基于CNN-BiGRU-ATTENTION和知识图谱的变压器缺陷识别及检修决策方法
Transformer Defect Recognition and Maintenance Decision-making Method Based on CNN-BiGRU-ATTENTION and Knowledge Graph
韦明源1
作者信息
- 1. 南宁交通资产管理有限责任公司,广西 南宁 530000
- 折叠
摘要
传统变压器运维需依靠人工从繁杂的运行缺陷记录中判断设备缺陷严重程度,然后做出检修决策导致缺陷处理效率偏低.为此,提出由卷积神经网络(convolutional neural network,CNN)、双向门控循环单元(bidirectional gated recurrent unit,BiGRU)与注意力机制组合而成的CNN-BiGRU-ATTENTION 算法实现变压器缺陷分类识别;然后,根据变压器运行标准、试验规程和运行经验基于知识图谱构建变压器检修决策库;最后,根据缺陷记录分类识别结果和检修决策库中缺陷检修措施的对应关系,由检修决策库推送缺陷检修措施,实现变压器缺陷记录到运维检修的智能化决策过程,以某一市级电网 35 kV 及以上变压器缺陷记录为训练样本,试验结果表明模型缺陷记录分类准确率达 88%以上,基于缺陷记录分类结果由检修决策库智能化推送检修决策,有效提高运维效率,具有一定的实用价值.
Abstract
Traditional transformer operation and maintenance(O&M)requires manual judgment of the severity of equipment defects from complex operation defect records and then makes maintenance decisions,leading to low defect handling efficiency.To address this issue,this paper proposes a CNN-BiGRU-ATTENTION algorithm composed convolutional neural network(CNN),bidirectional gated recurrent unit(BiGRU),and attention mechanism to realize transformer defect classification recognition;then,based on transformer standards,test procedures,and operation experience,a transformer maintenance decision-making knowledge graph is constructed;finally,according to the classification recognition results of defect records and the relationship between defect maintenance measures in the maintenance decision-making knowledge graph,the maintenance decision-making knowledge graph pushes the defect maintenance measures,realizing the intelligent decision-making process from defect records to O&M maintenance.Taking the 35 kV and above transformer defect records of a city-level power grid as training samples,the experimental results that the accuracy of defect record classification of the model is over 88%,and based on the defect record classification results,the intelligent push of maintenance decisions from maintenance decision-making knowledge graph effectively improves the O&M efficiency,which has certain practical value.
关键词
变压器缺陷/文本分类/知识图谱/变压器检修Key words
transformer defects/text classification/knowledge graph/transformer maintenance引用本文复制引用
出版年
2024
NETL
NSTL
万方数据