提出一种基于改进遗传算法的彻底进化神经网络算法,并且命名为IGA-BP算法。在IGA-BP算法中,首先用遗传算法来全面解决和设计神经网络中的结构以及所有训练参数,然后用训练的样本来找到最优解。该网络的缺点是,他的结构和神经训练参数是随机的,或由一个人的经验来解决的。IGA-BP算法可以用于电力变压器故障的诊断。在神经网络的基础上可以建立电力变压器故障诊断的模型。的例证结果表明,该算法在速度的收敛精度方面优于传统的BP算法。可以用该算法实现对电力变压器故障的快速且准确的判断。 A completely evolutionary neural network algorithm based on improved genetic algorithm is proposed and named as IGA-BP algorithm. In IGA-BP algorithm, genetic algorithm is firstly used to solve and design the structure of neural network and all the training parameters in an all-round way, and then use the trained samples to find the optimal solution. The disadvantage of this network is that his structural and neuro-training parameters are random or solved by one’s own experience. IGA-BP algorithm can be used for power transformer fault diagnosis. On the basis of neural network, a model of power transformer fault diagnosis can be established. The results show that the proposed algorithm outperforms the traditional BP algorithm in terms of speed convergence accuracy. This algorithm can be used to achieve fast and accurate judgment of power transformer faults.