Systematic Comparison of Aggregation Methods for Input Data Time Series Aggregation of Energy Systems Optimization Problems

Abstract We present a systematic comparison of how aggregation methods represent input data in the operational part of complex energy systems optimization problems. We compare both traditionally used methods such as k-means, k-medoids, and hierarchical clustering, and shape-based clustering methods such as dynamic time warping and k-shape in the domain of the objective function of sample operational optimization problems. Centroid-based approaches show improved performance as the number of clusters increases, whereas medoid-based approaches do not show a clear pattern as to when they best represent the objective function of the optimization problem of the full representation. Furthermore, clustering based on shape-based algorithms can improve performance compared to the traditionally used algorithms.