基于CEEMD联合TGSCSO-LSTM算法的变压器油中气体浓度预测方法  

Prediction Method of Gas Concentration in Transformer Oil Based on CEEMD and TGSCSO-LSTM

作　　者：彭继慎[1] 夏玲云 王燚增 PENG Jishen;XIA Lingyun;WANG Yizeng(Faculty of Electrical and Control Engineering,Liaoning Technical University,Huludao 125105;Beijing EHV Power Transmission Company,State Grid Jibei Electric Power Co.,Ltd.,Beijing 102488)

机构地区：[1]辽宁工程技术大学电气与控制工程学院,葫芦岛125105 [2]国网冀北电力有限公司超高压分公司,北京102488

出　　处：《电气工程学报》2024年第4期407-415,共9页Journal of Electrical Engineering

基　　金：辽宁省自然科学基金(2019-ZD-0039);辽宁省教育厅科学技术研究创新团队(LT2019007)资助项目。

摘　　要：油中溶解气体浓度的预测可为电力变压器状态评估与早期故障诊断提供重要的数据依据。由此,针对长短期记忆网络(Long short-term memory network,LSTM)预测模型参数选择困难的问题,同时为提高变压器油中溶解气体浓度预测的精度,提出一种基于CEEMD联合TGSCSO-LSTM算法的变压器油中气体浓度预测方法。利用互补集合经验模态分解算法(Complementary ensemble empirical mode decomposition,CEEMD)将原始气体浓度序列分解为一系列具有一定频率特征的分量,以提高原始序列的可预测性能;针对各分量分别建立LSTM预测模型,同时利用经Tent映射随机初始化种群与高斯扰动改进的沙丘猫群优化算法(Sand cat swarm optimization,SCSO)对LSTM网络参数进行优化选取,以提高算法的预测精度;最后重构各个分量的预测结果以获取最终的油中溶解气体浓度预测结果。利用某500 kV变压器实际气体浓度数据对所提方法进行对比试验,试验结果表明,所提方法油中溶解气体浓度预测性能优良,具有较好的应用价值。The prediction of dissolved gas concentration in oil can provide important data basis for power transformer condition assessment and early fault diagnosis.Therefore,in order to solve the problem of difficult parameters selection of long short-term memory network(LSTM)prediction model,and to improve the prediction accuracy of dissolved gas concentration in transformer oil,a gas concentration prediction method in transformer oil is proposed based on complementary ensemble empirial mode decomposition(CEEMD)combined with TGSCSO-LSTM algorithm.The CEEMD algorithm is used to decompose the original gas concentration series into a series of components with certain frequency characteristics to improve the predictable performance of the original series.The LSTM prediction model is established for each component,meanwhile,the LSTM network parameters are optimized and selected by using the Tent mapping random initialization population and Gaussian disturbance improved sand cat swarm optimization algorithm(SCSO)to improve the prediction accuracy of the algorithm.Finally,the predicted results of each component is reconstructed to obtain the final predicted results of dissolved gas concentration in oil.The proposed method is tested by using the actual gas concentration data of a 500 kV transformer.The experimental results show that the proposed method has excellent prediction performance of dissolved gas concentration in oil and has good application value.

关 键 词：油中溶解气体 互补集合经验模态分解 沙丘猫群优化算法 长短时记忆神经网络 

分 类 号：TM411[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]