基于BP神经网络和专家系统的变电站报警信息处理系统

对BP神经网络的容错性进行了研究，将网络的容错能力与测试样本所形成的模糊区的 大小相对应，通过消除模糊区来提高网络的容错能力。针对南昌500 kV变电站自动化系统， 开发了变电站的实时故障诊断系统。该系统以3层前向BP网络作为故障诊断的核心部分，以 开关动作信息、保护动作信息等作为人工神经网络的输入。同时结合专家系统，利用其推理 判断能力，对变电站运行方式进行识别，并对神经网络的某些输出结果进行必要的修正。

SUBSTATION ALARM MESSAGE DEALING SYSTEM BASED ON ANN AND ES

The fault- tolerance performance of BP neural network NN is studied firstly.The fault- tolerance perform ance is corresponding to the fuzzy area formed by detecting sam ples,and the fault- tolerance performance is enhanced by eliminating fuzzy area.A substation real- time fault diagnosis system is developed on the basis of Nanchang 5 0 0 k V automation substation.Triple layers BP NN is the core of fault diagnosis system in which inputs include the alarm ing information of tripped breaker,protection inform ation and so on.Because of its inference capability,expert system is used in the system to recognize substation operational modes and m odify som e output of ANN.