Research On Gas Emission Forecasting Based On The Least Square Support Vector Machine

In the mine ventilation design, gas drainage engineering design, gas prevention and control work, the gas emission prediction is an essential link. Forecasting accuracy directly affects the normal production in coal mine. For a long time, many scholars have conducted a lot of research for gas concentration prediction, and put forward many effective methods. Based on the sufficient study and research of some classical prediction method and the current hot spot prediction method, prediction model based on least squares support vector machine (SVM) has been established. Research results show that, prediction model based on least squares support vector machine has better prediction effect and better generalization performance compared with classical methods.Based on the classic forecasting methods, I used separate source method theory and selected mining coal seam in coal mining face, adjacent coal seam and gob area respectively as research object. For three different areas, I selected different subset of gas emission.In the further study of support vector machine (SVM) and least squares support vector machine (SVM) theory, least squares support vector machine (SVM) model of gas emission prediction has been established in this paper. Radial basis kernel function is chosen as the kernel function, with self-adjusting grid search method for kernel function parameter optimization. MATLAB simulation has carried on the analysis, testing the accuracy of the model. Artificial neural network method is selected to predict the same actual measurement data. Error analysis shows that the LSSVM prediction accuracy of forecasting model was obviously higher than that of the BP artificial neural network model. The least squares support vector machine (SVM) theory in gas emission prediction has a good prediction performance and generalization performance, and this theory has the rationality and superiority.Finally, meeting the needs of coal mine gas emission prediction real-time actual, I use MTTLAB GUI function design created gas emission forecast software based on least squares support vector machines. The software is the practical application of the theory, has friendly interface and easy to use.