Applications of Clustering Techniques to the Definition of the Bidding Zones

Abstract

The definition of the bidding zones is an essential aspect to develop an electricity market on large power systems, by trading off between situations with uniform price and nodal prices. In this paper, the results of the application of a set of clustering algorithms (k-means, k-medoids, hierarchical clustering, and price differential clustering) to the formation of the bidding zones are presented. The classical versions of the methods used require post-processing to identify connected bidding zones. The customised versions of the methods incorporate topology constraints in the clustering procedures. The input data are given by locational marginal prices (LMPs), power transfer distribution factors (PTDFs), and network topology information. The clustering algorithms have been applied to the study of a reduced model of the European transmission system, for different numbers of clusters. The results have been assessed by using two synthetic indicators that represent clustering validity and the possible occurrence of market power in the bidding zones formed. Finally, a conjecture has been expressed: methods that use LMPs tend to be more effective according with classical clustering validity-based indicators, while methods that use PTDFs tend to be more effective according with market power-based indicators.