Challenges in the Assimilation of PV Power Data

Challenges in the Assimilation of PV Power Data in the Convection-Permitting High-Resolution NWP Model COSMO-DE Stefan Declair*, Yves-Marie Saint-Drenan, Roland Potthast Erstellung innovativer Wetter- und Leistungsprognosemodelle für die Netzintegration wetterabhängiger Energieträger - Eine Kooperation von Meteorologie und Energiewirtschaft - 80. Jahrestagung der DPG und DPG-Frühjahrstagung Regensburg, March 08 th 2016

Motivation: Low Stratus Planetary boundary layer height free atmosphere Inversion layer Ekman layer (1000 m) Prandtl layer (100 m) laminar (mm) temperature

Agenda 1. Data Assimilation a) Method b) Observation Data 2. Impact Experiment a) Monitoring: Bias and Error b) Assimilation: Impact?

Agenda 1. Data Assimilation a) Method b) Observation Data 2. Impact Experiment a) Monitoring: Bias and Error b) Assimilation: Impact?

Forecast: Can I cross the street without getting hit? Information used: Forecast errors due to: • Observations • Observation (estimation) errors • Knowledge about cars, street, etc • Model errors (icy street) • Experience statistics • Case does not match statistics

Numerical Weather Forecast Model: First-guess Data assimilation tool Observations Analysis: Improved initial conditions for next integration step

Available Observation Data Ø Blue: large-scale PV power plants (~200) Ø Grey: small-scale PV power panels (~3. 7 k) Ø Meta data: Ø Lon/lat coordinates Ø Tilt angle Ø Azimuth angle Ø Degradation coefficients Ø Corrections (fitted 1)) for : Ø Temperature Ø Power 1) Source and Work: Yves-Marie Saint-Drenan, Fraunhofer IWES

Model Equivalent: PV Power Forward Operator Forward operator for PV module Model variables: Ø Shortwave surface irradiance (direct and diffuse downward) Ø Panel ambient temperature Ø Surface albedo Synthetic PV power Module meta data: Ø Panel azimuth/tilt angles Ø lon/lat Ø Degradation coefficients Ø Corrections for temperature and power Ø Compute angle between panel normal and sun position at current time Ø Transform horizontal model irradiation into tilted panel plane Ø Compute losses (soiling, shading, module temperature, optical losses)

Sources of Error Ø Inaccurate/unknown meta data Ø Angles Ø Large diversity in panel manufacturers Ø Local deviations (data) from Ø fitted corrections 1) Ø surface albedo Ø Local deviations (model) from Ø Aerosol optical thickness 2) Ø Cloud positions Ø Radiation scheme 3) Source and Work: Yves-Marie Saint-Drenan, Fraunhofer IWES Tegen et al. , Journal of Geophysical Research. , 102, pp. 23895 -23915 (1997) 3) Ritter et al. , Monthly Weather Review, 120, pp. 303 -325 (1992). 1) 2)

Agenda 1. Data Assimilation a) Method b) Observation Data 2. Impact Experiment a) Monitoring: Bias and Error b) Assimilation: Impact

Monitoring of Observations Ø Determination of PV power model equivalent bias and observation error Ø Cycling over 1 week, PV power data only passive in assimilation Ø Valid for 2014051600 - 2014052300

First Glimpse on First Results: Low Clouds Control First-guess Analysis Difference

First Glimpse on First Results: Low Clouds Experiment First-guess Analysis Difference

First Glimpse on First Results: Low Clouds 06 UTC 09 UTC 12 UTC

First Glimpse on First Results: Mid Clouds 06 UTC 09 UTC 12 UTC

Conclusion Ø LETKF successfully utilized to assimilate PV power obserations Ø Cloud cover correction pretty well for low and middle clouds despite strong non-locality in the morning Ø Spread increase due to PV power data assimilation better representation of model error Outlook Ø Stock up on PV power data! Ø Experiments with better localization lengths Ø Detailed analysis of increments in other atmospheric fields Ø Single observation experiments Ø Impact on forecast quality Ø Combi-experiment: add wind power data to assimilation cycle

Thank you for your attention! Now: Q & A