人工智能在煤矿瓦斯风险评估中的应用  

Application of artificial intelligence in coal mine gas risk assessment

作　　者：申小明 SHEN Xiaoming(Shanxi Coal Transportation and Sales Group Guxian Dongrui Coal Industry Co.,Ltd.,Linfen 042405,China)

机构地区：[1]山西煤炭运销集团古县东瑞煤业有限公司,山西临汾042405

出　　处：《陕西煤炭》2024年第9期168-172,共5页Shaanxi Coal

摘　　要：瓦斯事故是影响煤矿安全生产的主要事故类型,为降低瓦斯事故风险,提出一种创新实用的煤矿瓦斯风险等级评价方法,为煤矿瓦斯事故的防治提供帮助。研究主要包括3个步骤,首先,收集煤矿瓦斯事故的真实数据;其次,由于属性特征较多,数据集具有过于高维、大规模和高复杂性的结构特征,采用t分布随机邻域嵌入(t-SNE)方法处理复杂的高维气体事故数据;最后,利用遗传算法(GA)对支持向量机(SVM)进行优化,对煤矿瓦斯事故的严重程度进行预测。结果表明,通过对预测效果、误差分布、时间成本等性能的比较,引入t-SNE的评价模型可以准确预测89%的事故结果,同时节省约60%的时间成本。Gas accidents are the main type of accidents affecting the safety production of coal mines.In order to reduce the risk of gas accidents,an innovative and practical coal mine gas risk level evaluation method is proposed to provide assistance for the prevention and control of coal mine gas accidents.The study mainly includes three steps:first,the real data of coal mine gas accidents are collected;second,due to the large number of attribute features,the dataset is too high-dimensional,large-scale,and high-complexity structural characteristics,thus the t-distributed random domain embedding(t-SNE)method is used to process the complex high-dimensional gas accident data;and finally,the genetic algorithm(GA)is adopted to optimize the support vector machine(SVM)to predict the severity of the coal mine gas accident.The results show that by comparing the performance of prediction effect,error distribution,and time cost,the t-SNE-introduced evaluation model can accurately predict 89%of accidents,saving about 60%of the time cost.

关 键 词：风险评估 煤矿瓦斯事故 t-分布随机邻域 遗传算法(GA) 支持向量机(SVM) 

分 类 号：TD712[矿业工程—矿井通风与安全]