Adaptive CSOGI-Based synchronization using neural networks for distorted Grid conditions

Phase-Locked Loops (PLLs) are widely used in grid-connected systems for accurate phase and frequency estimation. Their performance greatly depends upon the quality of the preconditioning stage, particularly in non-ideal grid conditions. For improved robustness, a double cascaded second-order generalized integrator (CSOGI) is often utilized to obtain in-phase and quadrature clean signals from the input one. Though effective, the conventional CSOGI is founded on fixed design parameters, thereafter with limited capability in handling of time-varying disturbances. An adaptive synchronization method is introduced in this work, through which the CSOGI filtering gain is adaptively controlled in real time by an artificial neural network (ANN). The design CSOGI-ANN structure supports better dynamic response and improved harmonic rejection.