Stochastic modelling of variable renewables in long-term energy models: Dataset, scenario generation & quality of results. (1st December 2021)

• Record Type: Journal Article
• Authors: Seljom, Pernille; Kvalbein, Lisa; Hellemo, Lars; Kaut, Michal; Ortiz, Miguel Muñoz
• Is Part Of: Energy. Volume 236(2021)
• Publication Date: 2021-12-01
• Subjects: Variable renewable power generation -- Energy-system modelling -- Scenario generation -- Stochastic programming -- Stability tests -- Satellite-data; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2021.121415
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Variable electricity generation from wind and solar influences the design of a cost-efficient and reliable energy system. This paper presents a method that uses stochastic programming to represent variable renewable electricity generation in long-term energy system models, and demonstrates this on a Norwegian TIMES model. First, we derive hourly PV- and wind-generation data by modifying satellite-based data, based on a comparison with historical generation data. Second, the satellite-based dataset is transformed into a manageable set of scenarios that is used as an input to the stochastic energy-system model. This is done using six different scenario generation methods. Third, we solve the energy-system model with three of the scenario-generation methods and evaluate the quality of the corresponding model value by stability tests. We demonstrate that scenarios generated from the six methods have significantly different moment-based and Wasserstein distance error relative to the dataset. Further, the energy system model results show that the number of scenarios needed to achieve stability differs between the three used scenario generation methods. Highlights: A stochastic method to model VRES in long-term energy system planning is presented. There is a mismatch between satellite-based and historical wind power generation. The distribution error of six scenario generation methods is evaluated. Stability tests are used to evaluate the quality of model results. Number of scenarios needed to achieve good quality depends on scenario method.