基于改进D-S证据理论与深度学习的矿用电缆缺陷识别研究  

Research on defect assessment of mining cable based on improved D-S evidence theory and deep learning

作　　者：孟强 舒珊 秦晓梅 郭振振 孔宁宁 刘瑞国[2] MENG Qiang;SHU Shan;QIN Xiaomei;GUO Zhenzhen;KONG Ningning;LIU Ruiguo(Nantun Coal Mine,Yankuang Energy Group Co.,Ltd.,Jining,Shandong 273500,China;College of Intelligent Equipment,Shandong University of Science and Technology,Taian,Shandong 271019,China)

机构地区：[1]兖矿能源集团股份有限公司南屯煤矿,山东省济宁市273500 [2]山东科技大学智能装备学院,山东省泰安市271019

出　　处：《中国煤炭》2025年第1期181-188,共8页China Coal

摘　　要：准确识别局部放电缺陷模式在矿用电缆的缺陷评估中至关重要,煤矿供电环境复杂,矿用电缆缺陷识别也尤为重要。因此提出了一种融合改进D-S证据理论与深度学习的方法,构建基于Efficientnet-b0和Resnet-18的深度学习识别模型用于提取矿用高压电缆局部放电信号的关键特征并进行初步分类,引入D-S证据理论对单一模型的识别结果进行融合。针对证据冲突的情况,引入基尼不纯度改进D-S理论中的权重分配,从而提高矿用电缆缺陷识别的准确率。现场试验表明,融合后的模型平均识别率为94.2%,双模型融合的各项性能均比单一模型有所提高,有效提高了矿用电缆缺陷识别的准确度,为煤矿配电网安全可靠运行提供保障。Accurately identifying partial discharge defect patterns is crucial in the defect assessment of mining cables.Given the complex power supply environment in coal mines,defect identification of mining cables was also particularly important.A method of fusion-improved D-S evidence theory and deep learning was proposed,and a deep learning recognition model based on Efficientnet-b0 and Resnet-18 was established to extract the key features of partial discharge signals of high-voltage mining cables for mining and carry out preliminary classification,and D-S evidence theory was introduced to fuse the recognition results of a single model.In response to conflicting evidence,the Gini impurity was introduced to improve the weight allocation in D-S theory,thereby enhancing the accuracy of identifying defects in mining cables.The field test showed that the average recognition rate of the fused model was 94.2%,and the performance of the dual-model fusion was improved compared with the single model,which effectively improved the accuracy of the defect recognition of the mining cable,and provided guarantee for the safe and reliable operation of the coal mine distribution network.

关 键 词：矿用电缆 局部放电 D-S证据理论 深度学习 模式识别 双模型融合 

分 类 号：TD687[矿业工程—矿山机电]