Temperature Prediction for Electromechanical Equipment in Open-Pit Coal Mines Under Complex Working Conditions Using Wavelet Packet Decomposition and Graph Attention Network

The electromechanical equipment in open-pit coal mines is influenced by perturbing factors such as load, ambient temperature, and frequent startups and shutdowns, which result in low accuracy and poor generalization performance of the prediction model for its operating state. This paper considers both long-term temperature fluctuations and episodic changes caused by these perturbing factors. Additionally, multidimensional time and spatial data are integrated to propose a temperature prediction model for mining electromechanical equipment based on wavelet packet decomposition and a graph attention network (WPD-GAT). The temperature data of electromechanical equipment in an open pit coal mine in Xinjiang are used for experimental verification and comparison with four models. The experimental results show that the model proposed has an improvement in prediction performance, indicating the superiority of the model in the prediction of the temperature of electromechanical equipment of open-pit coal mines under complex working conditions.