基于指标气体的 BP 神经网络煤自燃预测系统

摘要

针对煤矿自然发火的预测问题，在指标气体分析法的基础上，构建 BP 神经网络，选取CH4/ CO、O2/ CO2这两组指标气体浓度比作为网络的输入以降低通风条件的影响，经过训练后，判断检测点是否发火并以0或1的形式输出。网络经过43次训练后，误差达到预设的范围（﹤0.0001）。研究表明，利用 BP 神经网络处理从煤层收集到的气体浓度并作出发火预报是可行的且具有相当优势的。%In order to forecast coal spontaneous combustion,take advantage of BP neural network. The input of the neural network is the concentration of CO、CO2 and CH4 in different temperature and use CH4-to-CO、O2-to-CO2 ra-tio. In this way,the influence of the wind will be little. After trained,the network can show 0 or 1 which representing fire or not. After trained 43 times,the error is lower than 0. 000 1. It proves that BP neural network can deal with the date of coal mine. What′s more,BP neural network has huge advantages.