UL_MultipleSatelliteModelsMCP_2020-06-18_Slides

UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential. Eva Plaza Sanz – Senior Renewable Energy Analyst Peter Johnson – Senior Project Engineer Lucy Tafur-Gamarra – Renewable Energy Analyst Multiple Satellite Models for On-Site Long-Term References UL and the UL logo are trademarks of UL LLC 2020. Proprietary Confidential. 3 Global HQ Chicago Advisory US Offices Albany, San Diego Using multiple satellite models as long-term references for measure-correlate-predict analyses 1 improves accuracy of results and 2 mitigates prediction risk. UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential. 6 Presentation Outline Measure-Correlate-Predict Approach MCP with On-Site Weather Stations MCP with Operational Energy Estimates Why Multiple References are Important UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential. 7 Presentation Outline Measure-Correlate-Predict Approach MCP with On-Site Weather Stations MCP with Operational Energy Estimates Why Multiple References are Important 1. Measure collect one year of high quality irradiance/power measurements 2. Correlate to long-term reference data from multiple satellite models 3. Predict adjust modeled estimate for observed model bias 4. Uncertainty for measurements, correlation, long-term adjustment Measure-Correlate-Predict Approach On-Site Data High Accuracy Short POR 1 yr Reference Source Moderate Accuracy Long POR 20 yrs Tuned TMY High Accuracy Represents Long-Term UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential. 10 Presentation Outline Measure-Correlate-Predict Approach MCP with On-Site Weather Stations MCP with Operational Energy Estimates Why Multiple References are Important Two Class A pyranometers heating/ventilation Weekly maintenance for cleaning and leveling Weekly meteorological desktop screening At least one complete year of measurements, which mitigates seasonality risk o Modeled seasonal biases may differ from annual biases in magnitude and direction o Shorter PORs result in more uncertainty than models themselves UL and the UL logo are trademarks of UL LLC 2018. Proprietary Confidential. 11 Solar Met Measurements UL and the UL logo are trademarks of UL LLC 2018. Proprietary Confidential. 12 Measure-Correlate-Predict with Satellite Models 1450 1500 1550 1600 1650 1700 1750 1800 1850 1900 I r r a d i a t i o n k W h / m 2 / y e ar Year GHI UL and the UL logo are trademarks of UL LLC 2018. Proprietary Confidential. 13 Measure-Correlate-Predict with Satellite Models Establish linear relationship y m*x b R-squared strength of correlation Input modeled resource Output long-term result Case study UL reviewed 8 long-term Solar Resource Assessments using on-site data o Locations across contiguous USA o Three satellite modeled reference datasets for MCP Comparisons  Climate of on-site period compared to long-term  Predictions from individual models vs. average of three models  Risk of using one model vs. three models On-Site Solar Resource Assessment with MCP MCP with On-Site Weather Stations -1 0 1 2 3 4 5 6 7 8 9 10 CA1 CA2 OH1 OH2 TX IL AZ NM Climate Adjustment from Individual Models Observed Low Resource Period Model 1 Model 2 Model 3 MCP with On-Site Weather Stations -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 CA1 CA2 OH1 OH2 TX IL AZ NM MCP Results from Individual Models vs. Three Models Model 1 Model 2 Model 3 All models usually show same direction of climate adjustment e.g., observed periods being high or low On average, long-term estimates ranged by 1.4 depending model selection maximum difference of 2.9 Sites in deciduous environments showed more of a range than sites in desert environments  greater need for model diversity depending on environment Multiple models for MCP mitigates risk of relying on one single model MCP with On-Site Weather Stations UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential. 19 Presentation Outline Measure-Correlate-Predict Approach MCP with On-Site Weather Stations MCP with Operational Energy Estimates Why Multiple References are Important Correlation between POA irradiance and monthly energy Monthly correlations less data points  higher correlation uncertainty Adjust for availability, curtailment, one- off events, degradation Can be completed on individual projects or a portfolio of projects UL and the UL logo are trademarks of UL LLC 2018. Proprietary Confidential. 20 MCP with Operational Energy Estimates Case study UL completed OEPEs at approximately 20 projects across a large region of the world o MCP results for two separate modeled datasets o Evaluated 1 as individual projects, and 2 as a portfolio Hypotheses 1. Does one reference dataset yield stronger correlations 2. How different are results on a project-by-project basis 3. How different are results on a portfolio basis MCP with Operational Energy Estimates MCP with Operational Energy Estimates 60 70 80 90 100 A B C D E F G H I J K L M N O P Q R S R-Squared Comparison - Model 1 vs. Model 2 Model 1 Model 2 1. Strength of correlation Evaluating multiple reference models may result in excluding models at some sites when they have weaker correlations MCP with Operational Energy Estimates -1.0 -0.5 0.0 0.5 1.0 A B C D E F G H I J K L M N O P Q R S MCP Energy Results from Individual Models vs. Two-Models Model 1 Model 2 2. Difference in results for individual projects For individual projects, energy accuracy is improved by multiple models max difference of 1.8. MCP with Operational Energy Estimates 0 1 2 3 4 5 A B C D E F G H I J K L M N O P Q R S Annual Energy Uncertainty - MCP with Individual Models vs. Two Models Model 1 Model 2 Both Models 3. Uncertainty Reduced by 1-1.5 by using two models rather than a single model. 3. Difference in results for portfolio More projects greater modeled data cost Portfolio energy differed by 0.3 within uncertainty of analysis Uncertainty already mitigated by regional variation and portfolio benefit Conclusion portfolio analyses may be able to save on cost and effort by using a single model for the entire portfolio. MCP with Operational Energy Estimates UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential. 28 Presentation Outline Measure-Correlate-Predict Approach MCP with On-Site Weather Stations MCP with Operational Energy Estimates Why Multiple References are Important Using multiple satellite models as long-term references for measure-correlate-predict analyses 1 improves accuracy of results and 2 mitigates prediction risk. 31 Questions UL and the UL logo are trademarks of UL LLC 2019. Proprietary Confidential Eva.PlazaSanzUL.com Peter.JohnsonUL.com