An optimized edge-assisted control system for low inertia photovoltaic microgrids in a cloud coordinating environment

Abstract

The increasing integration of inverter-based resources such as solar photovoltaic is reshaping modern power systems, leading to a significant reduction in system inertia. This transition introduces challenges to both frequency and voltage stability, especially during faults and dynamic loads. To overcome these limitations, this paper proposes an optimized framework based on edge computing and a cloud-based coordination environment for virtual inertia support. Unlike conventional droop control methods, the proposed control strategy involves dynamically optimizing the control gains to minimize voltage and frequency fluctuations. Specifically, the proportional-integral-derivative (PID) parameters for both the frequency and voltage control loops within a droop control system are adjusted using a multi-objective particle swarm optimization (MOPSO) approach. The goal of this tuning is to enhance the system's ability to support inertia and improve its overall transient response characteristics. The control management system combines a local cloud node for real-time data exchange and oversight, allowing the optimizer to adjust system parameters to emulate the inertia characteristics of traditional synchronous machines. The controller is deployed on an embedded system, connected to a Synology network-attached storage (NAS) through the Internet of Things (IoT) for edge processing and data synchronization. A controller-in-the-loop (CIL) experimental setup has been developed to test and validate system performance under various operating conditions. This proposed approach significantly reduces voltage overshoot and keeps frequency deviation within a narrow range. Specifically, the hybrid algorithm achieves a 47% reduction in post-fault voltage overshoot and confines frequency deviation to within ±0.015 Hz. Additionally, it achieves a settling time of under 200 ms during severe disturbances. The timing performance demonstrates a latency of approximately 1 s per cycle, without compromising frequency or voltage stability.