基于无量纲数的PCA-MRA在煤矿瓦斯涌出量预测中的应用

Application of PCA-MRA Based on Dimensionless Quantity in Prediction of Gas Emission in Coal Mine

为解决煤矿瓦斯涌出量预测过程中存在的指标繁杂致使预测精度低的问题,首先运用R语言中的主成分分析法(PCA)对煤矿瓦斯涌出量的影响因素进行降维分析;然后引入量纲分析理论,采用无量纲化处理方法消除量纲对数据的影响;最后利用多元回归分析法(MRA)结合多种多元回归模型,最终选用多元线性回归预测模型对煤矿瓦斯涌出量进行预测,并选取部分数据对所建立的煤矿瓦斯涌出量预测模型进行验证。结果表明:主成分分析法能有效减少预测变量的个数,经主成分旋转后构建的多元回归预测模型的预测精度较高,其相对误差平均值为1.99%;主成分分析与多元线性回归分析相结合的方法适用于煤矿瓦斯涌出量的预测,且满足工程需要,可为井下瓦斯涌出量预测技术的选择提供参考。

In order to deal with the poor prediction accurac of the coal mine gas emission for the existing comlicated prediction index, this article analyses the influencing factor of gas emission by using principal component analysis method in R language.Then the paper introduces dimension analysis theory, and uses dimensionless processing to eliminate the influence of dimension on data. At the end the paper applies the multivariate linear regression combined with a variety of multiple regression models to predicting the gas emission in coal min. The paper selects some data to verify the predicted function of gas emission. The results show that the principal component analysis method can effectively reduce the number of predictors, the multivariate regression prediction model constructed by the principal component rotation has higher precision,and the average value of its relative error is 1. 99%.The combination of principal component analysis and multiple linear regression is applicable to the prediction of gas emission in coal mine,meets the engineering needs.This method can provide some reference for the selection of underground gas emission prediction technology.