Deep prediction model of coal spontaneous combustion temperature based on multi-feature fusion

In order to effectively prevent coal spontaneous combustion disasters in goaf of coal mines and improve the accuracy of coal spontaneous combustion disaster prediction models, a deep prediction model of coal spontaneous combustion temperature based on multi-feature fusion is proposed. Firstly, each feature of coal spontaneous combustion data is denoised through an autoencoder network to enhance the robustness of the data. Secondly, the denoised data is transformed into a two-dimensional feature matrix in chronological order. A sliding window is used to slice the feature matrix, and a convolutional neural network in deep learning is used to extract effective features from the feature matrix, followed by feature data fusion. In the process of denoising, feature extraction and fusion, differential evolution algorithm is used to optimize the parameters of the denoising autoencoder and convolutional neural network. Finally, the fused data is inputed into the gated recurrent unit neural network for predicting coal spontaneous combustion temperature. The experimental results show that the mean absolute error of the data after denoising and feature fusion is reduced by 6.55% and 69.26% respectively compared to directly using the original data, and the root mean square error is reduced by 13.23% and 63.49% respectively. This indicates that the accuracy of coal spontaneous combustion temperature prediction can be effectively improved after encoder denoising and feature fusion processing.