基于机器学习的油浸式变压器故障状态诊断模型

打开文本图片集

中图分类号：TM411;TP368.39 文献标志码：A文章编号：1001-5922（2025）09-0188-05

Machine learning-based fault state diagnosis model for oil-immersed transformer

Abstract：The fault characteristics of oil-immersed transformers are complex，and they are verysusceptible to external factors，and eventuallyfail.Afteratransformer failureoccurs，itnotonlycauses power outages，but canalso lead to serious acidentssuchasfires ifnot diagnosed promptly.Intraditional diagnostic methods，itisverysusceptibleto environmental factors，but in fault diagnosis，neither theaccuracy nor the diagnostic performance can meet the currentrequirements.Based on this，the source and composition of insulating oil in transformers were analyzed in this paper，and its data were normalized.Due to the short training timeandeasy implementation ofthe Extreme Learning Machine（ELM），the fault diagnosis wascariedout，and theoptimal numberof hidden layer nodes was 4O after simulation analysis.In this study，an improved Black Widow Spider Optimization Algorithm（IBWOA）was proposed，and its performance was tested bytest functions，and the effctiveness of the improved strategy wasverified. Theresults showed that compared with the traditional feature combination，the ELM fault diagnosis rate was more than 89% ，and the IBWOA fault accuracy rate was increased by 4.28% . Compare machine learning algorithms to verifythe superiority of the models.

Key words：machine learning;oil immersed transformer;fault diagnosis ;fault location

油浸式变压器作为电力系统中的核心设备，承担着电压变换、电能传输与分配的关键任务，其运行状态直接关系到电网的安全性与可靠性。（剩余5722字）