Optimal bandwidth-based pseudo-centralized droop control mechanism for grid-forming microgrids using tri-layered neural network with real-time feasibility

In modern power systems, particularly in microgrid architectures, controlling and managing power is critical for ensuring stability and efficient operation, especially in grid-forming mode (GFM). This paper proposes an optimal limited bandwidth-based droop control strategy, leveraging a tri-layered artificial neural network (TLANN) to enhance the robustness of power sharing during the islanded operation of a microgrid. The proposed method calculates the optimal bandwidth of 10.8 kSa/s for the communication channel of the enhanced droop control mechanism based on the prediction performance analysis acquired through TLANN. At this low bandwidth level, communication among inverters can be established through a narrowband power line communication channel (NB-PLC) without incurring additional costs. Considering the delay associated with the Tx-Rx process, the performance of the proposed optimal bandwidth-based droop control mechanism has been evaluated using a hardware-in-the-loop (HIL) test-bench system with the dSPACE DS-1103 rapid control prototype (RCP). The real-time test results demonstrate a high feasibility percentage of 95.62%.