论文部分内容阅读
为分析矿井瓦斯涌出的非线性关系、指标间复杂联系和准确预测瓦斯涌出量,基于主成分分析和灰色关联理论在克服指标的共线性、相关性对瓦斯涌出量影响,兼顾二者关联性之上,确定主要指标,建立瓦斯涌出量预测的距离模型和灰色模型,进而基于支持向量机非线性,构建非线性组合预测模型.利用训练样本学习和最小绝对百分比误差确定预测模型参数,并以沈阳某矿某工作面为例,运用已构建模型预测瓦斯涌出量.研究结果表明:日产量、采出率与其他指标的共线性相对较强,煤层间距、临近层厚度及层间岩性与其他指标的共线性相对最弱;该模型绝对百分比误差最大为5.83%,预测精度相对高于各个单项预测模型,大幅降低预测风险.
In order to analyze the non-linear relationship between gas emission from mines, the complicated relationship between indicators and the accurate prediction of gas emission, based on principal component analysis and gray relational theory, the influence of co-linearity and correlation of indexes on gas emission is considered The main index and the distance model and the gray model of the gas emission prediction are established, and then the non-linear combination prediction model is constructed based on the support vector machine nonlinearity.The prediction model parameters are determined by the training sample learning and the minimum absolute percentage error , And taking a working face of a mine in Shenyang as an example, the gas emission has been predicted using the constructed model.The results show that the collinearity of daily output and recovery rate is relatively strong with other indicators, and the coal seam spacing, the thickness and layer The collinearity between the lithology and other indexes is relatively weakest. The absolute percentage error of the model is 5.83%, and the prediction accuracy is relatively higher than that of the individual prediction models, which greatly reduces the prediction risk.