为了高效地对工作面矿压变化趋势进行有效预测，减少顶板突发事故和引导煤矿井下安全生产，提出了融合相邻多个支架影响的矿压数据预测模型Prophet+LSTM。该模型首先基于数字信号处理技术，将原始矿压数据经过算术平均值滤波和小波去噪方法减少噪声和随机误差；其次在深入分析Prophet模型和LSTM模型特性基础上，通过添加额外回归变量方法融合相邻多支架矿压数据；最后为充分利用Prophet模型和LSTM模型的优势，构建了一种Prophet+LSTM组合模型对工作面矿压进行预测，通过线性加权组合方法获取最有权重系数使模型预测结果误差逐步减少。以均方根误差(RMSE)和平均绝对误差(MAE)来评估Prophet+LSTM模型对矿压时间序列的预测效果。预测模型在某矿工作面支架矿压时序数据的应用结果表明：经过数字信号处理后的支架矿压预测结果RMSE和MAE分别下降了约20%和16%，Prophet模型和LSTM模型的预测结果优于传统的BP神经网络和ARIMA模型，且Prophet+LSTM模型预测方法较其单项模型具有更强的稳定性和更高的准确性，实现了对工作面在推进过程中矿压变化的有效预测。该预测方法对煤矿井下工作面矿压预测提供了研究思路。

This paper tries to effectively predict the change trend of mining pressure in the working face, thereby reducing roof accidents and guiding coal mine safety production, a mining pressure data prediction model Prophet+LSTM is proposed that integrates the influence of multiple adjacent supports. The model is first based on digital signal processing technology, the original mine pressure data is filtered through arithmetic mean filtering and wavelet denoising methods to reduce noise and random errors; secondly, based on the in-depth analysis of the characteristics of the Prophet model and the LSTM model, the method of adding additional regression variables is used to combine the rock pressure data affected by multiple adjacent supports; finally, in order to make full use of the advantages of the Prophet model and the LSTM model, a Prophet+LSTM combined model is constructed to predict the rock pressure of the working face, and the most weighted weight is obtained through the linear weighted combination method. The coefficients gradually reduce the error of the model's prediction results. The root mean square error(RMSE) and mean absolute error(MAE) are used to evaluate the prediction effect of the Prophet+LSTM model on the mine pressure time series. The results of the application of the prediction model on the time series data of the support rock pressure in a certain mine show that the RMSE and MAE values of the support rock pressure prediction results after digital signal processing have decreased by about 20% and 16%, respectively. The prediction results of the Prophet model and the LSTM model.It is better than the traditional BP neural network and ARIMA model, and the prediction method of the Prophet+LSTM model has stronger stability and higher accuracy than its single model, which can effectively predict the changes in the mine pressure during the advancing process of the working face. This prediction method provides a research idea for the prediction of underground pressure in coal mines. 

0 引言

1 煤矿工作面矿压预测模型

2 工作面矿压时间序列数据处理

3 基于Prophet+LSTM的矿压预测模型

   3.1 融合相邻多个支架影响因素的Prophet模型

4 试验结果分析

   4.1 数据集介绍

   4.2 评估指标

   4.3 融合滤波技术和相邻多个支架的影响因素的预测结果分析

   4.4 不同模型预测性能试验对比

5 结论