Research On Gas Emission Prediction Based On Improved GA-SVM Algorithm

In the process of coal production safety, the coal mine gas emission is a very important link. When monitoring and forecasting the gas concentration, we need to establish a suitable model, a complete and advanced gas safety monitoring and control system, so as to promote the improvement of the coal industry automation system, and these systems play a key role on the safety of the downhole operation production. Therefore, this paper explores a kind of based on improved genetic and support vector machine algorithm (GA-SVM) with the combination of mining the prediction method of gas emission in the process of gas monitoring data.The development of coal gas system can be considered as nonlinear dynamic system with time, and it has the characteristics of high complexity and randomness. In addition, the coal gas monitoring is affected by many factors. So it produces partial problems when using single algorithm during the coal gas emission forecast. In order to solve these problems, firstly, this paper analyzes the inherent nature of coal gas emission formation and the emission influencing factors, and then it takes advantage of the characteristics of support vector machine (SVM) which can solve the small sample, nonlinear. Afterwards the predict concept is introduced in the genetic algorithm, and it provides a way to solve the problem of the gas emission forecast.In this paper, we make the following work:Firstly, using of the gas emission model to deathly research the selected impact factor. Secondly, it proposes the genetic support vector machine forecasting model and improve genetic and support vector machine algorithm. Selecting suitable kernel function and using genetic algorithm to optimize and support parameters of support vector machine to improve the prediction performance. Finally, the analysis and error analysis of improved GA-S VM prediction model is compared with constructed models with the other two algorithms, and the accuracy of the proposed algorithm is validated by simulation and experiment.