针对神经网络结构和参数设计的随机性及依赖于人们经验的缺陷,提出一种改进的遗传优化神经网络BP算法(简称IGA-BP算法),算法首先利用改进遗传算法对神经网络结构和所有参数进行全面的进化设计,然后在进化神经网络中用训练样本再次寻优。将IGA-BP算法应用于电力变压器故障诊断之中,实例表明该算法在收敛速度和精度方面均优于传统BP算法,能实现对电力变压器故障的快速、准确的诊断。 In view of the stochastic nature of neural network structure and parameter design and the defects that depend on people’s experience, this paper proposes an improved genetic algorithm (BP) algorithm based on genetic algorithm (IGA-BP). The algorithm firstly uses improved genetic algorithm to optimize the neural network structure and all the parameters Conduct a comprehensive evolutionary design, and then use training samples again in evolutionary neural network optimization. The application of IGA-BP algorithm in power transformer fault diagnosis shows that the proposed algorithm is superior to traditional BP algorithm in terms of convergence speed and accuracy, and can quickly and accurately diagnose power transformer faults.