摘要
为了解决粒子群算法的早熟收敛问题和BP神经网络梯度下降训练法收敛速度慢、容易陷入局部极小值的问题,将免疫学中的克隆、变异理论用于粒子群算法的优化,建立免疫粒子群算法并给出算法步骤及免疫粒子群算法训练BP神经网络的步骤,将其应用到电力变压器的故障诊断中.仿真实验证明所提出的方法对变压器故障的诊断准确率可达95%以上,能够满足工程应用的需要.
In order to solve the premature convergence of PSO and slow convergence speed and easily getting into local dinky value of gradient descent algorithm, an immune particle swarm optimization （ IPSO ） algorithm was proposed, which used clone and hypermutation of immune theory to optimize particle swarm optimization （PSO） algorithm. The steps of BP Neural Network trained by IPSO and BP Neural Network trained by IPSO was established to use fault diagnosis of power transformer, the accuracy is higher than 95% after experimental verification and it can meet the requirements of engineering application.
出处
《北华大学学报（自然科学版）》
CAS
2014年第2期269-273,共5页
Journal of Beihua University（Natural Science）
基金
吉林省教育厅科学技术研究项目(2013178)
