Abstract
We present a scenario generation method for representative days of wind and solar power availability for use in energy-system models. The method uses principal component analysis (PCA) such that the correlations between solar and wind can be captured. PCA is applied to daily time series of hourly profiles of regional solar and wind power availability to yield low-dimensional scenarios, which can be used in regional energy system or energy market models that represent the year with a limited set of representative days. Subsequently, the scenarios generated with PCA are used as building blocks for daily multi-regional scenarios under different assumption on dependence, which can include extreme joint events. As an application, the impact of variability of intermittent renewables with a – numerically tractable – low number of scenarios is applied in an electricity market model, where the increase in resulting price variation caused by solar and wind variability is investigated. Strengths and limits of the approach are also shown in terms of dimensional extensions and by comparison with hierarchical clustering. – The documented software code of the statistical analysis is freely available.
Highlights
Electricity models over sufficiently long time horizons must scope with both short- and long-term variations of supply and demand
We consider representative periods for wind and solar availability, which implies that we target models that are formulated with representative days, for example models that use time slices (e.g. TIMES family [2], and the electricity market model used in this paper)
To fulfill the aforementioned properties, we use as statistical method to capture correlations a principal component analysis (PCA) on the 2 × 24 hour vector of wind and solar availability
Summary
Electricity models over sufficiently long time horizons must scope with both short- and long-term variations of supply and demand. To fulfill the aforementioned properties, we use as statistical method to capture correlations a principal component analysis (PCA) on the 2 × 24 hour vector of wind and solar availability. For cross-regional analysis, PCA may not be used directly because it is based on the covariance matrix between variables, that is, the dependence structure is given entirely by the (Pearson) correlation, which is not suitable to capture joint extreme events; see e.g. We are able to split solar and wind availability for Central Western European regions into few statistical significant components for which we can give simple and novel interpretations These components are suitable to generate low-dimensional representative days. The script files for the PCA and the scenario generation are freely available (link in supplementary material section)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
