An adaptive framework based on multi-model data fusion for one-day-ahead wind power forecasting

This paper proposes a novel framework for one-day-ahead wind power forecasting based on information amalgamation from multiple sources. The final objective is to provide a better solution than could otherwise be achieved from the use of single-source data alone. The proposed framework combines multiple forecasting models and adaptive machine learning techniques for information processing. The input data sources are the wind forecast profiles computed by synoptic physical models and measured data coming from meteorological stations which are amalgamated via an adaptive supervised learning system. The latter is based on a local learning algorithm, called the Lazy Learning (LL) algorithm. This algorithm is sequentially updated, in order to adapt the whole architecture to “new” operating conditions. Experimental results obtained on a one-year time scenario show the effectiveness of the proposed data fusion paradigm in addressing the problem of one-day-ahead wind power forecasting.