A novel combined model based on echo state network optimized by whale optimization algorithm for blast furnace gas prediction

Abstract

The accurate prediction of blast furnace gas (BFG) holder level is crucial for reasonable gas dispatch. Therefore, we propose an improved echo-state network (ESN) model, based on variational modal decomposition (VMD) and whale optimization algorithm (WOA), for the BFG system. This model extracts inherent characteristic components of the sequence using VMD for prediction and determines model inputs via partial autocorrelation functions (PACF). ESN effectively handles the shortcomings of other classical neural networks with many training parameters and slow calculation speed. The important parameters are optimized through WOA to improve the prediction accuracy of the proposed model. Based on our proposed model, the average MAPE of three datasets is 0.1805% for 1-Step prediction, 0.1829% for 3-Step prediction, and 0.1886% for PACF prediction. With high prediction accuracy for BFG holder level, our model takes less time to establish and calculate, meeting requirements for short-term level prediction and more reasonable BFG scheduling decisions. The experiments and discussions support the correctness, effectiveness, and superiority of our model.