基于HS-ELM的油浸式变压器故障诊断

Fault Diagnosis of Oil Immersion Transformer Based on HS-ELM Algorithm

针对ELM模型的预测精度受初始的输入权重Wi和隐含层偏置矩阵bi的选择影响,运用和声搜索算法(HS)优化选择初始的输入权重Wi和隐含层偏置矩阵bi,提出一种基于HS-ELM的油浸式变压器故障诊断方法。将5种气体体积分数数据(H2,C2H2,CH4,C2H6和C2H4)当作HS-ELM变压器故障诊断模型的输入特征参数数据,不同故障类别标签作为HS-ELM的输出,建立HS-ELM油浸式变压器故障诊断模型。研究结果表明,在各个故障类别的诊断正确率和总体正确率上,HS-ELM均要高于GA-ELM,ELM和IEC三比值法,有效提高了变压器故障诊断的正确率。

The prediction accuracy of ELM model is affected by the selection of initial input weight Wi and implicit layer bias bi.Using Harmony Search(HS)to optimize the selection of initial input weight Wi and implicit layer bias bi,a fault diagnosis method of oil immersion Transformer based on HS-ELM was proposed.The volumetric fractions of H2,CH4,C2H6,C2H4,and C2H2 were selected as the characteristic parameters.Different fault categories were used as the output of HS-ELM to establish a fault diagnosis model for HS-ELM oil immersion Transformers.The results show that HS-ELM is higher than GA-ELM,ELM and tri-ratio method in the diagnosis accuracy and overall accuracy of various fault categories,which effectively improves the fault diagnosis accuracy of Transformers.