Evaluating Carrier-Specific Temporal Resolutions in Multi-Energy System Optimization

Kurzfassung

Increasing temporal, spatial, and technological detail improves the representation of interactions within multi-energy system optimization models but can lead to substantial computational requirements. At the same time, different energy carriers exhibit different characteristic time scales, raising the question of whether applying a uniformly fine temporal resolution throughout a multi-energy system is computationally necessary. Carrier-specific temporal resolution addresses this issue by retaining fine temporal detail where required while representing other carriers at coarser resolutions. Although the approach has previously demonstrated substantial computational benefits, its performance across different model structures, increasing spatial model sizes, and different choices of carrier-specific resolutions has received limited investigation.

This thesis evaluates these aspects using a scalable synthetic multi-regional benchmark model implemented in AnyMOD. Twenty instances containing between 2 and 40 regions are compared using a uniform hourly formulation, a carrier-specific configuration based on the temporal characteristics of the represented carriers, and a deliberately coarse configuration. An equivalent uniform-resolution implementation in REMix is additionally used for cross-framework validation.

The carrier-specific configuration reduces the number of non-zero matrix coefficients by approximately 50\%, peak memory requirements by approximately 33\%, and total runtime by approximately 54--57\% relative to the uniform AnyMOD formulation. The resulting objective value lies approximately 3\% below the uniform reference. The computational benefit persists as spatial model size increases, with the clearest scaling improvement occurring in model construction, whereas no systematic improvement in solver-time scaling is observed. Furthermore, reductions in mathematical problem size do not translate proportionally into reductions in computational effort. More aggressive temporal aggregation provides only limited additional problem-size reduction, does not consistently improve runtime, and produces substantially larger deviations from the uniform-resolution solution.

The results demonstrate that carrier-specific temporal resolution can provide substantial computational savings in multi-energy system optimization, while emphasizing that its effectiveness depends on the selected resolutions and the structure of the resulting formulation. Temporal detail should therefore be allocated selectively according to the characteristics and requirements of the modeled system rather than reduced as far as possible.