Deep Reinforcement Learning for Secondary Energy Scheduling in Steel Industry

Abstract

Considering that the blast furnace gas(BFG) tank level scheduling is of great significance for the steel plant's secondary energy system balance, this paper proposed a scheduling model based on deep reinforcement learning. In this model, BFG gas tank scheduling was transformed into searching the best production state under a certain operating condition, and a deep Q-learning network was used to search this state. Moreover, in order to speed up convergence and improve algorithm stability, an experience based pre-training was added to the training session. In order to verify the effectiveness of the proposed method, experiments are carried out with the secondary energy system production data of a domestic steel enterprise. The results show that the proposed method is more effective than artificial scheduling.