Prediction Method of Mixed Gas Consumption in Hot Rolling Based on Improved K-Medoids Working Condition Division

The prediction of mixed gas consumption in hot rolling production is of great significance in gas production-consumption balance and energy saving. However, in the actual production process of hot rolling, the complexity of production environment and the strong coupling of production processes lead to the complexity of working conditions, which brings great challenges to the prediction of mixed gas consumption. A prediction method of hot rolling mixed gas consumption based on improved K-Medoids working condition division is proposed in this paper. Based on Chebyshev's theorem, the method uses more robust median and median absolute deviation to evaluate the outliers in the cluster at each iteration in terms of distance, and further verifies from the perspective of density to complete the division of normal working conditions and abnormal working conditions. On this basis, considering the time dependence of hot rolling data, the prediction model of mixed gas consumption under corresponding working conditions was established by using the Long Short-Term Memory (LSTM) neural network. In order to verify the effectiveness of the proposed method, the actual industrial field data are used. The simulation results show that the proposed method has high prediction accuracy and is suitable for the prediction of mixed gas consumption in hot rolling.

I. Introduction

The hot rolling process is one of the most important production processes in modern iron and steel enterprises. The heating furnace is the main equipment in the hot rolling process and the main energy consumed is the by-product energy generated in the production process of iron and steel enterprises, mainly including blast furnace gas (BFG), Lindz-Donawitz Gas (LDG), coke oven gas (COG) [1]. Due to the large consumption of mixed gas and low utilization efficiency in actual production, there are various working conditions in the production line [2]–[3], resulting in waste of gas. Therefore, it is of great significance to accurately predict the consumption of mixed gas in the hot rolling process, which can improve the utilization rate of by-product gas and reduce costs and increase efficiency for iron and steel enterprises [4]–[5].