Full-Parameter Identification Model Based on Back Propagation Algorithm for Brushless Doubly Fed Induction Generator

Abstract

Brushless doubly fed induction generator (BDFIG) has great potential due to its high reliability and low maintenance cost. To achieve high-performance modeling and control, BDFIG resistances and inductances are necessary. However, the existing identification methods either required professional structure knowledge or special excitations and setups, or only estimated part of parameters. Thus, this letter proposes a multilayer full-parameter identification model based on the back-propagation (BP) algorithm for BDFIG, which is constructed with electric quantities as nodes and parameters as adjustable weights, and utilizes the electric quantities measured from regular operations as data. According to the fitting error obtained by comparing the model outputs with the easily measured references, the BP algorithm is applied to update the weights until the error is sufficiently small. Then, all resistances and inductances can be extracted directly from the weights. Such an identification methodology can be easily embedded into existing BDFIG systems. The simulations and experiments verify its feasibility and accuracy.