Frigg: Soft-linking energy system and demand response models

Frigg 2.0: Integrating price-based demand response into large-scale energy system analysis

“…[32]). Modelling demand response as a direct control problem can ease the modelling task, since it might not lead to a higher problem class (staying (mixed-integer) linear) [12]. A lower problem class, in turn, can reduce the model's solution time [34].…”

“…Frigg, a first approach to this problem, was proposed in [12], which we extended in this study. Despite successfully integrating demand response into energy system optimisation, the backward dynamic programming approach in [12] scales badly with a larger number of intertemporal constraints, such as energy storage or investments. It also does not account for the costs of providing demand response in the form of consumers' unwillingness to be flexible.…”

“…In this study, we propose an update to Frigg, a first version published in [12], and couple it to plan4EU, a dispatch model of the European electricity system (Section 3.2). Frigg is a softlinking framework for integrating energy system models and price-based demand response models.…”

“…Frigg is a softlinking framework for integrating energy system models and price-based demand response models. The price-based demand response model applied in this study, as in [12], is a Flexibility Function (see Section 3.3), consisting of a set of ordinary, nonlinear, differential equations that model end-consumer demand as a function of the price of (some commodity of) energy and a state variable that captures how responsive consumers have been in the past.…”

“…Heat pumps and intelligent heat storage specifically have been found to improve system efficiency [11]. Thus, an adequate representation of heat-demand flexibility, namely heat storage and demand response, in energy system analysis is crucial in providing deci-sion support [12]. However, end-consumer behaviour is traditionally not included in energy system modelling [13], where demand is mostly modelled as a fixed input rather than a model variable [14].…”