K-means clustering with a new initialization approach for wind power forecasting

The recent and increasing integration of renewable energy sources into the energy grid has brought the necessity of a more accurate forecasting framework, capable of predicting and modeling these sources with greater precision. This paper proposes a hybrid wind power forecasting method, which utilizes the K-means clustering technique with a new initialization approach that aims to increase the accuracy of the final forecast. By clustering historical data, selecting proper clusters' centroids and optimal groups of data to be used as input to the neural networks, the precision of the output is greatly increased. The performance of the proposed initialization method is evaluated and compared to other initialization techniques. Wind power datasets with diverse characteristics, from different wind farms located in the United States, are used to determine the accuracy of the hybrid forecasting method through various performance measures.