基于CEEMDAN-SG-BiLSTM的变压器油中溶解气体体积分数预测  

Prediction for Dissolved Gas Concentration in Power Transformer Oil Based on CEEMDAN⁃SG⁃BiLSTM

作　　者：陈铁[1,2] 陈一夫 李咸善[1,2] 陈卫东 冷昊伟 陈忠 CHEN Tie;CHEN Yifu;LI Xianshan;CHEN Weidong;LENG Haowei;CHEN Zhong(College of Electrical Engineering and New Energy,China Three Gorges University,Hubei Yichang 443002,China;Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station,China Three Gorges University,Hubei Yichang 443002,China;Yichang Electric Power Survey and Design Institute Co.,Ltd.,Hubei Yichang 443000,China)

机构地区：[1]三峡大学电气与新能源学院,湖北宜昌443002 [2]三峡大学梯级水电站运行与控制湖北省重点实验室,湖北宜昌443002 [3]宜昌电力勘测设计院有限公司,湖北宜昌443000

出　　处：《高压电器》2023年第12期168-175,共8页High Voltage Apparatus

基　　金：国家自然科学基金资助项目(51741907);梯级水电站运行与控制湖北省重点实验室开放基金(2019KJX08)。

摘　　要：变压器油中溶解气体体积分数可对变压器潜伏故障诊断提供重要依据,为了更精确的预测变压器油中溶解气体体积分数,提出了CEEMDAN结合BiLSTM网络的组合预测方法,先针对EMD中的模态混叠现象和EEMD重构误差较大的问题,提出了CEEMDAN,将气体序列分解,得到模态分量,并对高频波动分量进行Savitzky-Golay(SG)滤波,消弱高频分量中的极值点和噪声干扰。然后利用BiLSTM网络对各个分量进行预测,进一步提高特征提取的完整性。叠加重构所有分量的预测结果,得到变压器油中溶解气体体积分数预测值。经算例验证,相较其他模型,所提方法精度更高,证实其有效性。The dissolved gas concentration in transformer oil can provide an important evidence for potential fault di-agnosis of transformer.For predicting the dissolved gas concentration in transformer oil more accurately,a combined prediction method of CEEMDAN and BiLSTM network is proposed.Firstly,for the issue of modal aliasing in EMD and lager reconstruction error in EEMD,the CEEMDAN is proposed.The gas sequences are decomposed to get the modal component and the high frequency fluctuation component is subjected to Savitzky-Golay(SG)filtering to weak-en the extreme value point of high frequency component and the noise interference.Then,the BiLSTM network is used to predict each component to improve the global feature extraction further.Finally,the prediction value of dis-solved gas concentration in the transformer oil is obtained by superposition and reconstruction of the prediction re-sults of each component.It is confirmed by the calculation that,compared with other modules,the proposed method is more accurate and its effectiveness is verified.

关 键 词：变压器 油中溶解气体 添加自适应白噪声完全集合经验模态分解 Savitzky-Golay滤波 双向长短期记忆网络 

分 类 号：TM41[电气工程—电器] TP183[自动化与计算机技术—控制理论与控制工程]