With the increasing demand for gas, the problem of local anomalies in deep gas urgently needs to be addressed. This study takes the Lvjiatuo coal mine as the research area, analyzes the main influencing factors of coal seam gas accumulation law in the research area, and predicts the deep gas content in the research area based on a neural network model optimized by particle swarm optimization (PSO-BP). The research results indicate that the development of folds and fault structures only plays a certain degree of control over the accumulation and distribution of gas content in the coal seams of Lvjiatuo coal mine, and the burial depth of the coal seams is the dominant factor affecting the deep gas accumulation in the research area; The proportion of mudstone of coal seam roof has a significant impact on gas accumulation, and under similar geological factors, the higher the proportion of mudstone of coal seam roof, the more gas content there is. The PSO-BP neural network model based on the data of coal seam burial depth, structural complexity index, roof proportion of mudstone, and coal seam thickness in the research area can accurately predict the deep gas content in the area, and the predicted values are accurately fitted with the measured values. The research results have important theoretical and practical significance for studying the law of deep gas accumulation and predicting deep gas content.

随着天然气需求的不断增加，深层天然气局部异常问题亟待解决。本研究以吕家沱煤矿为研究区，分析了研究区煤层瓦斯成藏规律的主要影响因素，并基于粒子群优化优化的神经网络模型对研究区深部瓦斯含量进行了预测（ PSO-BP）。研究结果表明，褶皱和断层构造的发育仅对吕家沱煤矿煤层瓦斯含量的聚集和分布起到一定程度的控制作用，而煤层埋深是影响瓦斯含量的主导因素。研究区深层天然气成藏；煤层顶板泥岩比例对瓦斯成藏影响显着，在相似地质因素下，煤层顶板泥岩比例越高，瓦斯含量越多。基于研究区煤层埋深、结构复杂性指数、顶板泥岩比例、煤层厚度等数据的PSO-BP神经网络模型可以准确预测该区深层瓦斯含量，预测值为与测量值准确吻合。研究成果对于研究深层天然气成藏规律、预测深层天然气含量具有重要的理论和实践意义。