Research On Prediction Method Of Gas Emission In Working Face

The mining face of the coal mine is one of the main working faces of the gas accident.Accurately predicting the amount of gas emission from the mining face can greatly help reduce casualties and property losses.The current forecast of gas emission is mainly through two methods.First,it is a static prediction based on the geological conditions of coal mining and the influence of artificial construction to predict the amount of gas emission.The second is to use the coal mine gas emission detection system to collect a certain amount of gas emission time series,to deeply explore and analyze the timing law,and to use historical data to predict futxire data,which is a dynamic prediction.The static prediction method generally uses less sample data,and the system is complicated and difficult to predict due to the interaction of various factors.Therefore,SVR with good processing ability for small sample data is firstly used to search for SVR parameters by genetic algorithl.Global optimization,and then using the NM simplex algorithm to achieve fast convergence and precise optimization,find higher performance parameters based on genetic algorithm optimization,and ensure that the algorithm is stable and will not fall into local optimal values.The GA-SM-SVR model was established to predict the gas emission.Finally,the data of a coal mine is selected for modeling and analysis.The experimental results show that the proposed model has fast convergence and accurate prediction.The dynamic prediction method is based on the gas errussion quantity time series.According to the chaos of the gas emission quantity,the paper firstly uses CC to find the best embedding dimension and delay time of the gas emission quantity sequence for phase space reconstruction.The Takens theorem is used for prediction.The high-order sparse Volterra function is used for modeling analysis,and the concept of sliding window is introduced.The model is trained in the prediction process to ensure the validity of the model parameters.In order to accelerate the training speed of the model,the CG and RLS algorithms are coupled to increase the operation speed,so as to establish the CG-RLS-HONFIR model.Finally,the gas emission amount of a coal mine is selected as a sample for experimental analysis.The experimental results are presented in this paper.The timeliness of the model is high,and the prediction results are improved by about 5%compared with the prediction results of the SVR and BP neural networks.Figure 22 table 9 reference 55...