基于偏最小二乘回归的回采工作面瓦斯涌出量预测模型

Prediction Model for Gas Emission Quantity in Mining Face Based on Partial Least Squares Regression

针对回采工作面瓦斯涌出量回归建模过程中自变量间出现多重共线性问题,提出应用偏最小二乘回归(PLS)对瓦斯涌出量进行预测的建模思路。选取地质及采矿2个方面共12个参数指标作为回归因子,利用15个瓦斯涌出实例为建模样本,建立了回采工作面瓦斯涌出量的偏最小二乘回归模型。建立的模型对训练样本拟合效果良好,最大误差为6.09%,平均误差仅为2.06%;对其余几个案例进行预测,精度优于主成分分析和BP神经网络,与最小二乘-支持向量机模型相当。研究表明,基于偏最小二乘回归进行工作面瓦斯涌出量预测是一种有效可行的方法。

Aiming at the problem that the independent variables appear multi - collinear in regression modeling process of mining face gas emission, an idea using the partial least - squares （PLS） regression technology to find prediction model of the gas emission quantity in mining face is put forward. The partial least squares regression model which regards geological and mining two respects in all 12 in- dex as regression factor to predict gas emission quantity mining face is established by using 15 gas emission examples for modeling sam-ple. There is a good fitting result to training sample by established regression models, that the maximum error is 6.09%, the average error is only 2.06%. The PLS model is better than principal component regression analysis method and BP neural network in the pre-diction of the remaining cases, and which is consistent with the least squares - support vector machines method. It is an effective and feasible method by using partial least squares regression to predict gas emission quantity of mining face.