Microgrid Control Using Reinforcement Learning Operating in Grid Tied and Islanded Modes
Abstract
The growing demand for Distributed Energy Resources (DERs) has caused increased work in the development of Microgrids. It has been shown to have many economic benefits for utilities and consumers, reduce the amount of greenhouse gas emissions, allow user monitoring and controlling with local utilities, reduce strain during peak load events, and work in standalone operations. Unfortunately, Microgrids can pose threats to utilities and consumers by voltage instabilities, injection of harmful harmonics that can reduce Power Quality (PQ), and dependency on acceptable environmental conditions. These challenges can be resolved by implementing controllers that can monitor and control for voltage stability, improve PQ, and optimize the generation of power from the DERs. This work developed a Reinforcement Learning (RL) Smart Controller to monitor and control an 8 MVA Microgrid system consisting of a Photovoltaic (PV) System, Battery Energy Storage System (BESS), Wind Energy Conversion System, and Synchronous Generator. The RL agent was trained using the Proximal-Policy Optimization algorithm to output controls for Maximum Power-Point Tracking (MPPT) of the PV system, charging/discharging of the BESS, and adaptively tune the Proportional-Integral (PI) controllers for each DER. To prove the effectiveness of the proposed controller, it was compared to standard controller implementations using non-adaptive PI controllers and Perturb and Observe MPPT algorithm for different fault and load scenarios using MATLAB Simulink. The proposed controller performed better than the standard controller in terms of voltage stability, and PQ. It also showed improvement of the Total Harmonic Distortion up to 2% and increase efficiency by 7%.
Subject Area
Electrical engineering|Computer Engineering|Computer science|Energy
Recommended Citation
Richard Wiencek,
"Microgrid Control Using Reinforcement Learning Operating in Grid Tied and Islanded Modes"
(2024).
ETD Collection for Tennessee State University.
Paper AAI31140703.
https://digitalscholarship.tnstate.edu/dissertations/AAI31140703