Validation and Utilization of Numerical Weather Model Data in Energy Systems Analysis

1. Validation and utilization of numerical weather model data in energy systems analysis of decentralized electricity production Hans Schermeyer, Valentin Bertsch, Wolf Fichtner Hans Schermeyer Athens, October 9th 2014 7th International Scientific Conference on Energy and Climate Change

2. Context within our research Research Group Distributed Energy Systems and Networks • Model-based energy systems analysis considering grid restrictions • Design und analysis of distributed energy systems and smart markets/grids • Evaluation of technologies, investment options and business models for political or entrepreneurial decision support Goal ofthiswork: • Evaluation ofdatafrom a numericalweathermodelfortheapplication in energysystemsanalysis

3. Motivation: Market design to cope with grid constraints • High penetration of RES-E (about 10% of German wind power) • Low population and industry density  low grid capacity Figure: Google-Earth

4. Motivation: Market design to cope with grid constraints • Simulation of renewable electricity generation with a high spatial resolution • Identify grid constraints in the distribution grid • Find a market design for an efficient allocation during congestion Figure: EON Netz GmbH (2014): Netzübersichtkarte -Nordbereich- EisMan-Regionen - http://www.eon-netz.com/pages/ehn_de/Erzeugungsanlagen/Erneuerbare-_Energien-Gesetz/Einspeisemanagement/Einsaetze/files/pdf/Netzplan_nord_EisMan.pdf

5. Data Measurement data • Global irradiation [W/m²] • Temporal resolution: hourly • Spatial resolution: 57 measurement stations in Germany (run by the German national weather service) Model data • Global irradiation [W/m²] • Temporal resolution: up to 10 min • Spatial resolution: 20x20 km • NASA data applied to MM5 numerical weather model by PSU Picture: NOAA Earth System Research Laboratory (ESRL) http://www.esrl.noaa.gov/gmd/grad/surfrad/tmtpics/tracker1.jpg

6. MethodSelect and apply appropriate indicators measuring similarity Measuringsimilaritiesbetween time series Mean bias error (MBE) Describes the general trend of deviations Root mean square error (RMSE) Reflects the scattering of model vs. measurement data Linear correlation (R) Measures linear dependency

7. MethodSelect and apply appropriate indicators measuring similarity Measuringsimilaritiesbetween time series Mean bias error (MBE) Describes the general trend of deviations Focus on energy systems analysis Root mean square error (RMSE) Reflects the scattering of model vs. measurement data Measure the model‘s ability to produce realistic rather than real data with regard to fluctuations and extreme values Linear correlation (R) Measures linear dependency

8. MethodMaximum amplitude of radiation supply (MARS)

9. MethodMaximum gradient of radiation supply (MGRS)

10. MethodSpatial volatiliy

11. Results: MGRS

12. Results: MGRS • As one would expect the MGRS within 3 hours is significantly greater than within 1 hour  dominated by the diurnal cycle • The 3h gradients of the model data are generally higher. This leads to the conclusion that the diurnal cycle in the model data is more distinct and few cloud events are simulated (matches results of MBE and scatterplot) • MGRS shows that model data is less volatile with respect to 1h gradients of radiation supply and provides a quantification. • Altogether the difference in MGRS appears to be within acceptable limits when compared to standard deviation and RMSE as benchmark

13. Results: spatial volatility • Spatial volatility of the measured and modelled data respectively

14. Results:spatial vola • Hours during night and low data availability were excluded • The mean spatial volatility is 29% and 24% respectively • Peaks in the morning and afternoon can be explained by the time shift of sunrise at different sites • Regime-switching during the day: Model data shows higher spatial vola during morning and afternoon hours while the differences between measured sites peak during the day • The spatial volatility reveals substantial differences between modelled and measured data. But: Partly due to different data formats and their temporal resolution

15. Outlook and open questions • Fundamental attributes of a future energy system (grid capacity, demand for flexible back-up generation capacity and its distribution within the grid) caused by RES-E generation on a decentralized level can only be simulated appropriately if extreme and rare values are included in model data • In order to measure these relevant parameters we introduce the MARS, MGRS and the spatial volatility and apply these to the data set. • For future development of the MARS, MGRS and spatial volatility introduced in this paper, it might prove helpful to eliminate deterministic trends of the solar radiation time series. Thus frequent and repetitive changes in radiation supply that are easier to predict do not affect the indicators. • Moreover a higher temporal resolution will help to improve the significance of the MARS, MGRS and spatial volatility. This would enable a more accurate preprocessing of the data on the one hand and allow conclusions on the higher temporal resolution on the other hand.

16. Thank you very much. Questions?! Hans Schermeyer Karlsruhe Institute of Technology (KIT) , Chair of energy economics ‘Distributed energy systems and networks‘ Phone: +49 721 608 44458 Fax: +49 721 608 44682 Email: hans.schermeyer@kit.edu

17. Back Up

18. Mean bias erros MBE • Root mean square error RMSE • Correlation coefficient R

19. ResultsMBE, RMSE, R

20. ResultsFirst order statistics • Compared to literature the model data’s MBE appears to be too large (positive biased) while the scattering measured by the RMSE is within limits of a good model performance. The average correlation of 90% suggests decent model performance. • However, these indicators by themselves appear to be insufficient to assess a model’s performance as they do not capture fundamental characteristics of the time series: • maximum amplitudes • maximum gradients • and spatial generation differences (volatility)