Cluster Based ACDC Algorithm for Large Scale Unit Commitment Solution

This paper presents a new methodology for the most complex large scale unit commitment problem using agglomerative and divisive hierarchical clustering algorithms. Euclidean costs, a measure of difference in fuel cost and start-up costs of any two units are computed On the basis of these Euclidean costs, similar type of units is placed in a cluster and accordingly different clusters are formed This cost is also useful for preparing priority lists for the units in a cluster. The method can be broadly unfolded into three stages. First stage deals with the formation of clusters and priority list of units. Second stage presents an agglomerative hierarchical clustering algorithm that provides unit commitment solution while the load is increasing. Third stage presents a divisive hierarchical clustering algorithm that provides unit commitment solution while the load is decreasing. These algorithms search for an optimal solution that provides list of on-line units from a minimal number of clusters and cluster data points. Performance of the method is evaluated on a large scale power system consisting of 10 to 100 units in the multiples of 10 units. Input data is the standard ten thermal unit power systems for a period of 24 hours. Numerical results are presented and compared with the existing popular methods in most judicious way. Copyright (C) 2009 Praise Worthy Prize S.r.l. - All rights reserved.