AI-Powered Wave Forecasting: Enhancing Maritime Safety and Coastal Resilience with LLMs

On April 24, 2025, researchers introduced Improving Significant Wave Height Prediction Using Chronos Models, detailing a novel approach to wave forecasting with a large language model (LLM)-enhanced temporal architecture. This study presents a computationally efficient framework that achieves faster inference speeds and improved accuracy across short-term and extended-range predictions, offering significant advancements for maritime safety and coastal resilience.

Conventional wave forecasting models struggle with computational efficiency and nonlinear dynamics. This study introduces Chronos, a large language model (LLM)-powered temporal architecture optimized for wave height prediction. By analyzing historical data from three marine zones in the Northwest Pacific, Chronos achieves significant improvements: 14.3% reduction in training time, 2.5x faster inference speed (0.575 MASE units), superior short-term forecasting (1-24h), extended-range predictive performance (1-120h), and zero-shot capability maintaining median performance against specialized models. This LLM-enhanced framework sets a new standard for computationally efficient wave prediction, offering transferable solutions for complex geophysical systems modeling.

The quest to harness wave energy as a sustainable power source has gained momentum in recent years, driven by the urgent need for renewable alternatives. Central to this endeavor is accurate wave forecasting, which is crucial for optimizing energy extraction and ensuring operational safety. Traditional methods have relied on physical models based on hydrodynamics and meteorology, but these approaches often struggle to capture the dynamic and unpredictable nature of ocean waves.

Conventional wave forecasting has depended heavily on numerical models that simulate wave behavior using complex equations. While effective in many scenarios, these models face challenges with real-time data processing and adaptability to rapidly changing conditions. This limitation is particularly significant in operational settings where timely and precise predictions are crucial for maximizing energy yield and minimizing risks.

In conclusion, machine learning is transforming the field of wave energy forecasting by offering powerful tools that enhance accuracy, reliability, and efficiency. As the technology advances, it promises to make wave energy a cornerstone of the global renewable energy landscape.