ELM based LF temperature prediction model and its online sequential learning

Abstract

The accurate prediction of molten steel temperature is important for optimal control of Ladle furnace (LF) process. Under this conception, a novel LF temperature prediction model is constructed based on extreme learning machine (ELM), which is a new learning algorithm for single hidden layer feedforward neural networks (SLFNs). ELM is chose due to its good generalization performance and extremely fast learning speed. Furthermore, online sequential learning is adopted on the sequentially arriving data to correct the ELM based prediction model. We introduce a forgetting factor in this learning scheme for the sake of successfully accommodate to the variation in the production process. The simulation results show that the proposed predictor has a good accuracy and fast sequential learning speed, which ensure its ability for practical application.