基于灰色关联熵的煤与瓦斯突出概率神经网络预测模型*

煤与瓦斯突出是严重威胁矿井安全生产的重大自然灾害之一。为解决煤与瓦斯突出影响因素、突出危险性关联预测问题，在综合分析煤与瓦斯突出影响因素的基础上，利用灰色关联熵理论分析影响因素与突出危险性的关联度，得到各影响因素的权重及关联度排序，并结合概率神经网络（PNN）原理，构建基于灰色关联熵的煤与瓦斯突出PNN预测模型。用煤与瓦斯突出样本数据，对影响因素加权的PNN模型进行训练和测试。结果表明：用灰色关联熵分析可获得影响因素与突出危险性的关系，量化输入变量的重要性；瓦斯放散初速度、开采深度对于煤与瓦斯突出危险性的影响程度最大，可重点对瓦斯放散初速度、开采深度进行预处理以产生更为理想的预测效果；该预测模型能更好地考虑影响因素对突出危险性的综合影响，改善煤与瓦斯突出危险性预测的准确性。

Probabilistic neural network prediction model of coal and gas outburst based on grey relational entropy

Coal and gas outburst is one of the most serious natural disasters which threaten the safety production of coal mine. In order to solve prediction problem in influential factors of coal and gas outburst associated relationships with the outburst risk , based on a comprehensive analysis of influence factors of coal and gas outburst, using grey relational entropy theory to analysis correlation of factors with outburst risk, getting the weight and correlation order of influencing factors , and combining with principle of probabilistic neural network (PNN) , PNN prediction model of coal and gas outburst based on grey correlation entropy was built. Using the sample data of coal and gas outburst , the PNN model of influencing factors weighted was trained and tested. The results show that the relationship between influence factors and outburst risk by using grey correlation entropy analysis is obtained , and the importance of input variables is quantified; initial speed of methane diffusion and mining depth have maximum relevance to coal and gas outburst risk, and preprocessing of initial speed of methane emission, mining depth can be focused to produce more ideal prediction effect ; the prediction model can better consider that influencing factors have a comprehensive effect on outburst risk, and improve the accuracy of risk prediction of coal and gas outburst.