| Due to the characteristics of China’s energy resource reserves,coal accounts for a large proportion of China’s energy structure.In the process of coal mining,the occurrence of gas accidents has always threatened the life safety of coal mine staff and caused economic losses to coal enterprises.Gas emission can reflect the occurrence of gas in coal seams and is an important indicator for designing mine ventilation systems.At the same time,the occurrence of gas explosions,coal and gas outbursts in coal enterprises is closely related to abnormal gas outbursts.Therefore,accurate and fast prediction of gas emission data is of great significance to coal mine workers,enterprises and governments.In actual coal mining,the amount of gas emission is the result of a combination of various complex factors,with complex nonlinear characteristics of timing and randomness.In this paper,based on a large amount of analysis of the research status of domestic and foreign scholars on the prediction of gas emission,combined with machine learning and intelligent algorithms,a gas emission prediction method based on Pearson correlation coefficient and long-short-term memory network is proposed.Firstly,the gas related data collected from the mine gas monitoring system were preprocessed.Since the selection of the influencing factors of gas emission is closely related to the final prediction results,this article is based on relevant research and theoretical analysis at home and abroad Starting from the comparison of several dimensionality reduction algorithms,the Pearson correlation coefficient analysis method is finally selected to reduce the characteristic attributes of the factors that affect gas emission,and to reduce the dimensionality of the data while ensuring the integrity of the characteristic data.From the original 13 kinds of data,9 key factors affecting the gas emission were selected.Then,by analyzing the non-linear characteristics of the timing of gas emission,a long-term and short-term memory network is selected.The network can deal with the characteristics of long-term dependence of information between time series data,automatically excavate the potential correlation between the data,and overcome the problem of gradient disappearance and gradient explosion of the recurrent neural network.The sample data is divided into a training set and a test set.The training set is used to train the LSTM model.Considering that the long short-term memory network has the problem of easily converging to the local optimal solution,the gray wolf optimization algorithm is used to optimize the model parameters.The test data is used to analyze the model.Finally,the forecast results are displayed on the system.The results show that the prediction accuracy of the gas emission prediction model based on the improved LSTM can reach 97%,which can effectively predict the gas emission content.. |
