基于分布估计算法和遗传算法融合的神经网络故障诊断模型研究

本文构造了基于分布估计算法(Estimation of Distribution Algorithm,EDA)和遗传算法(GeneticAlgorithm,GA)融合的神经网络(Neural Network,NN)故障诊断模型。传统的GA看作是对生物进化"微观"层面上的模拟,则EDA是对生物进化"宏观"层面上的建模,是一种全新的进化模式。EDA与GA融合的实质是在解空间"宏观"和"微观"两个层面进行寻优,可克服NN陷入局部最小,提高NN的泛化能力,使故障诊断的容错性能得到有效改善。将该模型用于高压输电线系统的故障诊断,并作容错性能的评估。由仿真测试表明,研究模型的容错性能要优于传统的BP-NN模型和单纯GA优化NN模型。因此,新诊断模型是有一定的理论和实用价值的。

Study of fault diagnosis model based on neural network using estimation of distribution algorithm combined with genetic algorithm

A fault diagnosis model using NN(neural network) based on EDA(estimation of distribution algorithm) combined with GA(Genetic Algorithm) is constructed in this paper.The traditional GA is regarded as the simulation of biological evolution from microscopic level,while EDA is from macroscopic level.EDA is a kind of novel evolution mode.The essence of combining EDA with GA is to search the optimal solution from microscopic and macroscopic level,meanwhile to avoid NN to immerse in the local minimal points and improve generalization ability,so fault-tolerance performance of fault diagnosis model can be effectively improved.The presented model is used as the fault diagnosis in high voltage transmission line system,and their fault-tolerance performance is assessed.Through the simulation and test,it shows that the fault-tolerance performance of researched model is superior to that of the diagnosis model corresponding BP-NN and GANN.So researched diagnosis model possess theoretical and practical value.