基于L-MBP神经网络的煤自然发火期预测

针对不同矿井、矿井不同工作面预测煤自然发火期的难题，采用具有较好收敛性的L-M（Levernberg-Marquardt）BP算法的神经网络，开发出煤自然发火期预测仿真系统软件。根据煤自燃的内凼和外因影响因素，在东北矿区各自燃采场收集了152个样本训练数据，并应用于该软件训练、预测其结果误差〈10％。在神东矿区进行预测时，预测结果与实际情况吻合较好，为制定预防采场自然发火的技术措施提供了可靠的技术参数。

Forecasting technique for spontaneous combustion stage on basis of L-MBP neural network

It is difficult to forecast the spontaneous combustion stage of coal at different mines and different workfaces in a mine.Therefore,the neural network with the arithmetic of L-M(Levernberg-Marquardt) BP which has better astringency was adopted in this paper.And on the basis of this,the forecasting emluator for the spontaneous combustion stage of coal was developed.By the inner factor and the outer factor of the spontaneous combustion of coal,152 samples were gathered from the workfaces with coal spontaneous combustion at northeast mine area.And the forecasting error is lower than 10% by using the forecasting emluator.The forecasting results are close to the fact when the forecasting emluator was used at Shendong coal mine area.So it is credible parameter for establishing the technical measures for preventing coal spontaneous combustion in minied-out area.