Ransome_PVPMC_webinar_20_08_05

www.steveransome.com 6-Aug -20 1 How to use the Loss Factors and Mechanistic Performance Models effectively with PVPMC/PVLIB. Steve Ransome (SRCL, UK) www.steveransome.com mailto: steve@steveransome.com PVPMC Webinar on PV Performance Modelling Methods: 5 th Aug 2020 Acknowledgements : Juergen Sutterlueti (Gantner Instruments) for measurement datawww.steveransome.com 6-Aug -20 2 Contents of this talk 1. Introductions to Loss Factors Model and Mechanistic Performance Model (MLFM) 2. Performance and degradation analysis methods using MLFM and sample data from NREL and Gantner Instruments 3. Functionality of MLFM to be added to PVLIB python trying to fit in with their naming convention PV Modeling Glossarywww.steveransome.com 6-Aug -20 3 1 -diode model fit limitations 1. The 1 -diode model doesn’t always fit measured IV curves well – there are 6 knowns but 5 coefficients 2. Assumptions need to be made on how I 01 varies with temperature and R SHUNT with irradiance otherwise predicted low light performance and temperature coefficients may be wrong 3. Imperfect fit to smooth IV curve Residual fits 5 knowns but has wrong slope @V=Vmp (2% error) [A] 4. Imperfect fit to IV with mismatch Residualhas oscillations so loss of information on mismatch [B]and fitted R SHUNT value lower than expected Residual [A] Residual Measure -Fit (%) Current (A) Voltage (V) [B] Current (A) (3) (4)www.steveransome.com 6-Aug -20 4 Loss Factors Model -area and technology independent “Extracting normalised independent losses from the shapes of IV curves” ❶ Normalise IV curve currents and voltages by datasheet STC values so ref.imp.norm = 1 and ref.vmp.norm =1 ❷ Multiply by 1/ FF_ref to get to isc* voc ❸ Find Current losses ref.isc à meas.imp ❹ Find Voltage losses ref.voc à meas.vmp ❺ PR DC (= meas.pmp /ref.pmp /gi ) Any changes with time show degradation and cause • Some definitions updated since the original LFM paper in 2011 EUPVSEC Hamburg. • Naming convention in PV Modelling Glossary • For more information on spectral and reflectivity corrections not covered here PVSC46_Chicago.pdf PRdc = 1/ FF_ref* norm[( isc* rsc* ffi ) * ( ffv * roc * voc_Tcorr * t_mod )] www.steveransome.com 6-Aug -20 5 LFM curves vs. irradiance identify performance limits (NREL CdTe 1 year) Irradiance G I (kW/m 2 )→ If we don’t have soiling, spectral and aoi corrections for isc we can self reference. PRdc is then proportional to the product of the remaining LFM coefficients The 3 coefficients shown mainly determine the shape of PRdc vs. irradiance ↙ Voc_Tcorr (2) ↙ ~R SHUNT (1) ~ R SERIES ↘(3) MLFM Performance Value (%)→ PRdc !1/ FF_ref* norm[( isc* rsc* ffi )* ( ffv * roc * voc_Tcorr * t_mod )] www.steveransome.com 6-Aug -20 6 Mechanistic Performance Model – location and technology independent ”Fits PV performance vs. irradiance and temperature with robust coefficients” • C 1 = Tolerance = ( meas /ref) • C 2 = Temperature coefficient (e.g. gamma_pmp ~ -0.4%/K) • C 3 = low light loss (caused by Voc(G I )) • C 4 = high light loss (caused by R SERIES ) • Optional extra coefficients IF NEEDED C 5 = WS, C 6 =some modules R SHUNT C X = Beam Fraction/AOI, Spectral, non -linearity • Degradation studies – quantify changes with coefficients over time • Fault finding – coefficients that glitch or aren’t expected values • PVPVC 2017 Canobbio paper Predicted (lines) vs. Measured (dots) LFM values fitted with MPM (7 years cSi module in Tempe AZ) Reference :PVSC46_Chicago.pdf Irradiance G I (kW/m 2 )→ MLFM Performance Value (%)→ “#“$ % IV curve every minute, all environmental sensors, spectral parameters PV Module Power up to 500W/800W High quality digitalization, current accuracy 0.1% FS, voltage: 0.05% FS Scalable system (4 48 channels) with raw data access Local or cloud -based data streaming Derived parameters using Loss Factors and Mechanistic Performance Models Integrated Python Jupyter Lab for direct analysis and automatic reporting Continuous measurements in Arizona since 2010; Other sites available around the world Trusted by leading PV Module manufacturers, Technology providers and Research Labs 2D Tracker Sensors Fixed orientation Juergen to add his advert for OTF sales or datawww.steveransome.com 6-Aug -20 18 Summary • Loss factors model – meaningful, independent coefficient analysis of IV curve performance • Mechanistic Performance Model – fits meaningful, independent coefficients to LFM/PRdc/Matrices vs. irradiance and module temperature • Good quality data allows a lot of understanding better than 1 -diode model • Graphical analysis shown both multiplicative and stacked losses to find cause of any underperformance and quantify degradation • Code will be introduced into PVLIB • We are doing much more analysis that hasn’t been covered in this short talk e.g. mismatch, spectral, aoi/beam fraction … see our previous talks. • Feedback welcome !www.steveransome.com 6-Aug -20 19 Contact Steve Ransome (SRCL, UK) www.steveransome.com ; mailto: steve@steveransome.com Acknowledgements: Juergen Sutterlueti (Gantner Instruments) Contact us for OTF enquiries and high -quality data sets for your own research www.gantner -instruments.com/products/software/gi -cloud/ Thank you for your attention All papers are either available through SRCL website or writing to me directly Also NREL DATA https://www.nrel.gov/docs/fy14osti/61610.pdf NAMING CONVENTION https://duramat.github.io/pv -terms/ PVPMC https ://pvpmc.sandia .gov/