Neural Networks for Stable Control of Nonlinear DFIG in Wind Power Systems

Abstract

This paper presents an Artificial Neural Network (ANN) based Direct Power Control (DPC) strategy for controlling power flow, and synchronizing Double Fed Induction Generator (DFIG) with grid and Voltage Oriented Control (VOC). In order to cope with the complex calculations required in DPC, the proposed ANN system employs the individual training strategy with fixed-weight and supervised models. The ANN controller is divided into five subnets: 1) real and reactive power measurement sub-net (fixed-weight) with dynamic neurons; 2) reference real and reactive calculation sub-net (fixed-weight) with square neurons; 3) reference stator current calculation sub-net (supervised) with log-sigmoid neurons and tan-sigmoid neurons; 4) reference rotor current calculation sub-net (fixed-weight) with recurrent neurons; and 5) reference rotor voltage calculation sub-net (fixed-weight or supervised). The results obtained demonstrate the feasibility of ANN–DPC. The proposed ANN-based scheme incurs much shorter execution times and, hence, the errors caused by control time delays are minimized.