Toke Tobiasen: The influence of spatial aggregation of wind data in modeling of energy systems, MSc Thesis, DTU, 2016
Wind power has become a much larger part of the Danish energy system in recent years. Being a non-dispatchable power source, much more planning of the energy system is required in order to utilize the wind resource optimally. The Danish energy system was modeled using the mathematical modeling tool Balmorel. The focus of this thesis was to optimize the method for modeling wind power in relation to the remaining energy system. The optimization primarily looked into the effects of spatial aggregation of wind power, but also briefly investigated the effects of temporal aggregation. An investigation into usable wind data sets showed that CFSR wind data made by the NCEP, generally has low wind speeds compared to observed production. Therefore, a method was implemented such that the hourly variations from a reanalysis data set is scaled to distributions of the GWA, using the Weibull distribution parameters from both datasets. The GWA was used to correct the CFSR data,set, as this wind distribution corresponded better with observed production data. A method was developed for the creation of aggregated power curves for a modeled wind area. These were created using a reference power curve, a representative wind time series as well as observed wind turbine heights, installed capacities and yearly production for the chosen site. The method fits the power curves to observed production and therefore corrects the corresponding wind time series to produce power equivalent to observed values. A single representative time series was used for each area. The site of the time series is based on the site where most wind turbines currently exist. Five cases where created, dividing Denmark into 1, 2, 5, 10 and 20 areas respectively, to investigate the impact of spatial aggregation on the results based on total wind power production, curtailment and power prices. Above two areas, the effects of implementing more areas were seen to be miniscule. Two areas are seen to be enough to model Denmark, although adding more areas will increase the accuracy slightly. Investigations of the temporal resolution shows that when modeling Denmark as an island, the time aggregation does not yield the same results as hourly resolution. This is due primarily to hours with very high power prices that are extrapolated when using the time aggregation. Finally, it was found that the distance from a site, where wind speeds are no longer representative, is around 250 km. Therefore, the modeled wind areas should have a radius no longer than 250 km. This distance will be shorter if the terrain becomes more complex, i.e. mountains or similar. Furthermore the area size needs to consider the wind turbine capacities, as an area with almost no installed wind capacity will not change anything, if combined with an area with much larger existing capacity.
Available at http://production.datastore.cvt.dk/oafilestore?oid=579743c76bbf232e70001000&targetid=579743c86bbf232e7000100c