模糊神经网络负荷模型的内插外推能力

为了分析综合负荷模型的实用性,研究了综合负荷模型的内差外推能力。通过分析模糊神经网络综合负荷模型建模数据的特征,确定模糊模型的初始结构;通过模糊神经网络对建模数据的学习,获取模糊推理规则并调整隶属度函数的参数,进而辨识出模糊模型并调整隶属度函数;将电压、有功和无功作为模糊神经网络的广义模型输入,对综合负荷进行解耦建模。在综合负荷构成成分相同的3组数据中,用其中1组建模数据训练得出模糊模型的结构和参数去拟合其它2组建模实测数据,获得良好的拟合效果,从而验证了模糊神经网络综合负荷模型的良好的内插外推能力和收敛性。该特性对于综合负荷模型的实用化具有重要的指导意义。

Interpolation and Extrapolation Ability of Fuzzy Neural Network Load Modeling

A new load modeling method based on subtractive clustering and fuzzy neural network is proposed.The features of input and output data for modeling are analyzed by subtractive clustering to decide the number of fuzzy rules and membership function for presupposition conditions.The initial structure of fuzzy model is determined.By means of learning fuzzy neural network to modeling data,fuzzy reasoning rules are obtained and the parameters of membership function are adjusted,accordingly,the fuzzy model is identified and the membership function is adjusted,finally the load modeling is completed.In three groups of data with the same load constitution,one group of them are used for training and the structure of fuzzy model and parameters are obtained,and the obtained model and parameters are fitted with other two groups of data,thus the ability of interpolation and extrapolation of load model based on the fuzzy neural network is validated.The results of practical modeling examples show that the proposed model possesses high fitting accuracy and high convergence speed as well as good ability of interpolation and extrapolation.