GA-NN模型在输配电系统中诊断容错性能的评估  

Assessment on fault-tolerance performance using GA-NN model for fault diagnosis in transmission and distribution systems

作　　者：张智晟[1] 孙雅明[1] 

机构地区：[1]天津大学电气信息与自动化工程学院,天津300072

出　　处：《系统工程学报》2003年第5期470-474,共5页Journal of Systems Engineering

基　　金：国家自然科学基金资助项目(59877016).

摘　　要：在基于神经网络NN模型的高压输电线系统故障诊断和配网故障定位的两个目标的研究中,用遗传算法进行NN结构优化、权重和结构双优化两种模式来研究诊断系统的容错性能.通过仿真测试,并与BP NN模型相比较,证明可克服NN陷入局部最小、加快收敛和输出解空间重构,使故障诊断的容错性能和故障定位性能相应提高17.34%和13.86%.In the faulttolerance performance study on two objects for the fault diagnosis (FD) of high voltage transmission line system (HVTLS) and the fault section diagnosis (FSD) of distribution networks (DN) based on neural network, this paper presents two modes-the optimization of NN structure and the entire optimization of NN architecture and weights based on GA algorithms. It is tested that, the two modes can avoid NN to immerse in the local minimal points, accelerate the convergence rate, and reconstruct the output solving space to improve the corresponding faulttolerance performance 17.34% for HVTLS_FD and 13.86% for FSD through the comparison with conventional BPNN model. 

关 键 词：输配电系统 故障诊断 容错性能 评估 GA-NN模型 神经网络 遗传算法 配电网 供电可靠性 

分 类 号：TM732[电气工程—电力系统及自动化] TP183[自动化与计算机技术—控制理论与控制工程]