基于时间序列预测算法的变压器套管油中溶解气体预测研究  

Prediction of Dissolved Gas in Transformer Bushing Oil Based on Time Series Prediction Algorithm

作　　者：蒲曾鑫 李波 白洁 杨昊 黄宇 牧灏 吕黔苏 PU Zengxin;LI Bo;BAI Jie;YANG Hao;HUANG Yu;MH Hao;LYU Qiansu(Electric Power Research Institute of Guizhou Power Grid Co.,Ltd.,Guiyang 550001,Guizhou,China;School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710048,Shanxi,China)

机构地区：[1]贵州电网有限责任公司电力科学研究院,贵州贵阳550001 [2]西安工程大学电子信息学院,陕西西安710048

出　　处：《电力大数据》2025年第2期27-37,共11页Power Systems and Big Data

基　　金：中国南方电网有限责任公司科技项目(0666002023030103HX00010)。

摘　　要：套管作为电力变压器的关键组成部分,其油纸绝缘结构容易受到多种因素的影响,从而产生潜在缺陷。该研究旨在预估变压器套管内油中溶解气体的组分含量,进而评估变压器套管的绝缘状态。为此,综合选用了自回归积分滑动平均(autoregressive integrated moving average,ARIMA)模型、长短时记忆(long short-term memory,LSTM)神经网络以及灰色预测(discrete grey geometric model,DGGM)模型,对变压器套管绝缘油在未来时间点的溶解气体进行了预测。通过对比这些模型的性能,发现优化后的ARIMA模型在预测气体含量和评估绝缘状态方面表现最佳,而DGGM模型预测CO 2的平均绝对误差约为优化后ARIMA模型的2.5倍。该研究成果能够为保障电力系统安全稳定运行提供技术支撑。Bushing is a key component of power transformer,are prone to potential defects in their oil-paper insulation structure due to various factors.This study aims to predict the content of dissolved gases in transformer bushings oil and evaluate their insulation status.To this end,we used the autoregressive integrated moving average(ARIMA)model,the long short-term memory(LSTM)neural network,and the discrete grey geometric model to predict the dissolved gas content of transformer bushings oil in the future time points.By comparing the performance of these models,we found that the optimized ARIMA model performed best in predicting gas content and evaluating insulation status,while the DGGM model's average absolute error in predicting CO_(2)was about 2.5 times that of the optimized ARIMA model.This research result can provide important technical support for the safe and stable operation of the power system.

关 键 词：变压器套管 时间序列预测算法 油中溶解气体 绝缘性能 

分 类 号：TM41[电气工程—电器]