基于PCA—ELM的煤矿突水预测方法研究  

Research of coal mine water burst forecasting method based on PCA-ELM

作　　者：李培 陈颖[1] 马小平[1] 于得祥[1] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221008

出　　处：《工矿自动化》2013年第9期46-50,共5页Journal Of Mine Automation

基　　金：国家自然基金科学基金资助项目(60974126)

摘　　要：针对采用传统算法建立煤矿突水预测模型存在训练速度慢、泛化能力差、测试精度不高等问题,提出了一种将PCA与ELM相结合的煤矿突水预测方法,并利用该方法建立了煤矿突水预测模型。该方法以煤矿突水历史数据为样本,利用PCA得到煤矿突水主控因素,将仅包含主控因素的样本数据划分为训练集、验证集和测试集;然后把训练样本作为ELM的输入,对模型进行训练;最后利用样本数据验证模型。实验结果表明,相较于传统算法,该方法输入变量少,建模和运算时间短,模型的运行速度和预测精度较高。In view of problems of slow training speed,poor generalization performance and low test precision existed in coal mine water burst forecasting model with traditional algorithm,the paper proposed a coal mine water burst forecasting method based on PCA-ELM,and established coal mine water burst forecasting model with the method.The method takes mine water burst historical data as sample,uses PCA to obtain mine water burst controlling factors,and divides the sample data containing the main controlling factors into train set,validation set and test set;then,takes the train set as input of ELM to train the model;finally,uses validation set to validate the model.The experimental results show that input variables of the method are less than the traditional methods,time of modeling and computing is shorter,operating speed and prediction accuracy of the model are higher.

关 键 词：煤矿突水 预测模型 预测方法 极限学习机 主成分分析法 

分 类 号：TD745[矿业工程—矿井通风与安全]