1722. Analyzing Techniques to Connect Representative Periods in Energy System Optimization Models
Invited abstract in session MA-44: Advancing Energy System Models, stream Energy Economics & Management.
Monday, 8:30-10:00Room: Newlyn 1.01
Authors (first author is the speaker)
| 1. | Felix Clemens Alexander Auer
|
| Institute of Electricity Economics and Energy Innovation, Graz University of Technology | |
| 2. | Diego Tejada
|
| Energy Transition Studies, TNO | |
| 3. | Sonja Wogrin
|
| Graz University of Technology |
Abstract
In energy system optimization, time series aggregation using representative periods is a common technique to reduce computational complexity and solving times for large-scale models. However, accurately handling constraints at the edges of representative periods to model the transitions between them proves to be a challenge, since the overall deviations from the original problems (or rather their solutions) should be kept minimal.
Currently, three methods are mainly used to handle edges between representative periods: Not enforcing the constraints at the edges at all; fixing the variables at the edges; or, establishing cyclic constraints within each representative period. While easy to implement, each of those methods has its limitations with respect to accurately capturing the original, non-aggregated problem.
We introduce an innovative approach leveraging Markov chains to capture the stochastic transitions between periods. Importantly, the technique is not limited to constraints concerning continuous variables, but also applicable to those with binary variables.
We present a case study of a unit commitment model with ramping constraints, but the Markov chain method can also be extended to other types of constraints. Analyzing the differences in model size, solving time, objective function value and individual unit commitment, we show the potential of this method to improve energy system optimization models which are using representative periods.
Keywords
- Complexity and Approximation
- Optimization Modeling
- OR in Energy
Status: accepted
Back to the list of papers