Python language,and the prediction model of coke oven flue temperature is carried out.Simulation studies show that the prediction error of LSTM is further reduced,and the prediction of the flue temperature is more effective.Then,through in-depth analysis of predictive control,BP neural network and predictive control are combined to predict the control of the flue temperature.Firstly,the BP neural network is used to establish the prediction model of coke oven flue temperature.Then,the nonlinear model of neural network is linearized.By imposing the ladder control constraints on the future control variables,the work of solving the inverse matrix is improved,the corresponding computation is reduced,and the robustness and anti-disturbance ability of the algorithm are increased.The generalized predictive control algorithm and its improved algorithm are applied to study the control of flue temperature,and the important parameters of the algorithm are studied respectively.The simulation shows that the application of the stepwise control strategy has improved the response speed and parameter adjustment of the system.Finally,based on the actual production situation of the Baotou Steel Coke Oven,the simulation research under various working conditions was carried out.Intelligent predictive control algorithm combines neural network,step control strategy and generalized predictive control effectively.The simulation results show that the improved stepped control algorithm can effectively improve control performance and improve the dynamic response of the system.At the same time,the control action changes in the same direction,reducing the fluctuation and effectively protecting the service life of the field actuator.It shows that the intelligent predictive control algorithm has better control effect on the flue temperature of coke oven.