REMix (Renewable Energy Mix)

The goal was the assessment of the role of flexibility in Europe’s future energy system. We did this by comprehensive modelling of European networks, energy storage and sector coupling. A special focus was on battery vehicles, hydrogen infrastructures and the transformation of the heat supply. We learned that using flexibility in sector coupling is chapter than electricity storage. Flexible operation of battery vehicles, heat pumps and electrolyzers notably cut peak load. A focus on load balancing shifts from heat to hydrogen with storage with stronger emissions reductions.

Modelling the European energy in a high spatial and temporal resolution translates into solving more complex models and drives the need to introduce suitable approaches for performant computation. One of the main limitations are the state-of-the-art solvers which are used in order to solve the underlying mathematical optimization problem. In particular, they are typically designed to solve the problem on a single compute node. As a consequence, when the size of models increases both the available memory of the compute node and the time to solve the problem can become limiting factors.

To this end, we conducted a systematic evaluation of measures to reduce model complexity and thereby enhance computing performance. We did this by reducing the model size using difference heuristics and by the application of the newly developed parallel solver PIP-IPM++. The various heuristics allowed to speed up the solution time by a factor of 10, however this reduction comes at the cost of accuracy losses. Therefore, choosing the right heuristic for each problem becomes a crucial aspect in order to prevent biases when answering research questions. In contrast, the application of the massively parallel solver PIPS-IPM++ enables modelers to utilize modern high-performance computing infrastructure and benefit from a better scaling behavior. The application of parallel solvers however comes with an additional cost of adjusting the model formulation towards the diagonal block structure required for PIPS-IPM++.

Scenario analyses for future energy systems are often limited to the calculation of a small number of scenarios. On the one hand, this is due to the computational effort. On the other hand, knowledge transfer to human decision-makers requires that the results be condensed to a manageable number of scenarios. However, the uncertainties associated with future energy supply systems are enormous. To derive robust results, sophisticated approaches are needed to systematically test them against different assumptions. REMix's ability to cope with the computational challenges enabled us to perform ensemble simulations for more than 10,000 scenarios in total. Each scenario was evaluated using indicators that describe a scenario's characteristics in terms of affordability, environmental impact and security of supply, to identify scenarios that can be grouped as 'desirable' (SOI), 'not desirable' (AniSOI) or 'average' (Normal). For example, for our modelling setup describing the German electricity system in the next decade, we found that between 400 and 1000 scenarios were required to almost cover the space of possible developments in terms of indicator values. The resulting ranges for the most common indicators (system costs, GHG emissions) are shown below.

In this activity, a systematic comparison of nine energy system models was performed. The comparison focused on the use of flexibility options to balance variable renewable power generation. Nine models were harmonized and two experiments conducted. The first experiment identified significant differences in technology modeling, particularly for electric vehicles, hydro power, and demand response. The second experiment considered more complex scenarios with 16 stylized cases, revealing that differences in model scope had a greater impact on results than differences in optimization approach or technology representation. The project's findings contribute to a better understanding of the effects of different modeling approaches on system dispatch and flexibility options. In the comparison, REMix proved to be one of the most comprehensive frameworks of its kind.