样本不平衡下基于CNN的主变压器局部放电图谱识别

Recognition of partial discharge patterns of power transformers based on CNN under unbalanced sample

对主变压器进行局部放电的识别能提前发现缺陷,降低故障率。目前,现场检测的局放数据更多以图谱形式存在,且存在样本不平衡的特点。鉴于此,文中提出了一种基于卷积神经网络的主变局放图谱识别方法。通过将预处理的图谱作为输入,使用MobieNetV2模型进行训练,同时使用改进的损失函数以解决样本不平衡问题。最后验证了提出的方法能有效地解决样本不平衡的问题,且96%的识别率明显优于其他方法。

Partial discharge identification of the power transformer can find the defects early and reduce the failure rate.At present,the partial discharge data from on-site detection are more in the form of unstructured patterns,and there are characteristics of extremely unbalanced samples.On this basis,this paper proposes a Convolutional Neural Network-based method for indentifying the partial discharge patterns.Using the pre-processed power transformer partial discharge patterns as input,and the lightweight MobieNetV2 model is used for training,and the improved loss function was used to solve the problem of sample imbalance at the same time.Finally,it is verified that the proposed method can effectively solve the problem of sample imbalance,and the 96% recognition rate is significantly better than other methods.