Policy implications of downscaling the time dimension in power system planning models to represent variability in renewable output

Abstract

Due to computational constraints, power system planning models are typically unable to incorporate full annual temporal resolution. In order to represent the increased variability induced by large amounts of variable renewable energy sources, two methods are investigated to reduce the time dimension: the integral approach (using typical hours based on demand and renewable output) and the representative days method (using typical days to capture annual variability). These two approaches are tested with a benchmark implementation that incorporates full time representation in order identify their suitability for assessing power systems with high renewable penetration. The integral method predicts renewable representation, it may be more applicable to one-node models, while the representative days method is suitable for multi-regional models. In order to assess power systems with increasing renewable policy targets, models should be designed to handle at least the 160 time steps needed to provide results that do not systematically overestimate the renewable capacity share.

Impact of Germany's energy transition on the Nordic power market – A market-based multi-region energy system model

“…An alternative approach to account for temporal variations is to use a reduced time-scale technique, such as representative days (cf. Nahmmacher et al 2016;Reichenberg et al, 2018;Reichenberg and Hedenus, 2019). By doing so, computationally intractable problems (e.g., modeling a full year of operations) can be characterized with selected, representative time periods.…”

“…Because using the whole year for the model would be too expensive computationally, we use hierarchical clustering to create representative weeks (cf. Reichenberg et al, 2018) with a model developed by Reichenberg and Hedenus (2019). The selected four weeks have typical patterns in hourly demand and VRE generation and they are associated with corresponding weights based on how many weeks they exemplify.…”

“…Additionally, using a more sophisticated way to construct the representative days (and using more of them) in Paper III would improve the representativeness of results. This selection of these days is more challenging than selecting geographical or technical limits, but could be done for instance by following the representative days -methods of Reichenberg et al (2018) and Nahmmacher et al (2016) -similar to what we use for weeks in Paper II.…”

“…For instance, a company exerting market power uses storage differently from what is socially optimal in terms of scheduling and underusing storage (Bushnell, 2003;Schill and Kemfert, 2011;Sioshansi, 2010). Market power also affects storage investment sizes (Nasrolahpour et al, 2016;Siddiqui et al, 2018;Gonzalez-Romero et al, 2019). Finally, storage may even decrease social welfare (Sioshansi, 2014).…”

Opportunity cost including short-term energy storage in hydrothermal dispatch models using a linked representative periods approach

“…The most versatile method for grouping RPs comes from [15] and relies on clustering techniques (e.g., k-means or k-medoids) to group a number of hours with any number of normalized characteristics (solar energy, demand, wind energy, etc.). Furthermore, several authors have debated about the optimal length for RPs [21]. For instance, in [22], the authors suggested representative groups of days or representative weeks, which gives the advantage of increasing the amount of chronology preserved.…”