A Reinforcement Learning-based Multistep Prediction Strategy for Burn-Through Point Using State Feature Extractor
Junjin Mu,Chunjie Yang,Feng Yan,Chong Yang,Tao Lin
DOI: https://doi.org/10.1109/tim.2024.3502771
IF: 5.6
2024-01-01
IEEE Transactions on Instrumentation and Measurement
Abstract:Sintering represents a critical link in the steel production process, with the produced sinter ore serving as the core raw material for blast furnace ironmaking. Characterized by harsh working environment and multi-parameter, the sintering process causes many obstacles to academic research on measurement and prediction of key indicators. To address this problem, this article introduces a multistep prediction strategy, named the state feature extractor-based multi target networks delay updates algorithm (SFE-MDA), to predict burn-through point in sintering process. Initially, considering the characteristics of measurable interrelated variables in the sintering process, initial variables are selected and categorized. Subsequently, following the input standards of reinforcement learning framework, the state feature extractor is employed to derive latent variables and to model the Markov Decision Process. Finally, a novel multi target networks delay updates algorithm is designed. The proposed approach is validated on real sintering data, with results indicating its superiority over existing deep learning methods in terms of long-term multistep prediction accuracy, thereby confirming its feasibility and superiority.