Offshore wind power forecasting via trend-aware just-in-time learning with nearest neighbors

Accurate offshore wind power forecasting is vital for secure grid operation and cost-effective system dispatch but remains challenging due to the high volatility and non-stationarity of offshore environments. Existing forecasting models often rely on offline training and external meteorological data, limiting their adaptability to rapid variations in wind power. This study proposes a trend-aware just-in-time learning (tJITL) framework that integrates trend similarity into an online autor-egressive exogenous (ARX) model. The method dynamically constructs local models online by selecting trend-consistent samples from historical data, thereby capturing transient dynamics without the need for model retraining or external variables. Experimental results demonstrate that the proposed tJITL framework provides a reliable and data-efficient solution for online offshore wind power forecasting, with strong potential for application in intelligent power system operations.