Intelligent Renewable Energy Agent-Based Distributed Control Design for Frequency Regulation and Economic Dispatch

Abstract

The Distributed Renewable Energy Sources (DRESs) integrate hybrid microgrid and prosumer activities that constitute a dynamic system characterized by unknown network parameters. The dynamic system faces challenges, such as intermittent power supply due to low inertia, renewable intermittence, plug-and-play prosumer activities, network topology variations, and a lack of constraint handling. These complexities pose significant issues in designing effective control for frequency regulation and consensus-based economic load dispatch (ELD) within DRES to meet varying load demands. To address the above challenges, this research employs a machine learning-based distributed multiagent consensus design that offers a rapid and robust approach, mitigating the limitations associated with the Distributed Average Integral (DAI) control design. The proposed multiagent scheme empowers the successful implementation of ELD and frequency regulation, accommodating the intermittent DRES, diverse network topologies, and the dynamic plug-and-play activities of prosumers. Moreover, an optimization-based DAI tuning model is introduced to overcome tuning limitations. Intelligent renewable energy agents are trained through machine learning-based regression models that use root mean square error metrics for performance evaluations. The intelligent agents employ DAI control to overcome inherent limitations. The effectiveness of the machine learning-based DAI is thoroughly evaluated using the DRES-based IEEE 14-bus hybrid microgrid system. The quantitative results prove its efficacy in addressing the complex challenges of integrated microgrid dynamics.