基于逆向传播神经网络的工作面顶板矿压数据预测  

Roof pressure data prediction for working face based on back propagation neural network

作　　者：程海星 朱磊 宋立平 刘文涛 徐凯 CHENG Haixing;ZHU Lei;SONG Liping;LIU Wentao;XU Kai(China Coal Research Institute Co.,Ltd.,Xi'an 710054,China)

机构地区：[1]中煤能源研究院有限责任公司,陕西西安710054

出　　处：《煤矿安全》2021年第5期216-220,共5页Safety in Coal Mines

摘　　要：为了将人工智能技术中的逆向传播神经网络应用于顶板矿压数据预测,以11个影响工作面顶板矿压数据预测的主要因素作为输入参数,以4个工作面顶板矿压参数为输出参数,确定隐含层层数为1层,隐含层神经单元个数为24个,并在此基础上建立了基于逆向传播神经网络的工作面顶板矿压数据预测模型。利用王家岭及周边煤矿具有代表性的工作面顶板矿压数据建立学习样本,以王家岭煤矿12309工作面实测矿压数据为验证样本,对预测模型精度进行了检验。经分析,利用逆向传播神经网络模型得到的初次来压步距、初次来压强度、周期来压步距、周期来压强度预测值与实测值相对误差分别为0.043 343 653、0.006 077 606、0.006 401 138、0.020 236 088,即矿压数据预测总体上相对误差小于5%,符合工程应用允许的误差范围,说明所建立的逆向传播神经网络模型具有较高的准确性和可靠性。In order to apply the back propagation neural network in artificial intelligence technology to the prediction of roof strata pressure data,11 main factors affecting the prediction of roof rock pressure data of working faces are taken as input parameters,and roof rock pressure parameters of four working faces are taken as output parameters.The number of hidden layers is determined as 1 layer and the number of hidden layer neural units is 24.On this basis,the prediction model of roof pressure data based on the back propagation neural network is established.Based on the representative roof pressure data for working face of Wangjialing and its surrounding coal mines,the study samples are established,and the measured mine pressure data of 12309 working face in Wangjialing Coal Mine were used as the verification samples to test the accuracy of the prediction.After analysis,the relative errors between the predicted values and the measured values of the first weighting interval,the first weighting strength,the periodic weighting interval and the periodic weighting strength are 0.043343653,0.006077606,0.006401138 and 0.0202236088,respectively,which means that the overall relative errors of the mining pressure data are less than 5%,which is in line with the allowable error range of the engineering application.This indicates that the established back propagation neural network model has high accuracy and reliability.

关 键 词：逆向传播神经网络 矿压 数据预测 来压强度 来压步距 

分 类 号：TD323[矿业工程—矿井建设]