基于BP神经网络的煤与瓦斯突出预测系统开发  

Development of Coal and Gas Outburst Prediction System Based on BP Neural Network

作　　者：曲方[1] 张龙[1] 李迎业 李忠群 

机构地区：[1]中国计量学院质量与安全工程学院 [2]潞安矿业集团有限责任公司五阳煤矿,山西襄垣046200

出　　处：《中国安全科学学报》2012年第1期11-16,共6页China Safety Science Journal

摘　　要：煤与瓦斯突出影响因素多,难以为其建立合适的多指标非线性预测模型,为提高突出预测的准确性和增强预测预报方法的实用性,采用改进的BP算法建立煤与瓦斯突出预测数学模型。通过研究不同算法的突出预测效果,对已建模型的泛化能力进行检验,利用Matlab GUI和神经网络工具箱设计开发煤与瓦斯突出预测系统,通过向系统输入已知的突出样本数据,经过学习、训练,实现对未知参数的预测。仿真结果表明:网络在训练300次后,误差训练曲线的均方差(MSE)可以达到10-15,实际预测误差也小于0.1,系统得到的5组数据预测结果与实际情况相符。The factors affecting coal and gas outburst are so many that it is difficult to establish an appro- priate non-linear prediction model with many indicators for it. In order to improve the accuracy of forecast and enhance the practicability of prediction method, a coal and gas outburst prediction model was estab- lished by the improved BP algorithm. Different algorithms were investigated on outburst prediction model and generalization ability of the model was tested. The prediction system was designed by Matlab GUI and Neural Network Toolbox. After known parameters were entered into the system, forecasting results could be got through learning and training. Simulation results show that the MSE （ Mean Sguared Errov） of training error curves can get to 10 -15, when network is trained 300 times, and the actual prediction error is less than 0. 1, which shows that the forecasting results obtained by the system accord with the actual situation.

关 键 词：BP神经网络 改进模型 煤与瓦斯突出 预测系统 图形用户界面(GUI) 

分 类 号：X936[环境科学与工程—安全科学]