基于逻辑回归模型的瓦斯浓度异常值检测与预警

Detection and warning of abnormal gas concentration based on logical regression model

在煤层地质构造不确定性、开采强度变化、瓦斯抽采方案调整和人员作业管理等因素的影响下,瓦斯浓度监测数据通常会存在异常情况。仅依据瓦斯浓度阈值指标不能有效反映危险程度的演化过程,进行预警存在一定风险,为提升瓦斯浓度异常值检测的准确性,提出了一种基于逻辑回归模型的瓦斯浓度异常值检测与预警新方法。建立瓦斯数据检测逻辑回归模型后,得到了合理的回归系数并进行了检测试验。试验预测数据结果表明,利用建立的逻辑回归模型识别异常数据的正确率达85%以上,可以有效识别未达到阈值上限的瓦斯数据是否存在异常状态,为优化矿井瓦斯浓度的监测与预警提供了新方法。

Under the influence of coal seam geological structure,mining intensity,gas extraction and personnel operation management,gas concentration monitoring data usually have abnormal situation.Because only the gas concentration threshold value cannot effectively reflect the evolution of risk degree,there is a certain risk in using this index for early warning.In order to improve the accuracy of gas concentration abnormal value detection,a new method of gas concentration anomaly detection and early warning based on logical regression model is proposed.By establishing the logical regression model of gas data detection,the reasonable regression coefficient is obtained and the detection test is carried out.The results show that the accuracy of using the established logistic regression model to identify abnormal data is more than 85%,which can effectively identify whether there is abnormal state in the gas data that does not reach the upper limit of the threshold,and can provide a new method to optimize the monitoring and early warning of the gas concentration in the mine.