【摘要】 为提高故障诊断系统的容错能力,提出了将故障信息受随机因素畸变的扩展故障样本集引入神经网络(neural network,NN)的容错训练,以提高NN的容错性能,通过基于蚁群优化算法(ant colony optimization algorithm,ACOA)和遗传算法(genetic algorithm,GA)构造2种优化NN,用于高压输电线系统和配电网故障诊断,并进行容错性能的评估．仿真测试表明,基于ACOA法诊断模型的容错性能都要优于广泛应用GA的诊断模型,分别提高5．91%和4．95%．ACOA优化NN不仅具有较好的泛化能力,且具有快的收敛速率．更多还原

【Abstract】 In order to enhance the fault-tolerance performance(FTP)of the fault diagnosis system,the extend- ed fault samples considering the distortion of fault information because of stochastic factors are used in the fault- tolerance training of neural network(NN).Two kinds of optimized NN are constructed based on the ant colony optimization algorithm(ACOA)and genetic algorithm(GA)and are used in the fault diagnosis of transmission and distribution systems,and their FTP is assessed.The simulation results show that the FIT of ACOA-NN is superior to that of GA-NN though the structure of models and the distribution of sample space are different.It can enhance 5.91% and 4.95% respectively.ACOA-NN possesses not only excellent generalization ability but quick convergence rate.更多还原