改进萤火虫算法与小波神经网络相结合的变压器故障诊断

为了解决变压器故障诊断中诊断效率低的问题,本文对萤火虫算法(FA)进行了改进,并与小波神经网络(WNN)相结合应用于变压器故障诊断中。小波神经网络结构简单,预测精度高,收敛速度快,但是网络参数不好选择,易陷入局部最优。本文结合混沌算法、粒子群算法、可变步长的思想来改进萤火虫算法,用于优化小波神经网络的参数,再将处理后的数据带入神经网络中进行训练与诊断。实验结果表明,该算法与BP神经网络、支持向量机、小波神经网络、遗传算法改进的小波神经网络和粒子群算法改进的小波神经网络相比诊断正确率均有所提高。

Transformer Fault Diagnosis Based on Improved Firefly Algorithm Combined with Wavelet Neural Network

In order to solve the problem of low diagnostic efficiency in transformer fault diagnosis, the Firefly algorithm (FA) is improved and combined with wavelet neural network (WNN) for transformer fault diagnosis. The wavelet neural network has a simple structure, high prediction accuracy and fast convergence speed, but the network parameters are not well chosen, and it is easy to fall into local optimum. This paper combines chaos algorithm, particle swarm algorithm and variable step size to improve the firefly algorithm, which is used to optimize the parameters of wavelet neural network, and then bring the processed data into the neural network for training and diagnosis. The experimental results show that the proposed algorithm has improved diagnostic accuracy compared with BP neural network, support vector machine, wavelet neural network, genetic algorithm improved wavelet neural network and particle swarm optimization improved wavelet neural network.