基于BP神经网络的煤层自燃预测

在全面分析影响煤层自燃因素的基础上,建立了煤层自燃预测的人工神经网络模型.应用该模型对某煤田的多个煤层样本进行了训练和预测,网络经过10次训练后,误差达到设定的最小值,6次预测测试中最大误差仅为0.027 8,最小的为0.000 1.研究表明,该模型精度较高,可用于预测煤层自燃的实际应用.

Coal seam spontaneous combustion prediction based on neural network

A new forecasting method is presented using artificial neural network characterized with highly non-linear scanning and tracing power. As is known, although general engineers and technicians in the coal mine practice have accumulated a lot of data and materials in this way, coal seam spontaneous combustion is still an extremely complicated physical-chemical and environmental process to be harnessed. The inefficiency of the traditional methods for forecasting such combustions is largely due to the enormous complexity of the process, which can only be overcome, according to the present authors, by using the neural network for its massive implicit knowledge from these materials concerned and powerful tracing and mapping capability. In this paper, after analyzing all the factors that are likely to affect coal spontaneous combustion, we established a neural network model for forecasting such combustions. When we tried to use this model to some coal seam samples of a coalfield, first of all training and forecasting practice are given. Ten times of training would be enough to reduce the network error to the supposed minimum value. The biggest error is only 0. 027 8 in the six forecast tests we have done, with the smallest error being 0. 000 1, Thus, the forecasting model proves efficient enough, to meet the demands for its own tasks to be performed and reliable and precision enough for the forecasting with its computation and procession tasks performed conveniently and efficiently in the Matlab software.