Research On Prediction Of Gas Emission In Coal Mine Based On Improved PCA And SVR

The gas disaster is all the time one of the main disasters in coal mines,it will not only bring huge economic loss to the enterprise,but also bring threat to the safety of the coal mine workers,so gas prevention and control is very important.The amount of gas emission refers to the amount of gas released in the process of construction and production in coal mines,which is an important parameter to reflect the occurrence of gas in the coal seam.The gas emission is the main cause of gas explosion,gas poisoning,coal and gas outburst during mining operation in coal seam,it is also one of the main indicators to determine the mine ventilation,it is an important part of gas prevention and control.The gas emission can reflect the condition of gas comprehensively in a coal mine,whether the prediction results are correct,will not only directly affect the various technical and economic indexes of the coal mine,but also related to the safety of coal mine workers.Therefore,it is crucial to develop an appropriate and effective method to predict gas emission for guiding the safety production.In the actual production,the mine gas emission is affected by many complex factors,in the study of the classification or regression prediction ofsamples with such multidimensional data features,it may encounter problems such as the learning model is too complex,the training time is too long,the final result is inaccurate and so on.Therefore,it is necessary to deal with the raw data,reduce the dimension and remove the noise and redundancy.At the same time,as the support vector machine(SVM)model has many good properties,it puts glorious greatly in the field of machine learning,especially its advantage in small sample processing and better generalization performance characteristics,it is very suitable for prediction of gas emission.In this paper,it proposes a new gas emission prediction method which based on improved PCA algorithm and SVR algorithm,finally it achieved good results through experiments,the effectiveness and feasibility of the method are proved.This paper mainly involves the following work.(1)This paper summarized the research status of gas emission prediction and various prediction methods,then introduced the mechanism and influencing factors of coal mine gas emission.(2)It analyzed the main influencing factors of gas emission,determine the selecting principle of characteristic indexes for the prediction,according to the principle,selectd the appropriate index to forecast the gas emission.(3)It introduced the basic principle of the principal component analysis and steps of the algorithm,and aiming at the specific problems in the data processing,the weighted improvement was made,which makes the improved principal component analysis to deal with the data more objectively,and theability to reduce the dimension is better,it's contribute to the later training.Because of the data dimensionality reduction,the time required for training can be reduced when the sample size is large.(4)This paper introduced and studied the algorithm of support vector machine in depth,on the basis of many experiments,selectd the appropriate kernel functions and parameters,put the sample data which treatment with improved PCA in advance into SVR to make prediction training,then built gas emission prediction model which based on improved PCA and SVR algorithm.(5)The prediction model given by the paper was simulated by MATLAB software.The method of mine statistics,BP neural network,support vector regression were used to predict the same sample data,then campared their experimental results each other.The results showed that,the average relative prediction error of the prediction method proposed by this paper is 1.36%,the mean square error is 0.01,the prediction error is less than the other models,the experiment achieved good result.(6)Finally,the gas emission prediction software platform was designed and programmed by using MATLAB GUI,which can realize visual operation and friendly man-machine interface,it makes operation of gas emission prediction simple and feasible.