瓦斯浓度区间预测的灰色聚类与高斯过程模型

为有效分析矿井瓦斯监测数据,以实现准确、可靠的瓦斯浓度预测,基于灰色关联聚类分析与高斯过程回归模型,研究瓦斯浓度时间序列分析与预测的方法。以预测有效度为预测精度的评估指标来动态确定重构瓦斯浓度时间序列样本空间的维数;应用灰色关联分析方法将瓦斯浓度时间序列划分成若干样本集,将其中具有关联特征的样本作为虚拟变量进行预测以消除因随机、不确定因素干扰而引起的预测误差;应用高斯过程回归模型实现瓦斯浓度区间预测,将预测结果表示成一个具有较高可信度的取值区间,以此表达对未来一段时间内瓦斯浓度动态变化情况的预测。实例分析表明:预测结果准确、可靠,能够较好地反映瓦斯浓度的实际变化状况。

Grey Clustering and Gaussian Process Model for Gas Concentration Interval Prediction

For the purpose of achieving accurate and reliable gas concentration prediction through effective analysis on gas measuring data in mines,based on grey relevancy clustering analysis and the Gaussian process regression model,the method for gas concentration time series analysis and prediction was studied.The dimension of reconstructing the sample space of gas concentration time series was determined dynamically by making prediction validity as the evaluating indicators of prediction accuracy;gas concentratio...