基于改进变量预测模型的变压器故障诊断

针对变压器故障诊断中的小样本、非线性、参数寻优难等问题,提出改进的变量预测模型的变压器故障诊断方法.分析变量预测模型和布谷鸟搜索算法结合解决小样本和非线性问题,指出其后期收敛速度慢,稳定性差,收敛精度不高,易陷入局部极小值问题,在此基础上在谷鸟搜索算法位置更新中引入变异操作,提高解的多样性.引入动态步长和动态发现概率提...

Transformer fault diagnosis based on improved variable prediction models

Aiming at the problems of small sample size,non-linearity,and difficulty in parameter optimization in transformer fault diagnosis,an improved variable prediction model for transformer fault diagnosis was proposed.It was analyzed that the combination of variable prediction model and cuckoo search algorithm could solve the small sample and nonlinear problems,but the later convergence speed was slow,the stability was poor,the convergence accuracy was not high,and it was easy to fall into the local minimum problem.Mutation operation was introduced to the position update of the general cuckoo search algorithm to improve the diversity of the solution.The dynamic step size and dynamic discovery probability were introduced to improve the stability and convergence efficiency of the solution.By designing a transformer fault diagnosis experiment based on oil chromatographic data,it is verified that the combined method and its improved algorithm proposed have higher recognition efficiency than other optimization algorithms,and the average recognition accuracy is as high as 97%.