论文部分内容阅读
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。
In order to effectively analyze mine gas monitoring data and achieve accurate and reliable gas concentration prediction, a time series analysis and prediction method of gas concentration is studied based on gray relational cluster analysis and Gaussian process regression model. The prediction validity is used as the evaluation index of the prediction accuracy to dynamically determine the dimensionality of the time series sample space of the reconstructed gas concentration. The gray correlation analysis method is used to divide the gas concentration time series into several sample sets. The samples with the correlation features are regarded as virtual Variables to eliminate the prediction error caused by the random and uncertain factors. The Gaussian process regression model is used to predict the gas concentration interval, and the prediction result is expressed as a interval with high reliability, Prediction of Gas Concentration Dynamic Change in Future Period. The case study shows that the prediction result is accurate and reliable, which can well reflect the actual change of gas concentration.