Research On Prediction Model Of BOF Steelmaking Endpoint Based On Extreme Learning Machine

BOF is the most mainstream method of steelmaking in the world. The main task of it is processing temperature and component up to the requirements of high quality steel which the key technology of endpoint control. Raising the level of endpoint of BOF steelmaking can reduce costs and improve efficiency. Limited by a lot of reasons, a large number of small steel mills with lower automatic control level still rely on artificial experience and static control to direct the production in our country. In order to increase the control accuracy of the converter endpoint, combine neural network model with static control to predict the endpoint temperature and carbon content.Introducing the present research situation and achievements of converter endpoint control at home and abroad based on reading many foreign references and domestic. Analyze mechanism model of BOF steelmaking, according to the objectives and requirements and composition of raw material, using heat balance and material balance to calculate the ratio of raw materials. The core of static control is building a predictive model which based on historical data. Pretreatment original data including deleting inordinate values, normalizing and using grey relation analysis method to analysis the grey correlation between relevant parameters and endpoint temperature and endpoint carbon content. Taking parameters which with big gray connection as the input of the model. BP forecasting model and ELM forecasting model were built aiming at the defects of each model; PSO and DE were used to optimize the forecasting models for improving the generalization ability and the prediction accuracy.Take 60 sets of steelmaking production data as samples to check the simulation results of several models. Compare the prediction accuracy of models before and after optimization. Summarize the results, analysis the performance parameters and the prediction results of the models, and finally drawn the conclusion.