m‐ISODATA: Unsupervised clustering algorithm to capture representative scenarios in power systems.

Summary: This paper presents an unsupervised clustering algorithm, called modified Iterative Self‐Organizing Data Analysis Technique Algorithm (m‐ISODATA), to capture representative nonchronological scenarios for representing short‐term uncertainties in power system models. The proposed approach is suitable to automatically obtain the number of scenarios required to fully capture the variability of historical series, avoiding the need of adjusting the number of clusters as in techniques commonly used in the literature. The performance of the m‐ISODATA is discussed and compared with Monte Carlo simulation, the well‐known k‐means, and hierarchical agglomerate clustering algorithms. In addition, the obtained scenarios are applied to a wind‐solar‐thermal power system generation expansion planning and to a probabilistic optimal power flow, considering uncertainties over wind and load demand. Finally, the source codes are provided with the best parameters as default. [ABSTRACT FROM AUTHOR]