Physics-Informed Data-Driven Control of Wave Energy Converters: Predictive Wave-Force Modeling and Reinforcement-Learned Virtual Inertia for Stable Grid Integration

This project addresses a key challenge in connecting ocean wave energy systems to the electric grid: converting highly variable wave-driven motion into electrical power that is reliable, controllable, and compatible with grid requirements. Wave energy represents a significant electricity resource along U.S. coastlines, but its practical use requires advances in forecasting, energy storage, power electronics, and control. The project will bring transformative change by developing a predictive and coordinated control framework that enables an array of wave energy converters to operate cooperatively while using advanced hybrid energy storage to smooth power variations and support stable grid operation. This will be achieved by combining physics-informed wave and power forecasting, intelligent energy management, coordinated control of multiple devices, and grid-support functions in one integrated framework. The intellectual merit of the project includes new methods for forecast-aware control of wave energy systems, coordinated operation of multiple power converters and energy storage devices, and adaptive grid-forming control that uses available wave-power information to improve voltage and frequency stability. The broader impacts of the project include advancing wave energy as a reliable electricity resource, supporting energy resilience for coastal, islanded, and remote communities, and creating educational and research opportunities in marine energy, power electronics, machine learning, and power-system control. The project will study a grid-connected offshore energy system composed of an array of direct-drive linear-generator wave energy conversion systems, a hybrid supercapacitor/undersea energy storage system, and power electronic converters that interface the wave-energy array and energy storage system with the grid. The research will develop physics-informed machine learning models to forecast wave excitation forces and short-term power availability from sparse ocean measurements. These forecasts will be used by multi-agent reinforcement learning methods, in which individual wave energy converters act as cooperative agents that coordinate their converter actions to increase total array power production while respecting device and grid constraints. An integrative sizing/real-time energy management technique will be developed for the hybrid energy storage system by using reinforcement learning augmented by physics-informed machine learning forecasts, enabling cost-optimal sizing and robust operation under wave-energy uncertainty. The project will also develop a dc-link-aware grid-forming converter controller that adaptively tunes virtual inertia and damping to improve stability under changing sea states and grid conditions. The proposed framework will be validated through hardware-in-the-loop studies that combine buoy-derived wave data, real-time models of the wave energy array and energy storage system, and embedded implementation of the control algorithms. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.