Hybrid prediction model of coal stack temperature in power plant considering temperature fluctuation process

In thermal power plants, coal spontaneous combustion seriously affects the stability of the fuel delivery system. Therefore, preventing coal yard spontaneous combustion is very important for the safe operation of thermal power plants. In this paper, a hybrid prediction method for coal pile temperature prediction is proposed. First, the coal pile temperature is divided into different fluctuation processes. Meanwhile, the attention mechanism (AM) is used to excavate the factors that affect the coal pile temperature with different temperature fluctuations. Second, convolutional neural networks (CNNs) are employed to further extract spatiotemporal features. Third, a long short-term memory neural network (LSTM) is employed to predict the coal pile temperature. Finally, different prediction models on the experimental platform are built. The results show that the proposed prediction model has excellent prediction accuracy with different steps.