Study on Weather Typing Based Wind Power Prediction

It is an effective means to predict the output power of wind farms for ensuring secure and stable operation of power grid concentratedly connected with large-scale wind farms. A whether typing based wind power prediction algorithm is proposed. Based on wind speed vector in numerical weather prediction(NWP) and daily atomospheric pressure and using principal component analysis, the proposed algorithm performs dimensionality reduction for samples and classifies weather types by clustering analysis, and different prediction models are established respectively according to different whether types and compared with single prediction model. Research results show that using the method combining principal component analysis with clustering analysis the weather phenomena can be effectively classified; for more stable whether phenomena the prediction result by clustering model is far more accurate than that by single model; for instable whether phenomena, limited improvement of prediction accuracy by clustering model can be attained; as for overall samples, the accuracy improvement by whether typing based prediction method can reach to 2% and more than traditional prediction method.