7-5-GroundWork_PVPMC-SoilingDataProcessing_2022-08-24

© GroundWork Renewables, Inc. | All Rights Reserved Soiling Measurement Data Processing to Support Production Loss Prediction Julie Chard, Josh Peterson, Justin Robinson August 24, 2022 15th Annual PV Performance Modeling Collaborative© GroundWork Renewables, Inc. | All Rights Reserved 2 GROUNDWORK IS… 1,000’s Field visits #1 in US 45GW+ 500 Bankable Datasets Mission Driven To accelerate the success of solar in leading utility-scale energy production Founded in Values Trustworthy the other module is allowed to naturally soil • Soiling reduces the short circuit current (Isc) output of the soiled module relative to the clean control Maintenance and Ancillary Measurements • Weekly maintenance • Manual and natural resets • Concurrent rain measurement • One-minute data© GroundWork Renewables, Inc. | All Rights Reserved 6 Soiling Ratio Soiling Rate 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑅𝑅 𝑅𝑅 𝑅𝑅 𝑆𝑆 𝑆𝑆 = 𝐼𝐼𝐼𝐼 𝑐𝑐 𝑠𝑠𝑠𝑠 𝑠𝑠 𝑠𝑠 𝑠𝑠 𝑠𝑠 𝐼𝐼𝐼𝐼 𝑐𝑐 𝑐𝑐 𝑠𝑠 𝑠𝑠 𝑐𝑐𝑐𝑐 An instantaneous measure of the reduction in Isc due to soiling. A measure of how fast soiling is accumulating on the modules. 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑅𝑅𝑅𝑅 𝑅𝑅 𝑅𝑅 = Δ 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑅𝑅 𝑅𝑅 𝑅𝑅 𝑆𝑆 𝑆𝑆 Δ 𝑇𝑇𝑆𝑆 𝑇𝑇 𝑅𝑅 Soiling Loss 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝐿𝐿 𝑆𝑆 𝐼𝐼𝐼𝐼 % = 1 − 𝑆𝑆𝑅𝑅𝑅𝑅 𝑅𝑅 𝑆𝑆 𝑆𝑆 ∗100 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝐿𝐿 𝑆𝑆 𝐼𝐼𝐼𝐼 % ≈ 𝑓𝑓 ( 𝑆𝑆𝑅𝑅𝑅𝑅 𝑅𝑅 𝑅𝑅 , 𝑃𝑃𝑃𝑃 , 𝑊𝑊 𝑅𝑅𝑅𝑅 𝑅𝑅 𝑊 𝑅𝑅 𝑊𝑊 ,?) SOILING MEASUREMENT: TERMINOLOGY A derate of power production due to soiling. In preconstruction, loss is predicted using models, where soiling rate is an input. In post-construction, loss is expressed as a function of the soiling ratio. Post-construction soiling rates can be calculated retrospectively, working backward from loss. © GroundWork Renewables, Inc. | All Rights Reserved 7 GOVERNING STANDARD AND PUBLISHED FILTERING APPROACHES IEC 61724-1:2021 Photovoltaic system performance - Part 1: Monitoring, Annex C IEC 61724-1:2021 • Data are filtered • A daily average is computed • A quality number is obtained Published Filtering Approaches • Bessa et al., 2022. Estimation of photovoltaic soiling using environmental parameters: A comparative analysis of existing models. Advanced Sustainable Systems, 2100335. • Deceglie et al., 2018. Quantifying soiling loss directly from PV yield. IEEE Journal of Photovoltaics, 8(2), 547-551. • Skomedal et al., 2019. Endogenous soiling rate determination and detection of cleaning events in utility- scale PV plants. IEEE Journal of Photovoltaics, 9(3), 858-863. • Micheli et al., 2021. Improved PV soiling extraction through the detection of cleanings and change points. IEEE Journal of Photovoltaics, 11(2), 519-526. • Muller et al., 2022. Automated detection of photovoltaic cleaning events: A performance comparison of techniques as applied to a broad set of labeled photovoltaic data sets. Progress in Photovoltaics: Research and Applications, 30(5), 567-577. • Peterson et al., 2022. Extraction of prevailing soiling rates from soiling measurement data. IEEE 49th Photovoltaic Specialists Conference, PVSC 2022. “the instantaneously measured values shall be integrated to compute a daily average value… the data should first be filtered. the number of data points passing the filter should be recorded as a quality metric and calculation of the daily average should only be performed when a sufficient number of data points are valid”© GroundWork Renewables, Inc. | All Rights Reserved 8 DAY-TO-DAY CHANGES IN DAILY VALUES A quality number is helpful. Error bars are more helpful.© GroundWork Renewables, Inc. | All Rights Reserved 9 SOIING DATA MUST BE FILTERED Insufficient Irradiance Soiling ratios are skewed when module outputs are smaller. Angular Response Data are less accurate when angles of incidence are larger. Data are skewed across the solar noon window when modules aren’t coplanar. Outliers Anomalous instantaneous soiling ratios skew daily averages. Clean Control The clean module doesn’t stay clean between maintenance events. Inherent Noise Variable sky conditions, variable module surface characteristics, birds flying over, etc. Challenges with Soiling Data© GroundWork Renewables, Inc. | All Rights Reserved 10 INSUFFICIENT IRRADIANCE Common irradiance threshold = 500 W/m 2 Fixed thresholds are exceeded more often in summer than winter, and more often in equatorial than polar locations Limited irradiance = limited data Dynamic threshold, based on the modeled extraterrestrial GHI value (ETR): G Max(200 W/m 2 , 50% ETR) • G is the effective irradiance measured by the clean module • G should be greater than 50% of ETR. • Hard lower limit of 200 W/m 2 for extreme north/south locations© GroundWork Renewables, Inc. | All Rights Reserved 11 Remove outliers within the day INHERENT NOISE AND OUTLIERS – INTRA-DAY VARIABILITY Outliers Consider each day independently: P50 – 2*(P50 – P5) SRatio P50 + 2*(P95 – P50) • Find the P5, P95, and median (P50) soiling ratio values • Remove points that are less than twice the distance from the median to the P5 • Remove points that are more than twice the distance from the median to the P95© GroundWork Renewables, Inc. | All Rights Reserved 12 INHERENT NOISE AND OUTLIERS – INTER-DAY VARIABILITY Trend should be continuous forward or backward Soiling ratio should change slowly from day to day Too far away from local trends© GroundWork Renewables, Inc. | All Rights Reserved 13 ANGULAR RESPONSE – MODULE MISALIGNMENT Soiled Module Tilted To East Flattening Adjustment Morning: • Larger Isc soiled • Larger soiling ratio Soiling ratio continually decreases over the solar noon window Afternoon: • Smaller Isc soiled • Smaller soiling ratio Measured soiling ratios for a given day are adjusted to be more constant over the solar noon window Noise is preserved Reduced overall spread in soiling ratios = Reduced uncertainty West Clean Module East Soiled Module 0.5°© GroundWork Renewables, Inc. | All Rights Reserved 14 CLEAN CONTROL Select data immediately after cleaning event Between maintenance visits the “Clean” module accumulates dust at the same rate as the “Soiled” module. Because both modules accumulate dust at the same rate, the soiling ratio remains constant. When the “Clean” module is no longer clean, the soiling ratio is invalid. Only data immediately after a cleaning event should be selected for analysis.© GroundWork Renewables, Inc. | All Rights Reserved 15 BENEFITS OF DATA FILTERING Insufficient irradiance Inter-day outliers Intra-day outliers Failed previous filter Failed current filter Passed current filter Maintenance event Clean module clean? Flattening adjustment© GroundWork Renewables, Inc. | All Rights Reserved 16 BREAKING THE DATA INTO SOILING PERIODS Soiling periods are bracketed by reset cleaning events Reset events may be manual or natural. → Manual events are triggered by uncharacteristic localized soiling (e.g. bird droppings). → Natural events include rain, wind and/or dew. Detected using the stochastic rate and recovery (SRR) method [1] . When the soiled module is cleaned, the soiling ratio jumps up. This resets the soiling period. [1] Deceglie et al., 2018. Quantifying soiling loss directly from PV yield. IEEE Journal of Photovoltaics, 8(2), 547-551. When the soiled module is cleaned, the soiling ratio jumps up.© GroundWork Renewables, Inc. | All Rights Reserved 17 SOILING RATE DETERMINATION: SLOPES BETWEEN PAIRS OF POINTS The soiling rate between two points is the slope of the line connecting them. Lines are drawn between all valid pairs of points and the corresponding slopes are computed. → Pairs of points must fall within the same soiling period → ~ 500K slopes/year with one-minute resolution data From the collection of slopes, the median soiling rate is determined for a given time period. An estimate of the soiling rate uncertainty is given by the range in the slope values for the specified time period. 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑅𝑅𝑅𝑅 𝑅𝑅 𝑅𝑅 = Δ 𝑆𝑆 𝑆𝑆 𝑆𝑆 𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑅𝑅 𝑅𝑅 𝑅𝑅 𝑆𝑆 𝑆𝑆 Δ 𝑇𝑇𝑆𝑆 𝑇𝑇 𝑅𝑅© GroundWork Renewables, Inc. | All Rights Reserved 18 SOILING RATES WITH THE LOWEST POSSIBLE UNCERTAINTY Proper filtering significantly improves the quality of soiling data. Mean annualized soiling rate uncertainty for 32 GroundWork soiling measurement stations across the United States = 0.00083.© GroundWork Renewables, Inc. | All Rights Reserved 19 THANK YOU! Q A Questions? Contact me: jchard@grndwork.com