改进粒子群优化神经网络的变压器故障诊断

摘要

在分析传统误差反向传播(BP)算法和标准粒子群优化(PSO)算法的特征及其问题基础上，提出一种改进粒子群优化(IPSO)算法和改进BP(IBP)算法，建立基于IPSO-IBP混合算法的电力变压器神经网络故障诊断模型。通过85组训练样本和16组测试样本的仿真对比分析，该方法能够实现电力变压器不同故障的有效诊断，提高了电力变压器故障模式的识别能力及故障诊断准确率。%Based on analysis characteristics and problems of traditional error back propagation (BP) algorithm and standard particle swarm optimization (PSO) algorithm, this paper proposed an improved particle swarm optimization (IPSO) algorithm and an improved BP (IBP) algorithm, and established a model of neural network for power transformer fault diagnosis based on IPSO-IBP hybrid algorithm. By simula-tion comparison and analysis of 85 groups training samples and 16 groups test samples, this method can realize the effective diagnosis for different power transformer faults and improve the recognition ability of power transformer fault mode with high accuracy.