On the aggregation of input data for energy system models

The increasing share of variable renewable energy sources in power systems poses new challenges to policy makers and network planners alike because of the sources' variability and insufficient energy storage capability. This requires the development of new optimization models that consider the inter-temporal connection between different periods aggregating them on a more general level (e.g., days or weeks). However, aggregation models are often empirical and based on common clustering algorithms. In this paper, we carry out a numerical exploration of the relationship between the structure of the system and the hyperparameters required for these aggregation procedures. Our findings indicate that there is valuable information from the power system that can be used to improve the aggregation. This is important because, in most cases, the accuracy of the aggregation cannot be measured exactly.