Neural Networks Based Frequency-Locked Loop for Grid Synchronization Under Unbalanced and Distorted Conditions

摘要

Precise knowledge of amplitude, phase and frequency of the grid voltages is essential for grid synchronization of distributed generation units. Nevertheless, grid parameters estimation becomes complex under nonideal conditions. Therefore, a novel frequency adaptive neural network-based frequency-locked loop (FANN-FLL) strategy for unbalanced and distorted conditions is proposed in this paper. The FANN-FLL is based on a FANN configured as a quadrature signal generator, allowing fast and accurate fundamental symmetrical components estimation. To make the system frequency adaptive, the FANN weight vector is exploited in a new frequency estimation method. Simulation tests under highly unbalanced and distorted grid conditions are carried out. Obtained results demonstrated the superiority of the FANN-FLL compared with the decoupled double synchronous reference frame phase-locked loop.