Research On Prediction Of Gas Emission Based On Improved PCA-MEA-BP Neural Network

Accurate prediction of gas outflow is a key part of the gas extraction design and gas disaster prevention and control research on coal mining surface.In order to ensure the gas pre-pumping effect,and to reduce the harm caused by gas in mining,it has important practical significance and research value to study the gas gushing volume prediction on the coal mining surface.(1)The paper comprehensively analyzes the influence of geological conditions,coal seam characteristics and ventilation,mining factors and climate factors on gas outpouring volume,sifts out the gas outpouring influence factor of the working surface of the mine,constructs the gas surge prediction index system composed of 12 influencing factors,and analyzes the correlation between the forecast index and the gas out volume.(2)The main component analysis method(PCA)is used to reduce the dimensions of the established gas surge prediction index system.On this basis,the analysis of the main component sparing method is improved by spearman correlation coefficient analysis,and the analysis of the main component analysis method before and after the analysis of the main component is compared with the analysis,the results show that the traditional main component analysis method will reduce the original 12 variables to 7 variables,and the improvement of the main component analysis method can reduce the original indicator to 3 variables,to further simplify the data,reduce the error caused by redundant information in the original data,and thus improve the accuracy of prediction.(3)Selecting BP neural network algorithm as the basic prediction model,the value and threshold of BP neural network are optimized by increasing the thinking evolution algorithm(MEA),the accuracy of gas out pouring prediction result and the learning efficiency of the algorithm are improved,and the MEA-BP neural network prediction model is established.The error results show that the average absolute error is 0.0486,the average relative error is 3.58%,and the decision coefficient R2 is 0.939.(4)In order to verify the improvement of the quality of PCA-MEA-BP neural network prediction model proposed in this paper,the model is compared with the traditional BP neural network model,the improved PCA-BP neural network model and the MEA-BP neural network model,and the prediction results are analyzed by error.The results show that the improvement of PCA-MEA-BP neural network prediction model is reduced by 6.36%,3.06%,2.66%,and has strong practical value compared with the prediction error accuracy of the three comparative models.