Sensitivity analysis of time aggregation techniques applied to capacity expansion energy system models. (1st July 2020)

• Record Type: Journal Article
• Authors: Buchholz, Stefanie; Gamst, Mette; Pisinger, David
• Is Part Of: Applied energy. Volume 269(2020)
• Publication Date: 2020-07-01
• Subjects: Capacity expansion problem -- Solution time reductions -- Time aggregation -- Sensitivity analysis -- Performance comparison; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2020.114938
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: Aggregations based on representative days have better performance. An aggregation that groups input data prior to selection performs well. Aggregation performance is highly sensitive to imperfect knowledge of input data. Decreased aggregation quality in both high and low wind resourced systems. Improvements are needed to efficiently aggregate systems with seasonal storage. Abstract: Time aggregation techniques have shown great potential for efficiently solving comprehensive energy system models. Numerous studies document significant solution time reductions while maintaining high solution qualities. However, most of these studies lack providing guidelines for which aggregation techniques to use in a given setup, and why. A recent interest of comparing different aggregation techniques has arisen, and this paper contributes to that trend. We present a sensitivity analysis methodology that studies how different problem changes, influence the aggregation technique performances. The performance is measured as the ability of the aggregated problem to fully or partially replicate the solution achieved by the non-aggregated problem. The applicability of the methodology is illustrated through a case study considering three types of problem changes namely changes in the wind availability, changes in the aggregated problem size, and changes in the energy system design. By applying the suggested methodology to ten different aggregation techniques, key properties which seem to ensure promising performance are identified. Results show that, independently of the sensitivity measure applied, aggregations based on grouping strategies and day selections lead to better and more consistent performance and to the largest improvements in solution times. Strong and very consistent performance is also observed for the Optimized criteria Selection which furthermore outperforms any other technique in scenarios closely replicating real life applications. Contrarily, the Residual Load Duration Curve (RLDC) Selection consistently shows the worst performance which might be related to its configuration of constructing time elements from the selection of hours causing important within day/week chronologies to be lost.