Energy Consumption Prediction of Steelmaking Process Based on Improved Whale Optimization Algorithm and Stochastic Configuration Network

Abstract

To solve the problems of energy consumption, a prediction model based on an improved whale swarm optimization algorithm and stochastic configuration network (LWOA-SCN) is proposed. The advantage of the model is that the hidden layer structure can be generated adaptively according to the training effect and is characterized by strong generalization performance, fast convergence, high prediction accuracy, and the ability to jump out of local optimality. First, we used principal component analysis, mechanistic model derivation, and SPSS correlation analysis to determine the main influencing factors of the three sub-models. Second, LWOA was used in combination with a stochastic configuration network to establish the oxygen consumption model, gas recovery model, and steam recovery model. To verify the prediction effect of the model, it was compared and analyzed with BP (back propagation), RBF (radial basis function neural network), and SVM (support vector machines) prediction models. Finally, the LWOA-SCN model was tested in industrial production; the results showed that the hit rates (within ± 5 m³/ton) of models for oxygen consumption, gas recovery, and steam recovery were 91.08%, 92.14%, and 92.27%, respectively, and simultaneous hit rate was 85.71%. Those verify that the models can meet the actual demand of energy consumption prediction in a steel mill.