Development of Electricity Demand Forecasting Technology for Hot Rolling Lines Using Machine Learning

Reducing energy consumption and emission is a key environmental issue. Although industry requires high-quality steel sheets for forming, the production of these sheets by the steel industry consumes enormous amounts of electric energy and emits significant amounts of CO2. In addition to reducing energy costs, there is a need to control CO2 emission, improve production efficiency, stabilize operations and solve several difficult problems simultaneously. Traditionally, the hot rolling process has been empirically determined to realize the desired mechanical properties, with the quality of the hot-rolled product taking precedence over energy consumption. To solve these problems, it is important to have a predictive technology to determine the amount of energy required in the future and its timing. Therefore, in this study, we compared the power demand prediction system with conventional statistical methods and developed a power demand prediction system that uses deep learning technology to predict the power demand of a hot rolling mill. In this paper, we will outline the development of a visualization and forecasting function for power consumption in hot rolling mills, and introduce our method of constructing a simulation technology to optimize power consumption.