PCA-BP在回采工作面瓦斯涌出量预测中的应用

Application of PCA-BP to gas emission prediction of mining working face

为对回采工作面瓦斯涌出量进行准确预测,运用主成分回归分析以及BP神经网络原理和方法,结合现场实测数据,采用多元统计分析软件SPSS处理相关数据,研究影响回采工作面瓦斯涌出量各因素间的相关关系并提取主成分,以确定BP神经网络中的输入参数,建立BP神经网络进行预测.利用PCA-BP神经网络方法建立瓦斯涌出量预测模型.研究结果表明:采用PCA-BP神经网络方法的预测值与实际值最大相对误差为2.820%,最小相对误差为2.036%,平均相对误差为2.357%,较其他预测模型有更高精度.对降低事故发生率和矿井延深水平的回采工作面瓦斯涌出量预测具有较好的指导作用.

In order to accurately predict gas emission quantity in working face, this paper studied the relationships among 11 influence factors that influencing gas emission quantity in working face and extracted principal component, using principal component regression analysis and the principle and method of BP neural network. By combining field measured data and using multivariate statistical analysis software SPSS, the forecast model of gas emission quantity was established using multiple regression analysis, and input parameter of BP neural network was determined. The results of study show that the maximum relative error between predictive value of adapting PCA-BP neural network method and actual value is 2.820%, the minimum relative error is 2.036%, and the average relative error is 2.357%. The established model has a higher accuracy than that of other forecasting models. It plays a good guiding role for reducing accident rate and the prediction of gas emission quantity in working face of mine extended level.