Nonlinear Inverse Modeling of Synchronous Generator based on Improved Resource Allocating Networks

摘要

Synchronous generator inverse modeling is the basis of inverse control, analysis and design in the power systems. According to the strong nonlinear characteristics of synchronous generator, an inverse modeling method based on improved resource allocating networks (RAN) is presented in this paper. In view of the existing disadvantages of traditional resource allocating networks, a design method for RAN based on rough set theory (RST) and orthogonal least square (OLS) was proposed. With the advantage of finding useful and minimal hidden patterns in data, RST is first applied to intelligent data analysis for extracting typical characteristics and underlying rules from the training samples, followed by a second stage mapping the condition components of the rules into network centers candidate. And then OLS algorithm was used to select best centers as the hidden layer nodes with novelty criterion. The simulation results showed that the presented inverse modeling method has the advantages of simple network structure, high convergence rate and better generalization ability, etc.