Model-based approaches to demand side management of continuous industrial processes

  • Modellgestützte Ansätze zur Laststeuerung kontinuierlicher industrieller Prozesse

Schäfer, Pascal; Mitsos, Alexander (Thesis advisor); Grossmann, Ignacio E. (Thesis advisor)

Aachen : RWTH Aachen University (2020, 2021)
Book, Dissertation / PhD Thesis

In: Aachener Verfahrenstechnik 13 (2020)
Page(s)/Article-Nr.: 1 Online-Ressource (XVIII, 167 Seiten) : Illustrationen, Diagramme

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020

Abstract

Industrial demand side management is increasingly recognized as an important measure to align the network-wide electricity consumption with a volatile generation from intermittent renewable sources. At the same time, induced price fluctuations at liberalized electricity markets offer promising monetary incentives for flexible operation. Consequently, within the past decades, numerous works aimed at raising the economic potentials of adjusted production schedules by using model- and optimization-based approaches. Within this context, the contribution of the present thesis is fourfold. First, the relevant literature currently mainly confines to scheduling optimization problems embedding (piecewise) linearized process models to maintain computational tractability. In contrast, we herein focus on the development of algorithms that allow for the consideration of long planning horizons with fine discretizations in the case of scheduling formulations with nonlinear process models. In particular, we show that feasible near-optimal schedules can thereby be furnished, although exposing only a small fraction of the degrees of freedom to the optimizer, which leads to substantial savings in computational times compared to solution approaches considering the full dimensionality. Second, we build on existing works that argue to account for the transient process behavior during load changes in scheduling optimizations for processes with slow dynamics. In particular, we herein present a hybrid mechanistic/data-driven model reduction approach for distillation columns that allows for substantial reductions in computational times while providing similar prediction capabilities as full-order stage-to-stage models. Thereby, we enhance the real-time capability of economic nonlinear model predictive control schemes and enable participation in continuous electricity markets directly at the control layer. Third, we address the optimal decision-making at pay-as-bid electricity markets, such as the German market for primary balancing power, where a severe trade-off between a higher probability of acceptance of a bid and a higher potential compensation payment exists. To account for this trade-off, we herein propose a two-stage stochastic problem formulation, where we optimize the bid at the balancing market auction in the first stage and market the remaining flexibility at the spot market in the second stage by a scheduling optimization. Fourth, we critically question currently discussed measures increasing the flexibility potential of industrial processes from the perspective of the decarbonization of the electricity sector. For this purpose, we consider a generic process and evaluate whether measures that enhance the exploitation of time-variable electricity prices also result in a reduction of the consumption of fossil-generated electricity in favor of renewable electricity. Most importantly, we show that characteristic price fluctuation patterns prevent adequate monetary incentives for providing storage capacities that enable load shiftings on time scales in the order of one day, which would however be crucial for reducing the environmental impacts associated with the generation of the purchased electricity.

Identifier

Downloads