Low-Inertia Microgrid Synchronization Using Data-Driven Digital Twins

Abstract

We introduce data-driven and scalable digital twins (DTs) and decentralized observer-based control (DOBC) to enhance inverter synchronization in low-inertia microgrids. The proposed DT, serving as cyber-physical replicas, enables real-time monitoring and data-driven control. We employed the Kuramoto model as a reduced-order dynamic representation of the low inertia inverter-based microgrid. Additionally, we used finite state machines (FSM) to digitally integrate the states and operating modes of virtual oscillator controls (VOC) inverters and microgrid dynamics. We implemented generative adversarial imputation nets (GAIN) to impute missing states in real-time to address potential inconsistencies in data acquisition. For inverter synchronization and minimizing the control efforts in the presence of the grid topology changes and interruptions, we applied Gramian localized approximation (GLA) and DOBC. These techniques helped us identify an optimal subset of inverters for control. The efficacy of this approach was validated through several scenarios for normal operation and fault isolation cases. The proposed DT model with DOBC significantly reduces synchronization time to under 9 seconds for average to large topology connections, compared to conventional phase-locked loop (PLL) methods requiring around 3 minutes.