针对变压器缺陷样本存在不平衡性,现有诊断方法适用性不足的问题,提出一种面向油中溶解气体分析(Dissolved Gas Analysis,DGA)不平衡数据的变压器缺陷识别方法.首先,基于内蒙古西部地区部分电厂送检油样的DGA分析,重新划分5种变压器状态类型,并通过增加状态特征增大其差距;然后,构建综合考虑检测气体含量、无编码比值、三比值的输入特征,对智能诊断模型进行训练和测试;最后,选用改进动态多种群粒子群优化算法优化极限学习机神经网络参数,通过对未定义类型样本进行分析,验证方法的有效性.结果表明,本方法适用于DGA不平衡数据集,能够对变压器运行状态及缺陷进行准确识别.
A Transformer Defect Recognition Method for DGA Unbalanced Data
In response to the problem of imbalanced transformer defect samples and insufficient applicability of the existing diagnostic methods,a transformer defect recognition method for Dissolved Gas Analysis(DGA)imbalanced data in oil is proposed.Firstly,based on DGA analysis of oil samples sent by some power plants in western region of Inner Mongolia,five types of transformer states are reclassified and the differences are increased by adding state features.Then,input features considering gas content detection,uncoded ratio,and triration are constructed to train and test the intelligent diagnostic model.Finally,the improved dynamic multiple Particle Swarm Optimization(PSO)algorithm is selected to optimize the parameters of the Extreme Learning Machine(ELM)neural network.By analyzing samples of undefined types,the effectiveness of the method is verified.The results show that this method is applicable to DGA imbalanced datasets and can accurately identify the operating status and defects of transformers.
transformerdefect recognitiondissolved gas analysis(DGA)imbalancedynamic multiple particle swarm optimizationextreme learning machine
万方数据
维普
