采用NGA优化的新型BP神经网络模型及诊断应用

New Optimized BP Neural Network Model by Applying NGA and Its Diagnosis Application

针对传统BP神经网络模型收敛速度慢、易陷入局部极小点、网络结构不稳定等缺陷,提出了一种小生境遗传算法优化的BP神经网络模型。该模型充分利用小生境遗传算法的搜索能力和BP神经网络的非线性映射和学习联想能力,通过小生境遗传算法的选择、交叉、变异及小生境淘汰等操作,优化BP神经网络的初始权值和阈值,并采用BP算法对网络进行训练。有效解决了网络初值不合理的问题,提高了网络收敛速度、稳定性。最后结合变压器故障诊断实例。在Matlab7.0平台上进行仿真实验。实验结果证明:与传统方法相比,该模型具有很强的可行性和有效性。

According to shortcomings of BP neural network model, such as slower convergence speed, entrap ment in local optimum, unstable network structure etc. , an improved BP neural network model based on Niche Ge netic Algorithm （NGABP） was presented. The proposed model firstly makes full use of the global searching ability of genetic algorithm and the nonlinear reflection ability and the association learning ability of BP neural network to optimize the initial connection weights and thresholds of the neural network by means of selection operation, cross over operation, mutation operation and niche pass, and then adopts BP algorithm to train network, which can effec tively solve the questions of BP network about reasonable initial value and network misconvergence, and improve the convergence speed and the stability of network. Finally, the simulation experiment is carried out to diagnose transformer faults in Matlab7.0 platform. The experiment results show that the model is more feasible and effective than the traditional methods.