EURO 2025 Leeds
Abstract Submission

662. Modeling the stochastic and sequential real-time operation of energy systems during design optimization via multi-step Benders decomposition

Invited abstract in session WD-44: Advanced methods for designing and operating energy systems under uncertainty, stream Energy Economics & Management.

Wednesday, 14:30-16:00
Room: Newlyn 1.01

Authors (first author is the speaker)

1. Benedict Brosius
Institute of Technical Thermodynamics, RWTH Aachen University
2. Daniel Jost
Institute of Technical Thermodynamics, RWTH Aachen University
3. Frederike Kuperjans
Institute of Technical Thermodynamics, RWTH Aachen University
4. Benedikt Nilges
Institute of Technical Thermodynamics, RWTH Aachen University
5. Niklas von der Assen
Institute of Technical Thermodynamics

Abstract

Design optimization methods are widely used to identify the optimal design of integrated energy systems (IESs). For simplicity, most design optimization methods assume perfect foresight over the operational time horizon. However, the real-time IES operation is stochastic and sequential due to uncertain forecasts and limited forecast horizons. Thus, design optimization models that assume perfect foresight will estimate unattainable operational cost, leading to suboptimal designs. The perfect foresight assumption particularly affects the operation of storages, which are essential for integrating large shares of renewables.

This study proposes a novel multi-step design optimization method that increases the stochastic and sequential detail of IES operation in each step. The method is based on Benders decomposition, where the operational subproblem is a stochastic rolling horizon operational optimization, a common approach for the real-time IES operation.

In two IES case studies, we compare the designs resulting from our novel method to designs generated with the perfect foresight assumption. We evaluate all designs via stochastic rolling horizon operational optimization. Compared to assuming perfect foresight, our method reduces the total objective value by 1.3%. In addition, it greatly increases reliability, reducing the expected undersupply by up to 56.8%. These findings reveal the need to reflect the stochastic and sequential IES operation during design optimization.

Keywords

Status: accepted

Back to the list of papers