顶板瓦斯抽采巷因具有大流量和连续抽采的优点，被广泛用于高瓦斯或突出矿井回采工作面瓦斯治理。如何确定合理的顶板巷布置位置，以高效抽采采空区卸压瓦斯，是保障工作面瓦斯治理效果的关键。为此，在深入分析顶板瓦斯抽采巷布置原则及其布置位置影响因素的基础上，提出了一种基于GA–BP神经网络模型的顶板瓦斯抽采巷布置位置智能预测方法；采用灰色关联分析法确定了GA–BP神经网络模型的预测指标，并设计开发了顶板瓦斯抽采巷布置位置智能预测系统。研究结果表明：①工作面的采厚、埋深、覆岩结构、煤层倾角、倾向长度等5个物理指标是顶板瓦斯抽采巷布置位置的主控因素，且其权重值排序由大到小依次为采厚、埋深、覆岩结构、煤层倾角、倾向长度。②随着遗传代数的增加，GA–BP神经网络适应度不断减小，且当遗传代数为60时其适应度变化基本稳定，表明GA–BP神经网络初始权重和偏置效果较好。③在当前训练样本数据集的前提下，基于GA–BP神经网络模型的顶板瓦斯抽采巷布置位置的预测结果与实际工况值的相对误差仅为0.43%～11.27%，在可接受的范围内。该研究可为顶板瓦斯抽采巷精准设计提供一定的参考。

The roof gas drainage roadway, with its advantages of large flow and continuous extraction, is widely used in the gas control of high gas or outburst mine working faces. How to determine the reasonable arrangement position of the roof roadway to efficiently extract the pressure-relief gas in the goaf is key to ensuring the effect of gas control on the working face. Therefore, through a deep analysis of the arrangement principles of the roof gas drainage roadway and the main controlling factors of its arrangement position, an intelligent prediction method for the arrangement position of the roof gas drainage roadway based on the GA–BP neural network model is proposed. The prediction indicators of the GA–BP neural network model were determined using the grey correlation analysis method, and an intelligent prediction system for the arrangement position of the roof gas drainage roadway was designed and developed. The research results show: ① The mining thickness, burial depth, overlying rock structure, coal seam dip angle, and dip length of the working face are the main controlling factors for the arrangement position of the roof gas drainage roadway, and their weight values are ranked as: mining thickness > burial depth > overlying rock structure > coal seam dip angle > dip length; ② With the increase of genetic generations, the fitness of the GA–BP neural network continuously decreases, and when the genetic generation is 60, its fitness change is basically stable, indicating that the initial weight and bias of the GA–BP neural network are good; ③ Under the premise of the current training sample data set, the relative error of the prediction result of the arrangement position of the roof gas drainage roadway based on the GA–BP neural network model and the actual working condition value is only 0.43%～11.27%, which is within an acceptable range. This research can provide a certain reference for the precise design of the arrangement of the roof gas drainage roadway.