吴虹，崔登伟，张际涛，等. 基于XGBoost算法的煤矿导水裂隙带发育高度预测研究[J].河南理工大学学报(自然科学版)，doi:10.16186/j.cnki.1673-9787.2024090024

WU H，CUI D W，ZHANG J,et al．Research on the prediction of the development height of water-conducting fractured zones in coal mines based on the XGBoost algorithm［J］.Journal of Henan Polytechnic University( Natural Science) ，doi:10.16186/j.cnki.1673-9787. 2024090024

基于XGBoost算法的煤矿导水裂隙带发育高度预测研究(网络首发)

吴虹¹，崔登伟¹，张际涛²，丁恒³

（1.贵州省地质矿产勘查开发局 一〇六地质大队，贵州 遵义 563000；2.贵州黔诚力锦科技有限公司，贵州 贵阳 550081；3.贵州省地质环境监测院，贵州 贵阳 550081)

摘要: [目的] 为精确预测煤矿导水裂隙带发育高度，选取开采深度H、开采厚度M、工作面长度l和覆岩硬岩岩性比例b作为导水裂隙带发育的主要影响因子。[方法] 建立基于XGBoost 算法的导水裂隙带高度预测模型，选取决定系数（R2）与平均绝对百分比误差（MAPE）作为模型评价指标。[结果] 结果表明：XGBoost 的最优树深为8，最优棵数为250，相比于其他模型，该参数下 XGBoost 模型在训练集和测试集上 R2 更大，MAPE 更小，模型优于其他模型；选取赵固二矿工作面参数计算导水裂隙带发育高度，比较 XGBoost 模型预测结果与实测结果。分析XGBoost 模型预测值、相似模拟、理论计算与工作面实测导水裂隙带高度之间的差异，XGBoost 模型预测值与实测值相对误差与绝对误差均较小，优于相似模拟与理论计算的；模型误差处于工程容许误差范围内，后续可使用 XGBoost 模型对导水裂隙带高度进行预测。[结论] 研究结果可为准确预测煤矿导水裂隙带发育高度提供一定参考。

关键词: 导水裂隙带；XGBoost算法；集成学习；极端梯度；相似模拟

中图分类号：TD325

doi: 10.16186/j.cnki.1673-9787.2024090024

基金项目: 国家自然科学基金资助项目（51764010，51874019）

收稿日期：2024-09-11

修回日期：2025-04-25

网络首发日期：2025-05-07

Research on the prediction of the development height of water-conducting fractured zones in coal mines based on the XGBoost algorithm

WU Hong¹，CUI Dengwei¹，ZHANG Jitao², DING Heng³

(1. 106 Geological Group, Guizhou Bureau of Geology and Mineral Resources Exploration and Development, Zunyi 563000, Guizhou, China;2. Guizhou Qianchenglijin Technology Co., Ltd., Guiyang 550081,Guizhou,China;3. Guizhou Institute of Geological Environment Monitoring, Guiyang 550081, Guizhou, China)

Abstract: [Objective] In order to accurately predict the development height of the water-conducting fractured zone in coal mines, the mining depth H, mining thickness M, working face length l, and the proportion b of hard rock lithology in the overlying strata are selected as the main influencing factors for the development of the water-conducting fractured zone. [Methods] Establish a height prediction model of the water-conducting fractured zone based on the XGBoost algorithm, and select the coefficient of determination (R2) and the mean absolute percentage error (MAPE) as the model evaluation indicators. [Results] The optimal tree depth of XGBoost is 8, and the optimal number of trees is 250. Compared with other models, under this parameter, the XGBoost model has a larger R2 and a smaller MAPE on the training set and test set, and the model is superior to other models. The parameters of the working face of Zhaogu No. 2 Coal Mine are selected to calculate the development height of the water-conducting fracture zone, and the predicted results of the XGBoost model are compared with the measured results. By comparing the differences between the predicted values of the XGBoost model, similar simulations, theoretical calculations, and the measured height of the water-conducting fracture zone at the working face, the relative error and absolute error between the predicted value and the measured value of the XGBoost model are relatively small, which is better than similar simulations and theoretical calculations. The model error is within the allowable error range of the project, and the XGBoost model can be used to predict the height of the water-conducting fracture zone in the follow-up. [Conclusion] The results of this research provide valuable insights for accurately predicting the development height of the water-conducting fracture zone in coal mines, offering a reliable tool for practical applications.

Key words: water-conducting fracture zone; XGBoost algorithm; integrated learning; extreme gradient;analog simulation

CLC: