Seamless Grid-Following to Grid-Forming Transition During Nonlinear Load Conditions and Unintentional Islanding

Abstract

Unintentional islanding of inverter-based resources (IBRs) in low-voltage distribution networks introduces severe voltage, frequency, and power-quality challenges, particularly when nonlinear loads are present. Conventional grid-following (GFL) inverters depend on an external grid reference for synchronization and cannot autonomously regulate voltage and frequency upon islanding, while existing grid-forming (GFM) transition strategies frequently produce large transient oscillations during mode transfer. This article proposes a seamless GFL–GFM transition strategy based on an enhanced disturbance observer (DO) integrated within an indirect power control framework. The enhanced DO explicitly estimates and compensates for nonlinear load disturbances in real time, enabling smooth control authority transfer without phase discontinuities or transient overshoots. A coordinated feedback-form proportional-integral (PI) realization ensures that both GFL and GFM controllers remain internally synchronized prior to and during mode switching, eliminating the need for auxiliary synchronization controllers. The proposed method is validated through simulation and hardware-in-the-loop (HIL) experiments on a TMS320F28379D digital signal processor (DSP) interfaced with an OPAL-RT real-time simulator.Experimental results across three nonlinear load cases, including a three phase six-pulse rectifier with $LC$ filter, a 12-pulse rectifier with $LC$ filter, and an unfiltered six-pulse rectifier, demonstrate voltage total harmonic distortion (THD) of 3.5%, 2.8%, and 4.6%, respectively, all remaining below the 5% IEEE 519-2022 limit.Under linear loading, the voltage THD is further reduced to 1.7%. Compared with existing seamless transition strategies, the proposed method achieves significantly lower THD, reduced settling time, smaller transient overshoot, and attenuated power oscillations, with only two controller switching states and no dependence on communication or grid-status estimation.