基于PLS-BP神经网络模型的导水裂缝带高度预测  

Forecast of Water Flowing Fractured Zone Height Based on PLS-BP Neuron Network Model

作　　者：黄欢[1,2,3] HUANG Huan(China Coal Research Institute, Beijing 100013, China;CCTEG Xi’ an Research Institute, Xi’ an 710077, China;Shaanxi Province Coal Mine Water Disaster Prevention & Cure Key Laboratory, Xi’ an 710077)

机构地区：[1]煤炭科学研究总院,北京100013 [2]中煤科工集团西安研究院有限公司,陕西西安710077 [3]陕西省煤矿水害防治技术重点实验室,陕西西安710077

出　　处：《煤矿开采》2016年第6期6-10,共5页Coal Mining Technology

基　　金：“十二五”国家科技支撑计划课题(2012BAK04B04);国家级自然科学基金青年科学基金(41402220);陕西省自然基金项目(2011JQ5015)

摘　　要：针对淮南煤田走向长壁垮落式采煤法条件下导水裂缝带高度难以精确预测的问题,建立基于偏最小二乘法的BP神经网络模型,提高了导水裂缝带高度的预测精度。首先运用偏最小二乘法对导水裂缝带高度的影响因素进行分析,对原始数据降维处理提取主成分,优化了原始数据,克服了变量间因样本量小而产生的多重相关性影响,并对自变量、因变量具有很强的解释能力。再将提取的主成分作为BP神经网络模型的输入层,导水裂缝带高度为输出层,对网络进行训练。该方法既简化了网络结构,其精度也高于经验公式以及单一的偏最小二乘法模型与BP神经网络模型。To the problems of water flowing fractured zone height precise forecast based on long-wall caving mining along the strikemethod of Huainan coal field, BP neuron network model was built on partial least square regression, forecast precision of water flowingfractured zone height was improved. Firstly, the influence factors of water flowing fractured zone height were analyzed by partial leastsquare regression method, the main factors were extracted by dimension reduction of original data, and original data was optimized,the multi-correlation that induced by small sample capacity between different variable quantity was conquered, it has doughty explain abilityfor self-variable and dependent variable/ then the main factors that extracted was as input layer of BP neuron network model, andwater flowing fractured zone was output layer, and net was trained. The network structure was simplify according it, the precision wasexceed than experience formula, single partial least square regression and BP neuron network model.

关 键 词：导水裂缝带 多重相关性 偏最小二乘回归 BP神经网络 

分 类 号：TD325[矿业工程—矿井建设]