Wind power pattern clustering can potentially supply information about the effect of incorporating wind farms in smart electrical grid without in-depth analysis and studies of lengthy data. The present study… Click to show full abstract
Wind power pattern clustering can potentially supply information about the effect of incorporating wind farms in smart electrical grid without in-depth analysis and studies of lengthy data. The present study investigates the most effective clustering technique and optimum number of clusters for wind power pattern data through various unsupervised clustering techniques. It also presents the introduction of Ant Colony and Bat, swarm optimization strategies in clustering wind power patterns. Three clustering algorithms from two different unsupervised techniques were concerned. A total of eight validity indices were used; Davies Bouldin, mean square error, mean index adequacy, ratio of within-cluster sum-of-squares to between-cluster-variation, Dunn, Silhouette, Xie-Beni, and clustering dispersion indicator for evaluation of the unsupervised clustering algorithms in inclusive manner. Findings depicted that Bat bio inspired clustering is comparative to K-means clustering and the most effective combination of clustering algorithm and validity index was K-means and Silhouette index, respectively. Secondly, in order to achieve improved clustering of WPP, the best clustering algorithm (K-means with Silhouette index) was modified by integrating the Silhouette index as an objective function for K-means. To check the potency of the produced wind power pattern representatives during a wind system simulation, a short wind generation prediction model is presented. The results of those cluster representatives presented promising short-term prediction results and suggest that the produced wind power pattern cluster representatives can potentially be used in other wind power pattern simulations.
               
Click one of the above tabs to view related content.