基于拟牛顿优化算法BP神经网络的瓦斯灾害预测模型

Prediction Model of Gas Disaster Based on Quasi-Newton Algorithm BP Neural Network

矿山瓦斯突出与爆炸事故的预测预报是当前我国煤矿安全生产中急待解决的问题之一。引入BP神经网络的拟牛顿(Newton)优化算法,在保留空间实体相关和多种分布并存的前提下,讨论了建立拟牛顿优化算法BP神经网络瓦斯灾害预测预报模型的数学模型设计、网络结构设计和程序设计3个部分,并以济宁二号井为实例进行了测试。结果表明:该模型稳定、快速、预测精度高,能够较好地模拟矿山瓦斯突出与爆炸事故特征,对瓦斯灾害作出较准确的预测。

The forecasting of gas outburst and explosion accidents is one of the most pressing problem in current China＇s coal mine safety production.Introducing the Quasi-Newton optimization algorithm of BP neural network,this paper discusses the mathematical model,network architecture and programming design of establishing the gas disaster forecasting model of Quasi-Newton optimization algorithm BP neural network under the premise of keeping the relationship among the spatial entities and their distributions,and an instance of Jining No.2 coal mine is tested.The result shows that this model is stable,fast and high prediction accuracy,which can simulate the mine gas outburst and explosion accidents characteristics and make more accurate predictions on the gas disaster.