主成分分析与遗传算法优化ELM模型的矿井突水水源识别方法  

Identification method of mine water inrush source combined with PCA-GA-ELM mine

作　　者：郭中安 吴洪斌 姜海滨 李相通 张海龙 GUO Zhongan;WU Hongbin;JIANG Haibin;LI Xiangtong;ZHANG Hailong(College of Earth Science and Engineering, Shandong University of Science and Technology, Qingdao, Shandong 266590, China;Shaanxi Zhongtai Energy Investment Co. , Ltd. , Yulin, Shaanxi 719109, China)

机构地区：[1]山东科技大学地球科学与工程学院,山东青岛266590 [2]陕西中太能源投资有限公司,陕西榆林719109

出　　处：《中国科技论文》2021年第9期1010-1016,1022,共8页China Sciencepaper

摘　　要：为了实现矿井突水水源的快速准确识别,提出了主成分分析(principal component analysis,PCA)、遗传算法(genetic algorithm,GA)和极限学习机(extreme learning machine,ELM)相结合的突水水源判别模型。模型以ELM分类为基础,利用PCA将含水层6种水化学离子指标归纳为3种主成分,通过GA优化ELM,结合良庄煤矿51101工作面实测数据资料建立判别模型,并与传统ELM模型和BP神经网络预测模型进行对比,再实际应用到李楼煤矿1303工作面的突水水源识别。结果表明:PCA、GA和ELM相结合的突水水源判别模型能够有效消除水化学离子指标间的相互影响,优化分类模型的权值和阈值,使矿井突水水源判别更为准确。In order to realize the fast and accurate identification of mine water inrush source,a discriminant model of water inrush source combining principal component analysis(PCA),genetic algorithm(GA)and extreme learning machine(ELM)was proposed.Based on the classification of ELM,the model uses principal component analysis to classify the six hydrochemical ion indexes of the aquifer into three principal components,optimizes ELM through genetic algorithm,and establishes a discriminant model combining with the measured data of working face 51101 of Liangzhuang coal mine,and compares it with the traditional ELM model and BP neural network prediction model.The results show that the discriminant model combining PCA,GA and ELM can effectively eliminate the mutual influence between hydrochemical ion indexes,optimize the weight and threshold of the classification model,and make the discriminant of mine water inburst source more accurate.

关 键 词：矿井突水 水化学特征分析 主成分分析 遗传算法 极限学习机 水源识别 

分 类 号：TD163[矿业工程—矿山地质测量]