基于RBF神经网络的采空区煤自燃预测

Prediction of Coal Spontaneous Combustion in Goaf Based on RBF Neural Network

摘要: 本文以程序升温实验得到的指标气体实验数据为训练样本，建立了基于RBF神经网络模型的煤自燃预测模型，借助于MATLAB软件，对龙东煤矿7162工作面采空区的煤温进行了有效预测。该预测模型以CO浓度、C₂H₄浓度、C₂H₄/C₂H₆值作为输入单元，拓扑结构为3-12-1。预测结果表明，该神经网络模型预测效果很好，误差很小。可以将CO浓度作为煤自燃预测的主要指标气体，在煤温处于160℃~250℃之间时增加C₂H₄浓度作为辅助指标气体，在250℃以后增加C₂H₄/C₂H₆值辅助指标气体，提高煤自燃预测的准确度。该预测方法可为煤自燃防治提供科学指导。

Abstract: According to the training samples of the indicator gases data got from the procedural heating-up experiment, the model based on RBF neural network for predicting the coal spontaneous combustion was established, and the temperature of coal in 7162 goaf of Longdong coal mine was predicted effectively by MATLAB software. The prediction model takes CO concentration, C₂H₄ concentration and C₂H₄/C₂H₆ value as input unit, whose topological structure is 3-12-1. The prediction results show that prediction effect of the neural network model is very good and the error is very small. CO concentration can be taken as the main indicator gas for coal spontaneous combustion prediction, and the accuracy of coal spontaneous combustion prediction can be improved by adding C₂H₄ concentration as an auxiliary indicator gas when the coal temperature is between 160˚C~250˚C, adding C₂H₄/C₂H₆ value as an auxiliary indicator gas when the coal temperature is above 250˚C. The prediction method can provide scientific guidance for the prevention and control of coal spontaneous combustion.