邱梅, 许高瑞, 宋光耀,等.PCA-WNN模型在导水裂隙带高度预测中的应用研究[J].河南理工大学学报（自然科学版）,2023,42(6):27-36.

QIU M, XU G R, SONG G Y,et al.Research on application of PCA-WNN model in predicting the development height of water-flowing fractured zones[J].Journal of Henan Polytechnic University(Natural Science) ,2023,42(6):27-36.

PCA-WNN模型在导水裂隙带高度预测中的应用研究

邱梅1,2, 许高瑞1, 宋光耀1, 施龙青1,2

1.山东科技大学 地球科学与工程学院，山东 青岛  266590;2.山东科技大学 山东省沉积成矿作用与沉积矿产重点实验室，山东 青岛  266590

摘要:导水裂隙带是煤层顶板水害的主要通道，准确预计导水裂隙带高度是预测和防治煤层顶板水害的首要任务。基于导水裂隙带发育的复杂性及其影响因素之间的相关性，结合煤矿生产实际和工程地质理论，选取采高、工作面斜长、硬岩岩性比例系数、采深和煤层倾角作为主控因素，利用线性回归拟合及灰色关联分析法分析各主控因素与导水裂隙带发育高度的相关性。将主成分分析（principal component analysis，PCA）与小波神经网络（wavelet neural network，WNN）相结合，通过PCA消除各主控因素间的相关性及冗余信息，构造无相关性的主成分作为WNN的输入因素，建立导水裂隙带高度预测的PCA-WNN模型。结果表明：PCA-WNN模型能有效消除因素间的相关性，预测相对误差为-6.66%~6.13%，平均4.46%，较单纯的WNN模型预测精度高且稳定。将该模型应用于山东新巨龙煤矿1302N工作面，得到了较为准确的预测结果，为导水裂隙带高度预测提供了新方法。

关键词:PCA-WNN模型;导水裂隙带高度;相关性分析;主成分分析;小波神经网络

doi:10.16186/j.cnki.1673-9787.2022070055

基金项目:国家自然科学基金资助项目（51804184）；山东省大学生创新创业训练计划项目（S202110424107）

收稿日期:2022/07/24

修回日期:2023/03/01

出版日期:2023/11/25

Research on application of PCA-WNN model in predicting the development height of water-flowing fractured zones

QIU Mei1,2, XU Gaorui1, SONG Guangyao1, SHI Longqing1,2

1.College of Earth Sciences and Engineering，Shandong University of Science and Technology，Qingdao  266590，Shandong，China;2.Shandong Provincial Key Laboratory of Depositional Mineralization & Sedimentary Minerals，Shandong University of Science and Technology，Qingdao  266590，Shandong，China

Abstract:The water-flowing fractured zone serves as the primary pathway for roof water influx in coal mines.Accurate prediction of the development height of this zone is crucial for anticipating and mitigating roof water hazards.Given the intricacies of the water-flowing fractured zone and the interdependencies among predictive factors，we have combined practical coal production data with engineering geological theory. Five key factors were identified： mining height，inclined length of the working face，ratio coefficient of hard rock lithology，mining depth，and the coal seam dip angle.By combining Principal Component Analysis（PCA）and Wavelet Neural Network（WNN），correlations and redundant information among the main controlling factors were eliminated through PCA.The uncorrelated principal components were subsequently used as input factors for WNN to establish the PCA-WNN model for predicting the height of the water-flowing fractured zone.The results indicated that the PCA-WNN model effectively eliminated correlations among factors，resulting in higher prediction accuracy and stability compared to the conventional WNN model.The relative error ranged from -6.66% to 6.13%，with an average of 4.46%.The PCA-WNN model was applied to forecast the height of the water-flowing fractured zone in the No.1302N working face of the Xinjulong coal mine in Shandong Province，and the predictions were found to be reliable.Therefore，this study presents a viable method for predicting the height of the water-flowing fractured zone in coal seam roofs in coal mines.

Key words:PCA-WNN model;height of the water-flowing fractured zone;correlation analysis;principal component analysis;wavelet neural network

004_2022070055_邱梅_L.pdf