Performance Loss of PV systems. Giorgio Belluardo

Transcription

1 Performance Loss of PV systems Giorgio Belluardo

2 Content Importance of accurate estimation of PL Mechanisms behind performance loss Statistics Methodologies to assess PLR Novel method for estimation of PLR from field data and comparison with other common methods Future work Conclusions Giorgio Belluardo 2

3 Importance of accurate estimation of PL Investor side: need for accurate estimations of PL from installed PV systems and accurate module warranties foster bankability of PV systems PV components manufacturers side: need to collect as much field data as possible to improve products and establish warranty schemes Accuracy of PV module warranty MANUFACTURERS INVESTORS Field data Giorgio Belluardo 3

4 Importance of accurate estimation of PL Impact on LCOE Simple formula considering fixed rate of PL Monte Carlo simulation 100 kw PV NREL Colorado 40 tilt, South 17.3 c/kwh 18.2 c/kwh 16.7 c/kwh 18.4 c/kwh Starting point (initial production) Rate Pattern (linearity/non-linearity) Jordan, D.C., Kurtz, S.R., VanSant, K., Newmiller, J., Compendium of photovoltaic degradation rates. Progress Giorgio in Belluardo Photovoltaics: Research and Applications 4

5 Mechanisms behind Performance Loss Performance loss degradation quality of manufacturing & technology climate (cycles of temperature, humidity, irradiation) mounting system (mechanical stress, ventilation) soiling snow shading modules and cells mismatch wrong system design MATERIAL DEGRADATION Corrosion of metallic connections delamination discoloration (micro)cracks of cells hot spots bubbles etc. ELECTRICAL MIS-PERFORMANCE PERFORMANCE LOSSES Giorgio Belluardo 5

6 Some facts from a statistic analysis of values of PLR Jordan, D.C., Kurtz, S.R., VanSant, K., Newmiller, J., Compendium of photovoltaic degradation rates. Progress in Photovoltaics: Research and Applications PLR depending on technology Median PLR: c-si: %/year, stable HIT & µc-si: ~1%/year a-si: 1-1.5%/year from 1990s CIGS: 0.8-1%/year from 2000s CdTe: higher degradation (>1.5 %/year) PLR depending on climate Hotter climates and mounting configurations that lead to sustained high temperatures may lead to higher PLR in some, but not all, products Giorgio Belluardo 6

7 Methodologies to assess PLR Equipment Type of measurement Methodology INDOOR Solar Simulator I-V tracer Spot measurement of I-V curve & STC Spot measurement of I-V curve & STC Comparison with initial or nameplate values OUTDOOR Inverter or dedicated system MPP Continuos measurement of I-V curve & Pmax Continuous measurement of Pmax Adoption of a performance metric Adoption of a statistical method The methodology influences the value of PLR and the associated uncertainty Spot measurements of I-V STC outdoor show greater PL and higher spread than indoor (measurement uncertainty + soiling) Giorgio Belluardo 7

8 Methodologies to assess PLR Equipment Type of measurement Methodology INDOOR Solar Simulator I-V tracer Spot measurement of I-V curve & STC Spot measurement of I-V curve & STC Comparison with initial or nameplate values OUTDOOR Inverter or dedicated system MPP Continuos measurement of I-V curve & Pmax Continuous measurement of Pmax Adoption of a performance metric Adoption of a statistical method Giorgio Belluardo 8

9 Methodologies to assess PLR PLR evaluation using continuous measurements of Pmax Adoption of a performance metric (+ filtering/correction technique) Selection of a time scale Performance Ratio, PVUSA, Pmax, etc. weekly, monthly, etc. Adoption of a statistical method mathematical algorithms applied on the time series of performance estimators in order to extract a trend: Linear regression: possible initial decline (esp. for thin films) + stable PL Classical series decomposition (CSD) Locally weighted scatterplot smoothing (LOESS) Autoregressive integrated moving average (ARIMA) Etc. Giorgio Belluardo 9

10 Methodologies to assess PLR PLR evaluation using continuous measurements of Pmax 1. Select a performance index 2. Calculate monthly value 3. Linear regression: y=ax+b 4. PLR = 12a/b (%/year) regression uncertainty depends on: Seasonality (the lower the better) Number of available points (the higher the better) Giorgio Belluardo 10

11 Novel method for the evaluation of PLR from field data and comparison with other common methods Giorgio Belluardo 11

12 Methodology Selected performance metric: array generated power (P max dc side inverter reading) 0. Filtering: G > 800W/m 2, PPPP σ < PPPP < PPPP + σ 1. Correction for irradiance and temperature to Standard Test Conditions (G = 1000W/m 2, T = 25 C) G : irradiance on the plane of modules G STC : irradiance at STC (1000 W/m2) γ: temperature coefficient (%/ C) Tcell: module temperature T STC : module temperature at STC (25 C) T: ambient temperature NOCT: Nominal Operating Cell Temperature (from datasheet) T NOCT : temperature at NOCT conditions G NOCT : irradiance at NOCT conditions 15-min based values Giorgio Belluardo 12

13 Methodology 2. Spectral correction to Standard Test Conditions (AM1.5) a. simulation of solar spectra at ABD and calculation of average wavelength as index of spectrum shape and shift Validation of simulations Average wavelength Giorgio Belluardo 13

14 Methodology 2. Spectral correction to Standard Test Conditions (AM1.5) b. correction to AM=1.5 P G,Tcorr : P max corrected to irradiance and temperature λ ave : average wavelength G STC : irradiance at STC (1000 W/m2) AM; air mass c,d,e,f: parameters Giorgio Belluardo 14

15 Methodology 2. Spectral correction to Standard Test Conditions (AM1.5) Irradiance, temperature + spectral correction 15-min values, one year span, filtered data Giorgio Belluardo 15

16 Methodology 24 groups at ABD investigated 9 different technologies 3 years data ( , 6 th to 41 th operation month) Novel method (P max corrected to STC) Most common methods (PR and PVUSA) Performance Loss Rate, regression uncertainty Giorgio Belluardo 16

17 Results Performance metric / - Performance metric / - month month Crystalline silicon Thin film Giorgio Belluardo 17

18 Results Crystalline silicon Thin films Giorgio Belluardo 18

19 Results PR PVUSA Pmax_STC crystalline: novel method reduces uncert. from min. 18% to max. 84% (average 64%) Crystalline silicon average PLR values: max. -0.6%/year (ribbon) Thin film Thin film: unc. levels comparable to other methods average PLR values: microm.: -1.2%/year CIGS: -2.2%/year CdTe: -1.6%/year 1j-a-Si: -1.1%/year 2j-a-Si: -1.9%/year 3j-a-Si: gains likely annealing Giorgio Belluardo 19

20 Results Dependence on data time span Results converge, uncert. decreases Min. 3 years of data necessary Giorgio Belluardo 20

21 Results Novel method as result of optimization process (U2) (U4) 25% less uncertainty (on average) by applying spectral correction Giorgio Belluardo 21

22 Outlook New methodologies investigated (ARIMA) > 5 years data available Giorgio Belluardo 22

23 Future work Comparison of PV systems in different climates (Köppen Geiger climate classification): Italy (Bolzano and Milano): Temperate, without dry season Italy (Catania): Temperate, dry summer Cyprus (Nicosia) and Australia (Alice Springs): Arid, steppe In collaboration with: RSE University of Cyprus CAT Projects Giorgio Belluardo 23

24 Conclusions Importance of accurate estimation of performance loss Overview of methodologies Statistical analysis on values of PLR: dependency on technology, climate, methodology Novel method considerably reduces uncertainty than other common estimation techniques, for crystalline-silicon systems (60% on average) No improvement is seen for thin films Applying spectral correction generate benefit to method overall accuracy (25% on average) Outlook: new methods using data from 5 different locations Giorgio Belluardo 24

25 Thank you for the attention Giorgio Belluardo