基于BP神经网络的煤自然发火预报系统

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2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(0.000 1)。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。

Forecasting System of Spontaneous Combustion of Coal Based on BP Neural Network

BP neural network has been constructed to forecast the coal spontaneous combustion through gas analysis.In this system,the concentration ratio of CH4/CO and O2/CO2 are selected as the data input of the network,with little influence from the variation of the atmospheric conditions.After training,the network can forecast spontaneous combustion by outputting data in form of 0 or 1.After trained 43 times,the error limit reaches the expected range(< 0.000 1).It proves that BP neural network has feasibility and superiority in forecasting the coal spontaneous combustion through analyzing the concentration of gas collected from the coal seam.