Accuracy indicators for evaluating retrospective performance of energy system models. (1st November 2022)

• Record Type: Journal Article
• Authors: Wen, Xin; Jaxa-Rozen, Marc; Trutnevyte, Evelina
• Is Part Of: Applied energy. Volume 325(2022)
• Publication Date: 2022-11-01
• Subjects: Energy system models -- Energy projections -- Energy scenarios -- Accuracy indicators -- Optimization -- Model evaluation; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2022.119906
• Languages: English
• ISSNs: 0306-2619

Abstract

Graphical abstract: Highlights: Retrospective cost-optimal scenarios are generated in 31 European countries in 1990–2019. 24 accuracy indicators from literature are used for evaluating 34 model outputs. The most informative smaller set of five accuracy indicators is identified, using clustering-based dissimilarity analysis. Modeled scenarios in 31 countries are assessed by applying this small set of accuracy indicators. Abstract: Retrospective evaluation of energy system models and scenarios is essential for ensuring their robustness for prospective policy support. However, quantitative evaluations currently lack systematic methods to be more holistic and informative. This paper reviews existing accuracy indicators used for retrospective evaluations of energy models and scenarios with the aim to find a small suite of complementary indicators. We quantify and compare 24 indicators to assess the retrospective performance of D-EXPANSE electricity sector modeling framework, used to model 31 European countries in parallel from 1990–2019. We find that symmetric mean percentage error, symmetric mean absolute percentage error, symmetric median absolute percentage error, root-mean-squared logarithmic error, and growth error together form the most informative suite of indicators. This study is the first step towards developing a model accuracy testbench to assess energy models and scenarios in multiple dimensions retrospectively.