基于模糊神经网络的母线充电保护研究  

Study on Bus Charging Protection Based on Fuzzy Neural Network

作　　者：雷宇[1] 陈少华[1] 桂存兵[1] 马碧燕[1] 

机构地区：[1]广东工业大学,广东广州510090

出　　处：《现代电力》2007年第1期21-25,共5页Modern Electric Power

基　　金：广东省科技厅资助项目(040094)

摘　　要：针对当前母线充电试验时继电保护中存在的一些问题,提出了一种智能型的母线充电保护新方法。在分析母线正常充电和故障元件充电的基础上,找到了正确快速地识别两种情况的特征量,结合模糊神经网络这一新型的人工智能技术,综合利用母线充电时的多种电气量分别提取形成网络的特征输入量,并采用Simpson模糊极小-极大神经网络来形成区分母线正常充电和故障元件充电的模糊模式识别器。利用EMTP程序仿真来获取不同系统参数和各种不同故障情况下模糊神经网络训练和测试所需要的大量样本,通过对模糊神经网络的训练,使网络具备了很强的故障识别能力和较强的泛化能力,结果表明,训练后的网络能快速准确地区分母线在各种运行工况下的正常充电和故障元件充电,从而验证了该方法的有效性。Aimed at the problems of relay protection during charging test of the buses, a new method of intelligent buscharging protection is presented in this paper. After analyzing normal bus charging and fault component charging, the characteristic parameters are found to identify them based on the new artificial intelligence technique of fuzzy neural network. We make full use of many electrical values of bus charging and develop several corresponding characteristic inputs, and use Simpson＇s fuzzy min-max neural network to develop the fuzzy pattern classifier which can distinguish normal bus charging from fault element charging~ EMTP simulation programs are applied to obtain plenty of training and testing patterns in dissimilar system parameters and different fault condition. After abundant training, the fuzzy neural network has powerful fault discrimination and generalization ability. The training and testing results illustrate that the trained network can discriminate normal charging and fault charging under all kinds of operating conditions. It proves that this method is effective.

关 键 词：母线 充电保护 糊糊极小-极大神经网络 超盒 人工智能 

分 类 号：TM771[电气工程—电力系统及自动化]