Special-purpose optimization algorithms for demand side management

  • Spezielle Optimierungsalgorithmen für Laststeuerungsaufgaben

Varelmann, Tim; Mitsos, Alexander (Thesis advisor); Baldea, Michael (Thesis advisor)

Aachen : RWTH Aachen University (2022)
Book, Dissertation / PhD Thesis

In: Aachener Verfahrenstechnik Series 24
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen, Diagramme

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2022


Demand side management is a promising approach to adjust the energy demand to the variable renewable energy supply and thereby facilitate a stable grid operation. The idea of demand side management is that electricity users are exposed to different incentives in a variety of markets to adjust their demand to the current supply. For energy-intensive process operators, it can be difficult to decide in which markets to monetize the flexibility of their process and how to do that. In this thesis, we will develop both an algorithm and suitable models that allow energy-intensive processes to optimize the way they participate in an auction-based balancing reserve market with pay-as-bid mechanism while simultaneously exploiting price differences on an electricity spot market through a variable production schedule. With a Benders decomposition, we separate the bidding and the scheduling problem. Then we formulate the model for the scheduling problem in a way that allows to derive Benders cuts for the bidding problem. Such cuts describe the influence of the bidding decisions on the objective of the scheduling problem and thus facilitate the optimization of the overall problem. In a case study of an aluminum electrolysis process, we show the benefits of our algorithm in terms of computational efficiency while maintaining solution quality. Furthermore, the thesis at hand addresses issues in a recently conceptualized form of cooperation between electricity users and grid operators: In cooperative optimal power flow calculations, an electrical power generation and transmission schedule is computed that is capable of directly controlling the flexibility of energy-intensive processes, rather than indirectly nudging energy users to shift their demand by exploiting price profiles. In particular, we provide an algorithm that computes the solutions to cooperative optimal power flow calculations while maintaining confidentiality of dynamic process models. We provide an outer approximation of the feasible region of power profiles of cooperating plants to the grid. Then, the grid suggests potential power profiles to the plants based on its current approximation of their feasible region. In response to the suggestions, the approximation is iteratively refined until a solution with acceptable optimality gap is found. The refinement is based on cuts that the process models generate for the grid model. During these iterations, all information exchange takes place in the domain of power profiles only. We apply our algorithm in the IEEE 24-bus reliability test grid with cooperating chlor-alkali processes. Additionally, we extend the aforementioned algorithm by a different cut-generating subproblem formulation, which searches for a feasible power profile along the ray from the profile suggested by the grid towards a core point of the feasible region of power profiles, rather than minimizing the 1-norm of the required change to the profile suggested by the grid. The novel subproblem formulation is designed to provide high-quality cuts. By efficient utilization of the different formulations, we improve both the overall computational performance of the algorithm and its scaling behavior with the number of cooperating processes. We validate the desired effects of our algorithm and the maintained solution quality with a case study in the IEEE 24-bus reliability test grid.