Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

04 final - hobbs lave wvm solar portfolios - pvpmc

317 views

Published on

8th PVPMC Workshop, May 9-10 2017

Published in: Technology
  • Be the first to comment

  • Be the first to like this

04 final - hobbs lave wvm solar portfolios - pvpmc

  1. 1. Simulating high-Frequency Solar PV generation Profiles for Large Portfolios in the SE US Will Hobbs, Southern Company Services Matt Lave, Sandia National Laboratories 1
  2. 2. Background Southern Company (vertically integrated utilities in SE US) needed solar profiles that are: • Sub-hourly (10min interval down to 6sec interval) • Multi-year and concurrent with recent load • Adjustable by: – Locations – Capacities – Type (fixed vs. tracking) Applications include: • Resource planning studies on 10 min regulation requirements • Fleet operation studies on 6 sec AGC cost and performance Current solution: MATLAB toolbox that meets all of these requirements
  3. 3. For each site and configuration (fixed, 1-axis tracking): Repeat for all sites & configurations, then sum outputs Model Overview 3 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method Substitute “TAM” for “WVM” to compare methods
  4. 4. Outline 4 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 A.1 4 A.2 3 Results 5
  5. 5. Outline 5 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Results 5 4 A.1 A.2
  6. 6. Input Data (EPRI DPV Project) 6 • Clusters of pole-mounted PV modules (with Tbom) and POA pyranometers at 13 sites • 4+ years of data (2012 – much of 2016) Photo credit: EPRI (http://dpv.epri.com/)
  7. 7. Data Quality Challenges 7Photo credit: EPRI Notable fixed shading at many sites. Irrpoa plotted by solar azimuth, elevation (& filtered)
  8. 8. Outline 8 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Results 5 4 A.1 A.2
  9. 9. Cloud Speed • Cloud speed data obtained from NAM • Daily, monthly, and annual averages computed • Compared well with radiosonde data 9 Mean 24.2 mph Median 21.7 mph Min 1.1 mph Max 75.8 mph Std_Dev 14.6 mph 10th %-tile 8.1 mph 90th %-tile 43.7 mph Cloud speed statistics for Atlanta, 2014, based on weather balloon soundings. Cloud speed statistics across 13 DPV sites 2012-2016 NAM data (Jan Kleissl, Ellyn Wu, UCSD) .
  10. 10. Outline 10 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Results 5 4 A.1 A.2
  11. 11. Smoothing Models • Wavelet Variability Model (WVM) • Time Averaging Method (TAM) – Smoothing window = sqrt(Plant Area) ÷ Cloud Speed 11 M. Lave, A. Ellis, J. Stein, Simulating Solar Power Plant Variability: A Review of Current Methods, SANDIA REPORT SAND2013-4757, June 2013.
  12. 12. Outline 12 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Config. MW Results 5 4 A.1 A.2
  13. 13. Locational Capacities 253MW 1300MW - Fixed 1300MW - Tracking DPV Site Fixed MW Tracking MW Fixed MW Tracking MW AL Eufaula 0 0 100 100 AL Hoover 0 0 100 100 AL Mobile 0 0 100 100 AL Tuscaloosa 0 0 100 100 AL Wedowee 0 0 100 100 AL Wetumpka 0 0 100 100 GA Augusta 0 0 100 100 GA Columbus 29.9 0 100 100 GA Jonesboro 32.2 0 100 100 GA Macon 19.9 50.2 100 100 GA Rome 0 0 100 100 GA Savannah 0 0 100 100 GA Valdosta 17.2 103.3 100 100 13 • 253MW scenario • 1300MW scenarios For validation against measured power (6 month overlap) Sample application for utility planning Real plants (253MW total)
  14. 14. Outline 14 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Config. MW Results 5 4 A.1 A.2
  15. 15. Sample Day Comparison 15 Clear Day: Variable Day: Overhead shading at DPV site(s)
  16. 16. Monthly Energy Comparison 16 Accurate energy estimate is not a primary goal, but results are decent
  17. 17. Ramp Rate Comparisons at Different Time Intervals 17 • WVM matches actuals very well at 1 min • TAM underestimates ramps • WVM and TAM overestimate ramps at 10 min, but WVM is closer • Both match well to 95%-tile • Mixed (but still good) results Overestimation of ramps in WVM at 10 and 60min could be due to shading
  18. 18. Month by Hour 10min ramps, 95th %-tile* 18 Timing of 10 min ramps: TAM has notable concentration of high ramps: WVM looks better: NERC’s BAL standard (now replaced by BAAL) required monthly CPS2 compliance of 90% on 10-minute Area Control Error (we use 95% since solar generates ~1/2 of time) *
  19. 19. Month by Hour 10min ramps, 95th %-tile 19 Timing of 10 min ramps: TAM has notable concentration of high ramps: WVM looks better: Maps of difference from actual ramps: Morning shading at DPV site NERC’s BAL standard (now replaced by BAAL) required monthly CPS2 compliance of 90% on 10-minute Area Control Error (we use 95% since solar generates ~1/2 of time) *
  20. 20. Outline, part 2 20 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2b 3 Results 5 How important is this? 4 A.1 A.2
  21. 21. Daily, Monthly, or Annual Cloud Speed? 21 • How much benefit to using daily cloud speed over monthly or annual avg.? • Minimal change to broad 6-month ramp statistics • What about seasonal issues? Daily cloud speed is best. Overestimation gets worse in late spring/early summer when using monthly or annual avg. cloud speeds.
  22. 22. Outline 22 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Config. MW Results 5b (Sample Application) 4 A.1 A.2
  23. 23. 1300MW Portfolios 23 95th %-tile • Summary for generic planning scenario output 95%10min Ramp (%) 253MW (WVM) 8.3 % 1300MW Fixed 4.6 % 1300MW Tracking 6.8 % 95%10min Ramp (%) 95% 10min Ramp (MW) 253MW (WVM) 8.3 % 21.0 MW 1300MW Fixed 4.6 % 59.2 MW 1300MW Tracking 6.8 % 88.9 MW
  24. 24. Conclusions & Next Steps 24 Implemented and partially validated a method for developing solar profiles that are: • High frequency • Multi-year and concurrent with recent load • Scalable/Adjustable Next steps: • Validate with more recent actual generation (~1000MW) • Look at 6 second intervals • Possibly better address shading • Consider improved Tbom
  25. 25. Thanks to… 25 …EPRI for allowing us to use DPV data (Tom Key, Chris Trueblood, David Freestate, others) …UCSD for providing NAM cloud speed data (Jan Kleissl, Ellyn Wu) Questions? whobbs@southernco.com mlave@sandia.gov
  26. 26. Appendix 26 TBOM (EPRI DPV) Irrad30°S (EPRI DPV) Cloud Speed (NAM/UCSD) Simple PV Power Model Wavelet Variability Model Irradiance Translation Portfolio Table Site Config. MW AC Power (Site, Config) Plant size Plant Density Input Model/Calc Output Time Averaging Method 1 2 3 Results 5 4 A.1 A.2
  27. 27. A.1 Plant Density • Looked at range between 5 acres/MW and 100 acres/MW • Primary focus on 55 acres/MW – Assumption: max plant density of 5 acres per MW, 10% of land in region around measurement site is used for PV  55 acres/MW aggregate 27
  28. 28. A.1 Plant Density Impact 28 • 55 acres/MW is good for 1 min ramps • 55 acres/MW causes small overestimation at 10 min • 100+ acres/MW is better • Plant density has very little impact at 60 min interval WVM is intended to account for spatial smoothing, not weather diversity. This could explain the inconsistency here.
  29. 29. A.2 (Simple) Power Model 29 𝑃 𝐷𝐶,𝑚𝑜𝑑 = 𝑃𝑆𝑇𝐶 𝐺 𝑃𝑂𝐴 𝐺𝑆𝑇𝐶 1 − 𝛿 100 𝑇𝑆𝑇𝐶 − 𝑇𝐵𝑂𝑀 𝑃𝐴𝐶,𝑚𝑜𝑑 = 1 − 𝜇 100 𝑃 𝐷𝐶,𝑚𝑜𝑑 , 1 − 𝜇 100 𝑃 𝐷𝐶,𝑚𝑜𝑑 < 𝑃𝑖𝑛𝑣 𝑃𝑖𝑛𝑣, 1 − 𝜇 100 𝑃 𝐷𝐶,𝑚𝑜𝑑 > 𝑃𝑖𝑛𝑣 • PDC,mod is modeled average DC power (kW); • PSTC is plant DC capacity (kW); • GPOA is plane of array (POA) irradiance (W/m2); • GSTC is test condition irradiance (1000 W/m2); • δ is temperature coefficient for power of the modules (%/°C, typically negative); • TSTC is standard test conditions cell temperature (25°C); • TBOM is back of module (BOM) temperature (°C). • PAC,mod is modeled average AC power (kW); and • µ is a loss factor, including DC mismatch, DC wiring, inverter efficiency, etc. (%); and • Pinv is inverter AC nameplate capacity (kW).

×