基于支持向量机回归与分源预测法的瓦斯涌出量预测模型及其应用

Prediction Model for Gas Emission of Working Face Based on Support Vector Machine and Different-source Prediction and Its Application

针对目前瓦斯涌出量预测模型存在的局限性及精度低等问题,应用分源预测和支持向量机(SVM)的基本原理,将SVM回归与分源预测法相结合,并利用SVM对回采工作面的瓦斯涌出量进行回归分析和数值模拟,建立了SVM分源预测的数学模型,提出了SVM分源预测的新方法。数值实验表明,将训练成功的SVM模型对现场数据进行回归预测并对比预测结果与实际值发现,SVM比BP神经网络预测精度更高,训练样本期望输出与实际值的最大相对误差为1.45%,小于实际要求的5%,准确率较高,预测风险低,可以满足实际要求。

At present,for the problems existing in the gas emission prediction model such as low precision and so on,the basic principle of different source prediction and support vector machine are applied and SVM is used to make regression analysis and numerical simulation to mining face gas emission quantity,combining SVM regression and different source prediction. Mathematical model of the SVM different source prediction is established and a new method of SVM different source prediction is put forward. Numerical experiments show that using the successful raining support vector machine to make regression analysis of predicted field data and comparing the prediction results and the actual value,it＇s found that SVM prediction accuracy is higher than BP neural network. Maximum relative error of training sample expected output value compared to the actual value is 1. 45%,which is less than 5% of the actual requirements. The accuracy is higher,predict risk is lower,and it could meet the actual requirements.