Model-based scenario analyses of future energy systems often come to deviating results and conclusions when different models are used. This may be caused by heterogeneous input data and by inherent differences in model formulations. The representation of technologies for the conversion, storage, use, and transport of energy is usually stylized in comprehensive system models in order to limit the size of the mathematical problem, and may substantially differ between models. This paper presents a systematic comparison of nine power sector models with sector coupling. We analyze the impact of differences in the representation of technologies, optimization approaches, and further model features on model outcomes. The comparison uses fully harmonized input data and highly simplified system configurations to isolate and quantify model-specific effects. We identify structural differences in terms of the optimization approach between the models. Furthermore, we find substantial differences in technology modeling primarily for battery electric vehicles, reservoir hydro power, power transmission, and demand response. These depend largely on the specific focus of the models. In model analyses where these technologies are a relevant factor, it is therefore important to be aware of potential effects of the chosen modeling approach. For the detailed analysis of the effect of individual differences in technology modeling and model features, the chosen approach of highly simplified test cases is suitable, as it allows to isolate the effects of model-specific differences on results. However, it strongly limits the model's degrees of freedom, which reduces its suitability for the evaluation of fundamentally different modeling approaches.@en

The transition towards decarbonized energy systems requires the expansion of renewable and flexibility technologies in power sectors. Many powerful tools exist to find optimal capacity expansion. In a stylized comparison of six models, we evaluate the capacity expansion results of basic power sector technologies. The technologies under investigation include base- and peak load power plants, electricity storage, and transmission. We define four highly simplified and harmonized test cases that focus on the expansion of only one or two specific technologies to isolate their effects on model results. We find that deviating assumptions on limited availability factors of technologies cause technology-specific deviations between optimal capacity expansion in models in almost all test cases. Fixed energy-to-power ratios of storage can entirely change optimal expansion outcomes, especially shifting the ratio between short- and long-duration storage. Fixed initial and final-period storage levels can affect the seasonal use of long-duration storage. Models with a pre-ordered dispatch structure substantially deviate from linear optimization models, as missing foresight and limited flexibility can lead to higher capacity investments. A simplified net transfer capacity approach underestimates the need for grid infrastructure compared to a more detailed direct current load flow approach. We further find deviations in model results of optimal storage and transmission capacity expansion between regions, and link them to variable renewable energy generation and demand characteristics. We expect that the general effects identified in our stylized setting also hold in more detailed model applications, although they may be less visible there.@en

This paper presents the results of a multi-model comparison to determine outcome deviations resulting from differences in power system models. We apply eight temporally and spatially resolved models to 16 stylized test cases. These test cases differ in their renewable energy supply share, technology scope, and optimization scope. We focus on technologies for balancing the variability of power generation, such as controllable power plants, energy storage, power transmission, and flexibility related to sector coupling. We use harmonized input data in all models to separate model-related from data-related outcome deviations. We find that our approach allows for isolating and quantifying model-related outcome deviations and robust effects concerning system operation and investment decisions. Furthermore, we can attribute these deviations to the identified model differences. Our results show that trends in the use of individual flexibility options are robust across most models. Moreover, our analysis reveals that differences in the general modeling approach and the modeling of specific technologies lead to comparatively small deviations. In contrast, a heterogeneous model scope can cause substantially larger deviations. Due to a large number of models and scenarios, our analysis can provide important information on which investment and operation decisions are robust to the model choice, and which modeling approaches have an exceptionally high impact on results. Our findings may guide both modelers and decision-makers in properly evaluating the results of similarly designed power system models.@en