煤炭工程 ›› 2022, Vol. 54 ›› Issue (2): 140-146.doi: 10.11799/ce202202025

• 研究探讨 • 上一篇    下一篇

基于多方法耦合的矿井突水水源识别模型研究

宋立兵1,董东林2,王振荣3,李果4,杨茂荣3   

  1. 1. 中国神华集团神东煤炭集团公司哈拉沟煤矿
    2. 中国矿业大学(北京)
    3. 神东煤炭集团公司
    4. 国能神东煤炭技术研究院
  • 收稿日期:2020-12-15 修回日期:2021-11-23 出版日期:2022-02-14 发布日期:2022-07-06
  • 通讯作者: 董东林 E-mail:ddl@cumtb.edu.cn

Identification Model of Water Inrush Source Based on Piper-PCA-OT-Regression-Bayes

  • Received:2020-12-15 Revised:2021-11-23 Online:2022-02-14 Published:2022-07-06

摘要: 为快速准确地判别矿井突水水源,减少矿井突水事故带来的危害,以保德矿为例,选取Ca2+、Mg2+、Na++K+、SO2-4、Cl-、HCO-3共6种水化学指标作为判别指标,通过分析各含水层水化学特征,确定了各含水层代表水样,以此为基础建立了耦合主成分分析-离群值检验-回归填补法-贝叶斯判别法的矿井突水水源判别模型,并将模型判别结果与PCA-Bayes模型判别结果做出对比。结果表明:保德矿采空区、二叠系砂岩含水层、石炭系砂岩含水层、奥灰含水层的水质类型分别为HCO3-Ca·Na·Mg型、HCO3-Na型、HCO3-Na型和HCO3·SO4-Ca·Na·Mg型|保德矿水样主成分为Ca2+、Mg2+、Na++K+、SO2-4,可作为综合指标反映保德矿原始水样数据信息|待测水样中的异常值,可通过离群值检验和线性回归模型确定并校正|对比数据校正前后Bayes模型判别结果,校正后准确率为95%,判别准确度明显提升,可准确高效的识别突水水源。

关键词: 矿井突水, 水源判别, 水化学分析, 离群值检验, 回归填补法, 贝叶斯模型

Abstract: In order to identify the water source of mine water inrush quickly and accurately, and reduce the harm caused by mine water inrush, Baode mine was taked as an example, Ca2+、Mg2+、Na++K+、SO42-、Cl-、HCO3- were selected as the discrimination indexes. Through The representative water samples of each aquifer Were determined by analyzed the chemical characteristics of water samples. And then, the coupled Principal Component Analysis (PCA) - Outlier Tests (OT) - Regression - Bayes model were used to identify the water source of Baode mine. The results show that: the water quality types of the Goaf, Permian sandstone aquifer, Carboniferous sandstone aquifer and Ordovician limestone aquifer in Baode mine were HCO3-Ca.Na.Mg, HCO3-Na, HCO3-Na, HCO3-Ca. The main components of Baode mine water samples are Ca2+、Mg2+、Na++K+、SO42-, which can be used as comprehensive indicators to reflect the original water sample data information of Baode mine. The outliers in the tested water samples were determined and corrected by Outliers Test and Regression model. Compared the results of before and after data correction, the accuracy after correction was 95%, and the accuracy was significantly improved, which can accurately and efficiently identify water inrush source.

中图分类号: