Abstract
Global demand for clean, reliable energy continues to accelerate, driving the need for innovative renewable strategies that maximize power density and stability. Co-locating offshore wind (OSW) turbines with wave energy converters (WECs) has been recognized as a promising approach to mitigate resource intermittency, smooth variability, and optimize shared infrastructure [1, 2, 3]. However, the planning, forecasting, and grid integration of such hybrid systems remain challenging.
This study introduces a novel deep learning framework to forecast hybrid wind-wave energy output and spatial yield potential across transatlantic regions. We analyze two distinct offshore sites: the Amrumbank West wind farm in the German North Sea and the Coastal Virginia Offshore Wind (CVOW) zone in the U.S. Mid-Atlantic. Prior work suggests that the U.S. Mid-Atlantic region exhibits strong seasonal complementarity between wind and wave resources, making it a promising candidate for hybrid deployment [4]. Historical wind and wave datasets from ERA5 [5], the NDBC buoy network, and the German Weather Service (DWD) are utilized to develop long short-term memory (LSTM) networks [6] – an architecture proven effective for time-series energy forecasting [7, 8, 9].
The LSTM models jointly predict hourly to daily hybrid power output by incorporating turbine specific power curves and wave energy calculations based on IEC 62600-101 standards [10]. We also generate spatially resolved yield maps to identify sub-regions with optimal hybrid potential within each lease area, providing a lightweight, data-driven alternative to computationally expensive optimization methods [11].
To extend utility beyond forecasting, we interface our model outputs with techno-economic tools such as the NREL System Advisor Model (SAM) [12] to estimate Levelized Cost of Energy (LCOE) and simulate investment scenarios. Additionally, the model’s grid relevance is assessed in the context of major transmission bottlenecks highlighted in the U.S. DOE’s Atlantic Transmission Study [13]. This enables early-stage evaluation of how spatial generation variability may influence grid integration strategies.
Our comparative approach across two climatic regimes illustrates the value of co-located hybrid energy systems and their ability to offer complementary seasonal output, especially in temperate regions. While final training and validation results are forthcoming, this framework lays the foundation for an AI-augmented toolchain supporting hybrid offshore planning, resource modeling, and grid-aware deployment.
The associated presentation from OREC/UMERC 2025 can be found here.