An Energy Consumption Prediction LSTM Model of Metallurgy Enterprises

Abstract

Aiming at the characteristics of multi-dimensional production data, complicated sources and diverse data structures of metallurgy enterprises, it is of great significance to study how to use energy management-related data to predict metallurgy enterprises' energy consumption. Using with an accurate measurement method, the enterprises can not only reduce the cost of energy consumption but also develop economic efficiency in producing for metallurgy enterprises, which results in low energy use efficiency, high enterprise cost, and weak scalability. In this paper, we establish an LSTM model to achieve energy consumption prediction for metallurgy enterprises by optimizing the model's parameters using a grid search algorithm. We also compare our model's prediction results with other mainstream machine learning algorithms, i.e., MARS and SVM through indexes such as MSE, RMSE, MAPE, and MRAE to evaluate the prediction effect of the learning algorithm. According to our simulation, LSTM performs best in the task of energy consumption prediction.