摘要
为了对电力变压器进行更精确的故障诊断,提出一种基于粒子群优化(PSO)双向长短时记忆(Bi-LSTM)网络的变压器故障诊断方法。在5种变压器油气体成分的基础上加入三比值法构建的3个参量作为输入特征。采用粒子群算法对模型中的超参数进行优化,通过组合模型PSO-Bi-LSTM对变压器故障进行诊断分析并与其它方法进行对比。算例分析结果表明,所提模型的故障诊断准确率高达92.5%要优于传统方法,当样本特征数减少或数据集出现错误时,所提模型的诊断准确率仅有少许下降,说明该模型还具有较优的鲁棒性。
In order to perform more accurate fault diagnosis on power transformers,a transformer fault diagnosis method based on particle swarm optimization bidirectional long short-term memory(Bi-LSTM)network is proposed.First,on the basis of the five transformer oil gas components,three parameters constructed by the three-ratio method were added as input features.Then,the hyperparameters in the model were optimized using the particle swarm algorithm,and finally the combined model PSO-Bi-LSTM was used to diagnose and analyze transformer faults and compare with other methods.The analysis results of the calculation example show that the fault diagnosis accuracy rate of the proposed model is as high as 92.5%,which is better than the traditional method.When the number of sample features decreases or the data set has errors,the diagnosis accuracy rate of the proposed model only decreases slightly,which indicates that the model also has better robustness.
作者
樊清川
于飞
宣敏
FAN Qing-chuan;YU Fei;XUAN Min(School of Electrical Engineering,Naval University of Engineering,Wuhan Hubei 430033,China)
出处
《计算机仿真》
北大核心
2022年第11期136-140,共5页
Computer Simulation
关键词
电力变压器
故障诊断
粒子群优化
双向长短时记忆网络
Power transformer
Fault diagnosis
Particle swarm optimization
Bi-directional long short-term memory network