Wind resource clustering based on statistical Weibull characteristics

Abstract

This investigation presents results for the clustering of wind resource data based on statistical Weibull characteristics. The clustering of a chosen geographical area is based on the Weibull and mean wind speed characteristics for each geospatial point for the high energy demand period. The geographically clustered area is chosen from one of the renewable energy development zones, which were identified by the Council of Scientific and Industrial Research. The renewable energy dataset used throughout this study represents the eight renewable energy zones through a meso-scale wind and solar dataset, which spans a 5-year period, at a 15-min temporal resolution. The clustering exercise is aimed at the identification of various geographical areas which best represent a specific independent power producers energy site expectations, while balancing factors such as grid stability and economic and environmental considerations. The study looks into various clustering factors, namely the demand seasons and the energy time of use periods, which correlate to energy production demands for the South African region. The clustering algorithms compared within this study include k-means clustering, the Clustering LARge Applications algorithm, the hierarchical agglomerative algorithm and a model-based clustering algorithm. The initial comparison study yielded the k-means algorithm as the best performing algorithm based on the following internal validation metrics: the Silhouette index, Dunn index and the Calinski-Harabasz index. This clustering method is then subsequently performed on various topical case studies.