【PVPMC】EURAC-Lindig-性能损失率计算的理论与实践

Sascha LindigEurac Research - University of LjubljanaTheory STL – seasonal and trend decomposition using LOESS, YoY – Year on Year method S. Lindig, PVPMC China 2 0 1 9 2 01 0 /1 2 /1 9 Comparison of calculation methodologiesHigh quality datasets – Metrics S. Lindig, PVPMC China 2 0 1 9 2 11 0 /1 2 /1 9 Comparison of calculation methodologiesHigh quality datasets – Methodologies 1 . R. B. Cleveland, W. S. Cleveland, J. E. McRae, I. Terpenning, STL: A Seasonal-Trend Decomposition Procedure Based on Loess, Journal of official statistics 6 (1 ) (1 9 9 0 ) 3 –3 3 .2 . Sascha Lindig, David Moser, Alan Curran, Roger French, Performance Loss Rates of PV systems of Task 1 3 database, IEEE PVSC (2 0 1 9 ).3 . Ernest Hasselbrink, Mike Anderson, Zoe Defreitas, Mark Mikofski, Yu-Chen Shen, Sander Caldwell, Akira Terao, David Kavulak, Zach Campeau, David DeGraaff, Validation of the PVLife Model Using 3 Million Module-Years of Live Site Data, in: IEEE 3 9 th Photovoltaic Specialists Conference (PVSC), IEEE, 2 0 1 3 : pp. 0 0 0 7 –0 0 1 2 . https://doi.org/1 0 .1 1 0 9 /PVSC.2 0 1 3 .6 7 4 4 0 8 7 . Seasonal Decomposition of Time Series using Loess (STL)1,2 Year on Year (YoY)3 S. Lindig, PVPMC China 2 0 1 9 2 21 0 /1 2 /1 9 Comparison of calculation methodologiesHigh quality datasets – BenchmarkingVAR: yearly variation of powerYoY: Year on Year – Comparison of daily power valuesLR: Linear RegressionSTL: Seasonal Time Series Decomposition using LoessC S D : C l a s s i c a l S e a s o n a l DecompositionSCSF: Statistical clear sky fitting YbY: Year-by-Year comparison Power NOCT EURAC PV-systemPoly-crystalline silicon S. Lindig, PVPMC China 2 0 1 9 2 31 0 /1 2 /1 9 Comparison of calculation methodologiesHigh quality datasets – Benchmarking EURAC PV-systemPoly-crystalline siliconVAR: yearly variation of powerYoY: Year on Year – Comparison of daily power valuesLR: Linear RegressionSTL: Seasonal Time Series Decomposition using LoessC S D : C l a s s i c a l S e a s o n a l DecompositionSCSF: Statistical clear sky fitting YbY: Year-by-Year comparison Power NOCT S. Lindig, PVPMC China 2 0 1 9 2 41 0 /1 2 /1 9 Comparison of calculation methodologiesHigh quality datasets – BenchmarkingVAR: yearly variation of powerYoY: Year on Year – Comparison of daily power valuesLR: Linear RegressionSTL: Seasonal Time Series Decomposition using LoessC S D : C l a s s i c a l S e a s o n a l DecompositionSCSF: Statistical clear sky fitting YbY: Year-by-Year comparison Power NOCT S. Lindig, PVPMC China 2 0 1 9 2 51 0 /1 2 /1 9 Low Medium High S. Lindig, PVPMC China 2 0 1 9 2 61 0 /1 2 /1 9 S. Lindig, PVPMC China 2 0 1 9 2 71 0 /1 2 /1 9 Comparison of calculation methodologiesDigital power plant K. Radouane, A. Lindsay, P. Dupeyrat, B. Braisaz, An advanced model of pv power plants based on modelica, in: 2 8 th EU PVSEC Proceedings, Paris, France, 2 0 1 3 (Sep. 2 0 1 3 ). § Simulated PV data§ Total of 4 systems2 x 5 years repeated weather data2 x 4 years weather data + 5 th year colder weather§ PLR for each of the 4 cases2 x PLR = 0 %/a2 x defined linear PLR valueMakes it possible to validate methods against a KNOWN PLR value S. Lindig, PVPMC China 2 0 1 9 2 81 0 /1 2 /1 9 Comparison of calculation methodologiesDigital power plant – NO degradation – same yearVAR: yearly variation of powerYoY: Year on Year – Comparison of daily power valuesLS-LR: Least Square Linear RegressionR-LR: Robust Linear RegressionSTL: Seasonal Time Series Decomposition using LoessC S D : C l a s s i c a l S e a s o n a l Decomposition SCSF: Statistical clear sky fittingYbY: Year-by-Year comparison Power NOCT S. Lindig, PVPMC China 2 0 1 9 2 91 0 /1 2 /1 9 Comparison of calculation methodologiesDigital power plant – NO degradation – different yearsVAR: yearly variation of powerYoY: Year on Year – Comparison of daily power valuesLS-LR: Least Square Linear RegressionR-LR: Robust Linear RegressionSTL: Seasonal Time Series Decomposition using LoessC S D : C l a s s i c a l S e a s o n a l Decomposition SCSF: Statistical clear sky fittingYbY: Year-by-Year comparison Power NOCT S. Lindig, PVPMC China 2 0 1 9 3 01 0 /1 2 /1 9 Comparison of calculation methodologiesDigital power plant – unknown degradation – different yearsVAR: yearly variation of powerYoY: Year on Year – Comparison of daily power valuesLS-LR: Least Square Linear RegressionR-LR: Robust Linear RegressionSTL: Seasonal Time Series Decomposition using LoessC S D : C l a s s i c a l S e a s o n a l Decomposition SCSF: Statistical clear sky fittingYbY: Year-by-Year comparison Power NOCT