Recurrent Neural Network (RNN) Based Algorithm in Multi-Level Control of an Islanded DC Microgrid Connected to Variable Communication Networks

Abstract

The utilization of microgrids (MGs) and energy communities has surged in recent years, enabling numerous stakeholders to participate in the power distribution system. Unfortunately, communication infrastructure failures in rural networks has increased the operational blind spots. In the event of a failure, information sharing may be delayed. To address this problem in a multi-feeder MG, a resilient control approach utilizing RNN-based control has been proposed to manage load sharing and voltage regulation during communication delays. A recurrent neural network (RNN) is utilized to optimize the control scheme for the operating direction for each distributed generating point. Traditional control may become unstable during information breaks, but the proposed RNN method improves connectivity during such occurrences. Through this analysis, the research showcased the efficacy of the proposed RNN technique in precisely distributing the load and regulating voltage, particularly during information breaks. The study also confirmed that the RNN strategy is more efficient than conventional control methods. The RNN approach creates a resilient and stable network to information failures, and the study's findings were derived from the detailed mathematical analysis of DC microgrid (DC MG) load conditions and radial networks' uncertain line characteristics.