煤炭易自燃煤层采空区温度预测方法研究  

Study on prediction method of goaf temperature in coal spontaneous combustion seam

作　　者：张怡[1] 陈莹[1] 陈宠 Zhang Yi;Chen Ying;Chen Chong(Land and Space Application Research Center,College of Engineering and Technology,Chengdu University of Technology,Leshan 614000,China)

机构地区：[1]成都理工大学工程技术学院国土空间应用研究中心,四川乐山614000

出　　处：《能源与环保》2021年第7期94-99,共6页CHINA ENERGY AND ENVIRONMENTAL PROTECTION

基　　金：成都理工大学工程技术学院院级课题(C12201913)。

摘　　要：常用温度预测方法建立的温度预测模型,多采用正向传播的方式,导致温度预测结果与实际测量结果存在较大的误差。针对该问题,提出煤炭易自燃煤层采空区温度预测方法。根据采空区中存在的易燃气体,选择气体浓度传感器采集样本数据,并针对数据采集过程中存在的缺失、重复、多单位等问题进行删除、填充和归一化处理;引入神经网络,建立温度预测模型,采用正向传播和反向传播2种方式训练模型误差;从平均绝对误差、平均绝对百分比误差、均方根误差和判定系数4个方面检验温度预测模型,输出温度预测结果。实验结果表明:设计煤炭易自燃煤层采空区场景,研究方法的预测温度结果与测量温度结果之间的绝对误差小于1,相对误差小于0.01,可认为所述温度预测方法具有较好的精度,能够为煤炭开采提供一定的参考价值。Most of the temperature prediction models established by common temperature prediction methods adopt the forward propagation mode,which leads to a large error between the temperature prediction results and the actual measurement results.To solve this problem,the temperature prediction method of goaf in coal spontaneous combustion coal seam was put forward.According to the flammable gas in the goaf,the gas concentration sensor was selected to collect sample data,and the deletion,filling and normalization were carried out in view of the problems of missing,repetition and multiple units in the data collection process;introducing neural network,establishing temperature prediction model,and training model error by forward propagation and backward propagation;testing the temperature prediction model from four aspects:average absolute error,average absolute percentage error,root mean square error and judgment coefficient,and output the temperature prediction results.Experimental results showed that:the absolute error between the predicted and measured temperature results of the research method is less than 1,and the relative error is less than 0.01.It can be considered that the temperature prediction method has better accuracy which can provide certain reference value for coal mining.

关 键 词：煤炭 自燃煤层 采空区 温度预测 神经网络 正向传播 反向传播 

分 类 号：TD713[矿业工程—矿井通风与安全]