Clustering analysis of multidimensional wind speed data using k-means approach

Abstract

The energy capability of wind power plants is strictly correlated with the wind characteristics of the considered site. For this reason, it is very important to process the wind speed data for making the wind power more competitive with respect to other energy sources. This paper presents a detailed similarity analysis to discover the meaningful subsets within the monthly average wind speed data of 75 provinces in Turkey. In the similarity analysis, the k-means clustering method is adapted with Squared Euclidean, City-Block, Cosine and Pearson Correlation distance measures. In addition, the silhouette coefficient is used to validate how well-separated the resulting clusters are. As a result of the optimal silhouettes acquired for k = 5 and Squared Euclidean distance measure, many comparative assessments are made about the monthly average wind speed characteristics of all provinces.