Geographic smoothing of solar PV: Results from Gujarat AMS 2016 Kelly Klima, Jay Apt 1
Many forms of renewable energy exist. Some are variable, requiring smoothing. Wind Solar Biomass Geothermal Wave Hydropower 2
Can solar PV be smoothed? Examining Irradiance Data Correlation measured at two locations decreases as the distance between the sites increases. Long & Ackerman, 1995; Barnett et al., 1998; Lave & Kleissl, 2010, Hinkleman 2013.; Argonne National Labs 2013 etc. Changes in clear sky index for 23 locations show smoothing is likely for as few as 5 plants. LBNL 2010. Examining Generation Data In Germany, 5-min ramps in normalized PV power may exceed +/- 50% for 1 plant but never exceed +/- 5% for 100 PV sites. Wiemken et al., 2001. Correlation of real power output for three sites in Arizona is high, suggesting smoothing might not work here. Curtright & Apt, 2008. Sites at hourly resolution show smoothing. Rowlands 2014. 3
We can measure solar data at several points between the sun and the power plant output. Satellite Actual Generation Pyrheliometer 4
In this study we examined actual generation data. Satellite Actual Generation Pyrheliometer 5
Latitude We examined generation data from 50 power plants in Gujarat (eastern India). 26 o N 25 o N 24 o N 23 o N 22 o N 21 o N 20 o N 68 o E 69 o E 70 o E 71 o E 72 o E 73 o E 74 o E Longitude 6 Green = Plants used in rest of presentation 5-221 MW of installed capacity; changed over time. Source: Gujarat State Load Dispatch Center, downloaded every minute.
Generation (MW) Generation (MW) Generation (MW) Generation (MW) Time series show sunny & cloudy days. 15 15 10 10 5 5 0 0 100 200 300 Day after 2/17/2014 15 0 2 4 6 8 Day after 2/17/2014 15 10 10 5 5 0 2 2.5 3 Day after 2/17/2014 7 0 4 4.5 5 Day after 2/17/2014
Number of occurences Time stamp intervals were uneven, and generally within 1 to 2 minutes. 2.5 x 105 2 1.5 1 0.5 0 0 5 10 15 20 Length of timestamp in minutes after an index 8
To understand the variability, we looked in the frequency domain. Curtright & Apt 2007: Sample power spectral density (PSD) of a Tucson Electric Power Array; red line is f^-1.3 9
Magnitude of generation variability kw/sqrt(hz) The PSD for one Gujarat plant has a f^1.3 spectra. 10 2 10 0 Black = 1 plant 10-2 10-6 10-4 10-2 Frequency (Hz) 10
Magnitude of generation variability kw/sqrt(hz) Summing the generation of the 5 closest plants smoothens higher frequencies. 10 2 10 0 Black = 1 plant Red = 5 plants 10-2 10-6 10-4 10-2 Frequency (Hz) 11
Magnitude of generation variability kw/sqrt(hz) There appear to be diminishing returns with adding plants together. 10 2 10 0 10-2 Black = 1 plant Red = 5 plants Green = 10 plants 10-6 10-4 10-2 Frequency (Hz) 12
Magnitude of generation variability kw/sqrt(hz) The amount of smoothing achieved with 20 plants is almost the same as with 10 plants. 10 2 10 0 10-2 Black = 1 plant Red = 5 plants Green = 10 plants Magenta = 20 plants 10-6 10-4 10-2 Frequency (Hz) 13
Fraction of PSD of 1 plant Interconnecting a few solar plants achieves the majority of smoothing. 1 0.8 0.6 0.4 0.2 0 5 10 15 20 # of PV Sites 6 hours 4 hours 3 hours 1 hour 10 minutes sqrt(n) 14
MUCH less smoothing than for wind Wind PV 6 hours 1 hour That is probably because PV s deep power fluctuations lead to variability at many frequencies. 15
Conclusions Interconnecting 20 Gujarat plants yields a f^-1.66 spectrum and reduces fluctuations at frequencies corresponding to 6 hours and 1 hour by 23% and 45%, respectively. Half of this smoothing can be obtained through connecting 4-5 plants. The largest plant (322MW) showed an f^-1.76 spectrum. This suggests that in Gujarat the potential for smoothing is limited to that obtained by one large plant. 16
Contact Information Kelly Klima kklima@andrew.cmu.edu Jay Apt apt@cmu.edu This research is based upon work supported by the Solar Energy Research Institute for India and the U.S. (SERIIUS) funded jointly by the U.S. Department of Energy subcontract DE AC36-08G028308 (Office of Science, Office of Basic Energy Sciences, and Energy Efficiency and Renewable Energy, Solar Energy Technology Program, with support from the Office of International Affairs) and the Government of India subcontract IUSSTF/JCERDC-SERIIUS/2012 dated 22nd Nov. 2012. 17
References Melillo JM, Richmond TC, Yohe GW. 2014. Climate Change Impacts in the United States: The Third National Climate Assessment. U.S. Global Change Research Program. doi:10.7930/j0z31wj2 Curtright, A.E.; Apt, J. (2008). The character of power output from utility-scale photovoltaic systems. Progress in Photovoltaics: Research and Applications. 16(3), 241 247. Mills, A.D.; Wiser, R.H. (2010). Implications of Wide-Area Geographic Diversity for Short- Term Variability of Solar Power. Technical Report LBNL-3884E. Long, C.N.; Ackerman, T.P. (1995). Surface Measurements of Solar Irradiance: A Study of the Spatial Correlation between Simultaneous Measurements at Separated Sites. Journal of Applied Meteorology. 34(1), 1039 1046. Barnett, T.P.; Ritchie, J.; Foat, J.; Stokes, G. (1998). On the Space Time Scales of the Surface Solar Radiation Field. Journal of Climate. 11(1), 88 96. Lave, M.; Kleissl, J. (2010). Solar variability of four sites across the state of Colorado. Renewable Energy. 35(12), 2867 2873. Hinkleman, L. (2013). Differences between along-wind and cross-wind solar irradiance variability on small spatial scales. Solar Energy. 88(1), 192 203. Argonne National Labs 2013 Lave, M.; Kleissl, J.; Arias-Castro, E. (2012). High-frequency irradiance fluctuations and geographic smoothing. Solar Energy. 86(8), 2190 2199. Rowlands, I.H.; Kemery, B.P.; Beausoleil-Morrison, I. (2014). Managing solar-pv variability with geographical dispersion: An Ontario (Canada) case-study. Renewable Energy. 68(1), Pages 171 180. 18
Extra slides 19 19
NREL calculated solar data from 2002-2011. SUNY model: created for U.S. observations 1 x 1 Calculated direct normal irradiance, direct horizontal irradiance, and global horizontal irradiance Values almost as high as Arizona. 20
We compared NREL calculated irradiance with generation data in Arizona. Pairwise correlation, Actual vs NREL. Values are given as: GHI correlation, (DNI correlation) Springerville (1/1/2004-12/31/2005) Cimarron (10/12/2010-12/31/2005) NREL SUNY NREL Metstat 8am-8pm LT 10am-6pm LT Noon-4pm LT 0.79, (0.60) 0.79, (0.61) 0.79, (0.61) 0.77, (0.60) 0.76, (0.6) 0.76, (0.6) 0.75, (0.69) 0.44, (0.41) 0.19, (0.35) This suggests that we might be able to use the NREL insolation data to predict generation data in India. 21
Since correlations were not 1-1, we decided to use actual generation data. Data Locators: IITB s Rangan Banjeree & Rhythm Singh Data Source: State Load Dispatch Center for Gujarat, India Download and Conversion Advisors: Eric Brundick, Terrence Wong, Joe Jewell, Jeremy Boulton, Eric Morganson. 22