基于Keras长短时记忆网络的矿井瓦斯浓度预测研究

瓦斯浓度监测是煤矿瓦斯灾害事故预警的重要的手段,其浓度变化预测对于提升矿山安全生产具有重要意义。针对矿井瓦斯浓度预测问题,建立了一种基于Keras长短时记忆网络的矿井瓦斯浓度预测模型。该模型首先对矿井瓦斯浓度时间序列进行标准化处理,并将处理后的时间序列划分为训练集与测试集;然后通过调用测试集数据进行模型训练,利用提出的基于LSTM网络建立的矿井瓦斯浓度多步预测模型,实现了对矿井瓦斯浓度发展趋势的预测,并利用损失函数计算预测误差大小,评估模型的预测精度;最后以贵州某煤矿掘进工作面为工程背景,利用基于LSTM网络建立的矿井瓦斯浓度多步预测模型,开展了矿井瓦斯浓度预测研究,并通过与ARMA模型、ARIMA模型的预测结果进行对比,验证该模型的预测效果。结果表明:该模型预测结果的均方根误差RMSE值最小仅为2%,且预测步长约为ARMA模型、ARIMA模型的5倍,说明该模型的预测效果好,可为煤矿井下合理规避瓦斯灾害事故提供科学依据。

Prediction of Mine Gas Concentration in Heading Face Based on Keras Long Short Time Memory Network

Gas concentration monitoring is an important means of early coal mine gas disaster warning,and its concentration prediction is of great significance to improve the safe production of mines.This paper establishes a gas concentration prediction model based on Keras long-term and short-term memory network to solve the problem of gas concentration prediction.The model standardizes the time series of gas concentration,and divides the processed series into training set and test set.The study calls the test set data for model training,and uses the proposed multi-step prediction model to predict the development trend of gas concentration.The study uses the loss function to calculate and predict error,and evaluates the predictive accuracy of the model.Taking a mine in Guizhou as the engineering background,the paper establishes the prediction model of mine gas concentration based on LSTM network model,and compares the prediction results of gas concentration in mine heading face with ARMA and ARIMA models to verify the prediction effect of the model.The results show that the minimum RMSE value of this model is only 2%,and the prediction step is about 5 times of ARMA and ARIMA models,which can provide scientific basis for reasonably avoiding gas accidents in coal mines.