1_2023-05-10_PVPMC_UL_Albedo_HalleyDarling-1

© 2023 UL LLC. All Rights Reserved. Impacts of Albedo Estimation Method on Energy Estimates PVPMC 2023 Halley Darling, Renn Darawali, Lucila D. Tafur May 10, 2023Agenda 1. Background and Motivation 2. Albedo Measurement and Modeling 3. Sample Locations 4. Energy Modeling Assumptions 5. Results 6. Conclusions 7. Future Work 2Background and Motivation 3Albedo • Albedo is the ratio of reflected solar irradiance over global irradiance and depends heavily on the reflectivity of the surface material • Typical ranges are 0.15-0.20 for darker soil, 0.20-0.30 for very light soil, 0.20-0.25 for vegetation, and up to ~0.75-0.85 for fresh snow • Albedo can have a significant effect on bifacial energy; Albedo inc is the incident irradiation on the back of the panel, ρ is the ground surface albedo under the panel, and θ is the angle of the panel 4Inflation Reduction Act (2022) Impacts on Solar Installations 5Motivation • How much difference do different sources of annualized albedo produce? • Given different albedo estimates, how much does a generic bifacial project vary in energy? • Is there a relationship between differences in albedo and the resulting difference in energy? • Do on-site measurements capture typical long-term albedo conditions? 6Albedo Measurement and Modeling 7Satellite Modeled Albedo • Services offering solar resource modeling from satellite sources have begun to offer albedo estimates as well in the last few years • Each provider uses custom algorithms for albedo modeling, often involving imagery resolutions of 1-4km, and historical and modeled snow fall timeseries data • Satellite modeling can offer long-term solar resource (20+ years) at hourly or finer temporal resolutions; however, albedo statistics are generally reported as monthly (typical/average year or specific years) • This study tests PVsyst albedo default of 0.20, 3 monthly average albedo sources (A, B, C) and a monthly average of all three models • Additionally, IAV statistics are generated from 2 timeseries based albedo datasets (Models 1 and 2) 8Solar Meteorological Stations (SMS) • SMS are used to obtain point location solar resource, often at a future development site • Stations are often deployed for 1-2 years (See our MCP poster for deployment length uncertainty impact on long term GHI estimates!) • Solar resource often uses a combination of long-term satellite modeling and on-site measurements to derive a long-term adjusted solar resource values using a Measure- Correlate-Predict (MCP) approach • Stations often include several pyranometers, including a downward-facing pyranometer to calculate albedo 9Sample Locations 10UL Solutions Field Services • UL’s albedo calculation is derived from high- frequency pyranometer data • Having collected albedometer data at over 150 locations, UL has a strong understanding of irradiance and albedo distribution across the United States • From this data, it is possible to understand regional differences in albedo • Extensive manual validation and QAQC practices remove systematic biases and erroneous values due to dew, frost, unlevelness, soiling and other issues 11Locations of 34 Sample Sites with On-Site Albedo 12Energy Modeling Assumptions 13PVsyst Setup General • Single inverter block simulations in PVsyst • Generic PAN and OND from PVsyst library • 440W Mono-PERC panel with 80% bifaciality • Fresnel Anti-reflective coating • 1.6m height above ground • Unlimited trackers for bifacial and shading simulations Site-Specific • Site (.SIT) and Meteo (.MET) files – Meteo from solar resource MCP analysis where available • Soiling losses from dust and snow – Bifacial benefit from snow reflection and penalty from backside soiling included 14Modeling Assumptions • Project-level albedo specification remains default; no albedo benefit to front-side for these simulations • Losses from complex terrain and specific tracker shading not included • Horizon profile excluded from PVsyst as horizon impacts are accounted for from on-site solar resource data MCP analysis • Standard assumptions for electrical and transformer losses • No post-processing losses applied, such as POI clipping, wind stow, sub-hourly clipping, and availability • Energy evaluated on a First-Year basis; no material degradation 15Base Cases - 34 Sites Albedo Energy 16Results 17Differences from On-Site Scenario for 34 Sites Albedo Energy 18Differences from On-Site Scenario for 34 Sites (Tabular) Albedo Default (0.20) Model A Model B Model C Average mean -0.01 -0.01 -0.03 -0.01 -0.03 max 0.02 0.02 0.01 0.06 0.02 min -0.09 -0.09 -0.09 -0.07 -0.09 std 0.03 0.02 0.02 0.03 0.02 Energy 19 Default (0.20) Model A Model B Model C Average mean -0.4% -0.1% -0.7% -0.2% -0.3% max 0.5% 0.5% 0.1% 1.2% 0.2% min -2.6% -1.8% -2.2% -1.2% -1.3% std 0.6% 0.5% 0.5% 0.5% 0.3%20Model 1 Albedo IAV by State (1-8 sites per state) 21 Higher IAV (~0.04/yr std dev) Lower IAV (~0.00/yr std dev)Model 2 Albedo IAV by State (1-8 sites per state) 22 Higher IAV (~0.04/yr std dev) Lower IAV (~0.00/yr std dev)IAV over Two Years of On-Site Data • 5 sample sites were selected with two years of albedo data • Note that these data are a small sample of both sites and years of operation and serve to demonstrate potential areas of risk regionally regarding on-site albedo • Sites in the low IAV regions identified in the two long-term albedo estimates also show low IAV over this two-year period, whereas sites in regions identified as greater IAV potential also show similar results • Maximum monthly differences reported here are primarily winter and demonstrate the variability of snowfall dates Site 1 (CA) 2 (PA) 3 (PA) 4 (TX) 5 (OH) Year 1 0.23 0.29 0.28 0.19 0.25 Year 2 0.22 0.24 0.22 0.18 0.24 Avg 0.22 0.26 0.25 0.19 0.24 ± 0.00 0.02 0.02 0.01 0.01 Max Monthly 0.02 0.13 0.13 0.07 0.13 23Conclusions 24Conclusions • Annual albedo estimates differ from on-site measurements on average by - 0.017 ±0.023 across three models, but only translates to a -0.3% ±0.5% Energy difference • Averaging these three models monthly before analysis maintains the same average difference, but further reduces standard deviation to ±0.015 albedo and ±0.3% Energy • Regressions of albedo difference to energy impact suggest a 1/6 factor for this experiment, i.e., 0.06 annual albedo difference = 1% Energy difference • On-site albedo measurement of only 1 year does not inherently capture inter-annual variability, which 2 models predict 0.00-0.04 albedo standard year-to-year differences; 2 years of albedo data at highly varied sites in snow regions may provide additional coverage • In these high IAV environments on-site albedo measured at max or min could result in ~1% energy difference from a typical albedo year 25Motivation Revisited • How much difference do different sources of annualized albedo produce? 0.017 ±0.023 • Given different albedo estimates, how much does a generic bifacial project vary in energy? -0.3% ±0.5 • Is there a relationship between differences in albedo and the resulting difference in energy? Potentially, 1/6 for this system • Do on-site measurements capture typical long-term albedo conditions? Depends on the IAV of the region and length of on-site measurement campaign 26Future Work 27Future Work Albedo MCP • Measure-Correlate-Predict analyses have produced strong solar resource long-term results • High quality modeled albedo datasets may be used in conjunction with on-site measurements to produce statistical regressions on a monthly basis to capture the effects of IAV and measurement time periods • Best applied to regions with high IAV but verifying the low effects to low IAV regions may also provide a benefit Sub-Monthly Modeling • On-site albedo measurements can provide finer temporal resolution than monthly • Modeling daily or even hourly may produce different results than this study • Likely would need pvlib to customize this albedo application • Unlikely to affect current energy modeling practices but understanding the effects can help better categorize the effects 28© 2023 UL LLC. All Rights Reserved. Thank you UL.com/SolutionsQuestions? UL.com/Solutions Halley Darling halley.darling@ul.com Renn Darawali renn.darawali@ul.com Lucila D. Tafur lucy.tafurgamarra@ul.com 30