Science News

【Achievement】 2026/08/06: Marine Energy — Hybrid Transformer AI enables operational forecasting of turbine-scale Kuroshio power generation at optimal sites.

Congratulations to our Research Fellow, Dr. Chau-Ron Wu, whose study on using AI to forecast the efficiency of Kuroshio turbine sites has been published in the journal Energy Conversion and Management!
To address the bottleneck of imprecise site-scale ocean current forecasting, the research team developed an observation-constrained hybrid deep-learning (HyDL) framework tailored for western boundary currents such as the Kuroshio. This breakthrough enables accurate predictions of Kuroshio current velocity, power conversion, and practical industrial applications.
By integrating Transformer, statistical correction, and real-time learning, the framework achieves an approximate 30% reduction in RMSE compared to existing operational systems. It accurately converts ocean currents into turbine-scale power output under real-world operational constraints, reducing generation uncertainty and system-level costs. Looking ahead, this work integrates Kuroshio power generation, AI forecasting, and Taiwan's energy management system to offer a scalable blueprint that accelerates the commercial deployment of ocean current energy in Taiwan and worldwide.

Link: https://doi.org/10.1016/j.ecmx.2026.101923
Wu, C. R., & Chang, Y. C. (2026). Hybrid Transformer AI enables operational forecasting of turbine-scale Kuroshio power generation at optimal sites. Energy Conversion and Management: X, 101923.