Physics Informed Learning-Based Frequency Regulation and Virtual Inertia Control for Renewable Energy Generators in Power Systems

Integrating renewable energy systems (RESs) such as polar photovoltaic (PV), wind turbines, energy storage, and electric vehicles (EVs) into the power grid is a pivotal step toward sustainable, secure, and efficient energy. Transitioning to renewable energy is critical to addressing environmental challenges like climate change, reducing greenhouse gas emissions, ensuring reliable energy access for future generations, meeting global energy demand, and supporting economic development.

Off-grid renewable energy solutions can bring electricity to remote and undeserved areas, improving living standards and fostering economic growth. However, integrating renewable energy-based systems into the utility grid is challenging owing to the unpredictability of renewable energy sources and the lack of inertia of the power electronic converters (PECs) used for grid integration. Increasing integration of renewable energy generators reduces the overall grid inertia. Low-inertia power grids are susceptible to frequency instability and eventual system failure under severe contingencies.

This thesis proposed a learning-based predictive virtual inertia (LPVIC), a cutting-edge approach to addressing the challenges of low-inertia power grids with high renewable energy penetration. This method integrates machine learning (ML) techniques with predictive control algorithms to optimize virtual inertia responses of low-inertia power grids.

Specifically, the proposed method leverages the capability of the reinforcement learning algorithm in solving complex and high-dimensional tasks and the reliability of the physics-informed neural network (PINN) to develop a virtual inertia. This enables the integrated renewable energy generators to participate in grid frequency by providing rapid response to frequency deviations and compensating for lack of inertia of the PECs through active power adjustment. LPVIC offers dynamic, reliable and adaptive solutions for grid stability.