基于广义生长-剪枝径向基函数神经网络的谐波源建模  

GENERALIZED GROWING AND PRUNING RBF NEURAL NETWORK BASED HARMONIC SOURCE MODELING

作　　者：占勇[1] 程浩忠[1] 葛乃成[1] 黄广兵[2] 

机构地区：[1]上海交通大学电气工程系,上海市闵行区200240 [2]南洋理工大学,新加坡639798

出　　处：《中国电机工程学报》2005年第16期42-46,共5页Proceedings of the CSEE

基　　金：高等学校优秀青年教师教学科研奖励计划项目

摘　　要：采用广义生长-剪枝RBF神经网络建立稳态频域的谐波源模型。在该模型中,各次谐波电流的幅值和相角与各次谐波电压的幅值和相角以及负荷特征参数的非线性映射关系通过一种新颖的广义生长?剪枝RBF网络进行建模。该网络的学习算法是串行的,可以进行动态建模。算例计算表明,该模型具有训练时间少、精度高、可动态建模等优点。A novel generalized growing and pruning RBF neural network （GGAP-RBF） is proposed for modeling nonlinear electric loads in steady-state frequency-domain. In the model, the nonlinearity mapping between harmonic voltages and harmonics currents is established by GGAP-RBE GGAP-RBF is a sequential learning algorithm so that model can be established dynamically. Numerical results show that the proposed method, having the characteristics of short training time, high precision and dynamic modeling, is an effective technique for building up harmonic source model.

关 键 词：电力系统 谐波潮流 谐波源模型 径向基函数 神经网络 串行学习 广义生长-剪枝径向基函数 

分 类 号：TM711[电气工程—电力系统及自动化]