自适应神经网络在确定落煤残存瓦斯量中的应用

落煤残存瓦斯量的确定是采掘工作面瓦斯涌出量预测的重要环节，它直接影响着采掘工作面瓦斯涌出量预测的精度，并与煤的变质程度、落煤粒度，原始瓦斯含量、暴露时间等影响因素呈非线性关系，人工神经网络具有表示任意非线性关系和学习的能力，是解决复杂非线性，不确定性和时变性问题的新思想和新方法，基于此，作提出自适应神经网络的落煤残丰瓦斯量预测模型，并结合不同矿井落煤残存瓦斯量的实际测定结果进行验证研究，结果表明，自适应调整权值的变步长BP神经网络模型预测精度高，收敛速度快，该预测模型的应用可为采掘工作面瓦斯涌出量的动态预测提供可靠的基础数据，为采掘工作面落煤残存瓦斯量的确定提出了一种全新的方法和思路。

A NEW APPROACH TO DETERMINING THE REMNANT METHANE QUANTITY IN FALLEN COALS BY MEANS OF SELF-ADAPTING NEURAL NETWORK

The present article intends to introduce a new approach to solving the problem in determining the remnant methane quantity in fallen coals by means of self-adapting neural network. As is known, it has been a big problem for the coal production to predict the remnant methane quantity in fallen coals so as to ensure the safety and security of the regular coal mine production. Since the amount of methane emitted from the coal-digging face directly influences the accuracy of prediction, it becomes a nonlinear relation with such likely factors that may influencing such prediction accuracy, as degree of coal metamorphosing, fallen coals grain size, quantity of initial methane in coal, exposure time and so on. Such a problem of non-linear nature can actually be solved by adopting artificial neural network systems. Based on the above thought, the authors of this article come up with a predicting model for determining the remnant methane quantity in fallen coals based on self-adapting neural network. The new approach has further been tested and verified by actual determining results with the remnant methane quantity in different coal mines. The results show that the BP neural network model of adjusting step length self-adaptively is of high accuracy and fast convergence. The application of the new model supplies a reliable basic approach to dynamic prediction of the amount of methane emitted from digging face with fallen coals.