电力信息与通信技术2025，Vol.23Issue(9)：28-34,7.DOI:10.16543/j.2095-641x.electric.power.ict.2025.09.04
基于DBSCAN和VMD-TCN-LSTM的变压器油中溶解气体预测方法
Prediction Method of Dissolved Gases in Transformer Oil Based on DBSCAN and VMD-TCN-LSTM
摘要
Abstract
Predicting dissolved gases in transformer oil can provide early insights into the transformer's operational trends.To address the issues of data noise interference and incomplete exploration of temporal dependency features by a single neural network,a transformer oil dissolved gas prediction method is proposed.Firstly,the density-based clustering algorithm DBSCAN and the 3σ criterion were used to clean the data to eliminate outliers.Secondly,this paper designs a combined model for predicting dissolved gas content in oil.The model uses variational mode decomposition(VMD)to reduce the impact of noise,and utilizes temporal convolutional networks(TCN)and long short-term memory(LSTM)networks to comprehensively extract local and long-term dependent features of the data to improve the accuracy of prediction.Experimental results show that this method achieves a mean absolute error(MAE)of 0.611×10-6 and a mean absolute percentage error(MAPE)of 1.95%when predicting the content of dissolved gases in transformer oil within 7 days,improving by 0.661×10-6 and 3.79%compared to the benchmark model.关键词
油中溶解气体/DBSCAN/变分模态分解/时域卷积网络/长短期记忆网络Key words
dissolved gases in oil/density-based clustering with noise/variational mode decomposition/temporal convolutional network/long short-term memory network分类
信息技术与安全科学引用本文复制引用
袁和金,陈龙,李金波,马欢,刘占康..基于DBSCAN和VMD-TCN-LSTM的变压器油中溶解气体预测方法[J].电力信息与通信技术,2025,23(9):28-34,7.基金项目
国家电网有限公司总部管理科技项目"基于电网资源业务中台多模态数据的变压器状态感知与预测性运维技术研究与应用"(5700-202340289A-1-1-ZN). （5700-202340289A-1-1-ZN）
