人工智能在长壁煤矿井下瓦斯实时预测中的应用

Application of Artificial Intelligence in Real-Time Prediction of Underground Gas in Longwall Coal Mine

检测长壁开采面中爆炸性甲烷-空气混合物的形成仍然是一项具有挑战性的任务,尽管大气监测系统和计算流体动力学建模用于检查甲烷浓度,但它们不足以作为关键区域(例如靠近切削鼓)的实时预警系统,建立了长短期记忆算法,在爆炸发生前预测和管理长壁开采作业中的爆炸性气体区。介绍了一种采用人工智能算法的新方法,即改进的长短期记忆,以检测长壁开采面中爆炸性甲烷-空气混合物的形成,并在爆炸性气体成为危害之前识别可能的爆炸性气体积聚。该算法基于CFD模型输出对采煤机的六个位置进行了训练和测试,以匹配切割机的相似位置和操作条件。结果表明,该算法可以在3D中预测爆炸性气体区域,对于不同的设置,总体准确度在87.9%~92.4%;在将测量数据输入算法后,输出预测花费了2 min。结果发现,与计算流体动力学和大气监测系统相比,使用所提出的算法可以更快、更突出地覆盖准确的实时爆炸性气体聚集预测。

Detecting the formation of explosive methane-air mixtures in longwall mining faces remains a challenging task,and although atmospheric monitoring systems and computational fluid dynamics modelling are used to check methane concentrations,they are insufficient to serve as a real-time early warning system for critical zones(e.g.,close to cutting drums),and a long-and short-term memory algorithm was established to predict and manage explosive gas in longwall mining operations before an explosion occurs in the Zone.A new approach using an artificial intelligence algorithm,improved long short term memory,is presented to detect the formation of explosive methane-air mixtures in longwall mining faces and to identify possible explosive gas build-ups before they become a hazard.The algorithm was trained and tested on six locations of a coal miner based on CFD model outputs to match similar locations and operating conditions of the cutter.The results showed that the algorithm could predict explosive gas areas in 3D with an overall accuracy of between 87.9% and 92.4% for different setups;the output prediction took two minutes after feeding the measured data into the algorithm.It was found that accurate real-time explosive gas accumulation predictions can be covered faster and more prominently using the proposed algorithm compared to computational fluid dynamics and atmospheric monitoring systems.