Field Testing Meets Modeling: Validated Data on Bifacial Solar Performance
Though bifacial solar cell technology is not new to PV, recent advances have made the technology more commercially viable than ever before for PV power plants.
Along with that viability, however, come questions about the potential energy yield and best practices when choosing bifacial modules over their monofacial counterparts.
In particular, university research and other preliminary studies of the increase in gain from a switch to bifacial modules suggested a bump of 20-30% was possible – but is that really the case?
In short, no.
Array Technologies partnered with PV Lighthouse and CFV Solar Test Laboratory to conduct an independent study assessing the real-world impact of bifacial solar cells.
The study’s goal was to determine best practices for modeling the technology in order to get an accurate view of a utility scale PV plant’s power production. By using field data from CFV correlated by three-dimensional ray tracing modeling from PV Lighthouse, the study determined realistic modeling inputs. The results present bankable, real-world modeling best practices for bifacial power plants.
The recent Greentech Media webinar based on this research gave Array, PV Lighthouse and CFV the chance to answer common questions about the study and its findings.
Webinar Q&A Highlights
If you missed the webinar, below is a look at some of the highlights as well as answers to questions which were submitted by attendees.
Answers below were provided by Array’s Kyumin Lee (KL), PV Lighthouse’s Keith McIntosh (KM) and CFV’s Jim Crimmins (JC).
Q: Importing grey and white gravel for ground cover to boost albedo gain is economically unrealistic. Were multiple ground cover types considered during the simulation for variable albedo values?
JC: For a lab environment, the most important thing is to have a stable albedo. Gravel is a very stable surface and easy to control for vegetation. We have simulated higher albedos with various synthetic surfaces as well. A stable albedo allows us to fit models which can then be used to predict yields with varying albedos. It is best to think of albedo in the real world as a site-dependent time series, changing seasonally and also with precipitation and snow.
Q: Did SunSolve consider 2MIP with a torque tube that avoids the backside of the module or obstructs the backside of the module?
KM: Specifically we examined the case when the east-west module separation was 16 cm (the same width as the torque tube). As you’d expect, this led to less shading from the torque tube, but the resulting increase in yield was counteracted by the need for greater backtracking to avoid row-to-row shading. We found that including an east-west separation between modules led to a slightly lower yield.
Q: Is there a current partnership between SunSolve and Array?
KM: PV Lighthouse is an independent scientific consultancy and software provider that was engaged by Array Technologies to conduct simulations for this study. While we welcome the opportunity to work with Array Technologies in the future, we don’t have a formal partnership.
Q: Does SunSolve offer any guarantees regarding their modeling?
KM: Confidence in SunSolve’s accuracy is gained from validation studies, such as that presented in the webinar and others described in published papers. However, like other system-simulation software, SunSolve is not guaranteed to predict yield to within some level of accuracy, because that accuracy depends on the user’s experimental measurements and selection of SunSolve inputs. PV Lighthouse enters consulting agreements in which its SunSolve simulations and analysis are applied to a guaranteed professional scientific standard.
Q: Were the results consistent at low albedo values?
KM: As the albedo decreases, the relative advantage of 1MIP over 2MIP also decreases. The reason for this trend is that the optical advantage of the 1MIP to 2MIP pertains to the rear-side optics (for the tested configurations).
Q: Does gain vary as a function of array spacing?
KL: Yes, when the system is not backtracking. But the view factor does not increase proportionally with the ground cover ratio (GCR), and so there is a crossover point in terms of economics.
Q: Did you have dummy modules at the edges? In that case, wouldn’t you have reduced the likelihood of hotspots and bypass diode activation?
KL: There were dummy modules at the north end. The cell mismatch is greatest at the south end when the ground irradiance comes at normal angles.
Q: What are the efficiencies of each side, and the total, integrated efficiency?
KL: The bifaciality of a PV module is measured per IEC TS 60904-1-2. Basically, Standard Test Conditions peak power on backside / Standard Test Conditions peak power on frontside. In modeling, you do not use different efficiencies for the front, back and integrated. The module efficiency stays constant, and it is the total effective irradiance that gets calculated. Total effective irradiance = Frontside Plane of Array irradiance + Bifaciality * Backside Plane of Array irradiance.
Q: How much diffuse irradiance is actually there to draw from?
KL: That is totally climate-dependent. Across the U.S., the annual diffuse fraction (diffuse horizontal irradiation / global horizontal irradiation) can vary from 0.2 to 0.5.
Q: What’s the best way to figure out the albedo values for a bifacial system on a ground with seasonality?
KL: The best practice would be to make an actual albedo measurement at the site over a year. A fallback solution would be to use albedo values as estimated by various satellite-based services, such as PVGIS.
Q: For module for Array2, what tech is applied? N-type Topcon, PERT or HJT?
Q: What’s the difference between albedo and diffuse fraction?
KL: The “albedo” is the reflectance of the ground. The “diffuse fraction” is the fraction of sunlight that does not come directly from the sun. For example, some sunlight reflects from clouds before it reaches the panels; that sunlight contributes to the diffuse fraction. On a cloudless day, the diffuse fraction is about 10%, and on a fully overcast day, the diffuse fraction is 100%.
Q: I can see that shading loss between 1MIP and 2MIP is similar. But there is assumption that there is no spacing between top and bottom modules in 2MIP. But, in reality, there is gap between them in 2MIP.
KL: For a 2MIP setup with fixed array row spacing, having zero gap between the modules will give you more energy. A 2MIP configuration with a gap between the modules is just not an economically viable setup. You lose more energy from prolonged backtracking than any energy gain from reduced shading loss.
Q: What is the margin of error for the bifacial measured versus PVsyst?
KL: It really depends on each case. Currently, the biggest sources of uncertainty are (1) module characteristics and (2) albedo. If these are known, the accuracy of the energy calculated by PVsyst for a given weather data can be well below 0.5%.
Q: I installed my first solar system at the South Pole in 2008, and I will likely go back sometime within the next five years. If I were to mount bifacial modules perpendicular to the ground, would I get the same output from the module 12 hours later when the sun is at the same angle and intensity? Basically, what would the difference in output be if I mounted the module upside down?
KL: It depends on the module bifaciality. Widely available bifacial modules are 65% bifacial, and so the back side will produce about 35% less for the same irradiance. There are modules that are more than 70% bifacial (LG n-type, for example).
Q: Can you share the modeling assumptions used in the PVsyst model of the setup at CFV? (shade Loss, transparency, mismatch, etc.)
KL: The transparency was set to 2.5% for the Type 2 (clear backsheet n-PERT), and 0 for the Type 1 (white-backsheet polyPERC).
Q: Is CFV going to build a 2MIP test setup?
JC: No plans yet, but we are open to it.
Q: Is shadow intensity taken into account in the models? How much impact does it have on the results?
KC: Yes, it is taken into account. We didn’t examine its contribution specifically in the project.
Q: The photo of the test yard did not include any 2MIP rows. Did you deploy/measure 2MIP RPOA and Bifacial Gain at the CFV test site?
KL: The testing did not include 2MIP trackers. Our approach was to gain confidence in 3D ray tracing and 2D view factor modeling (PVsyst) via testing on our 1MIP tracker, and then extend the ray tracing and modeling to 2MIP cases.
Q: Do you intend to validate the 2MIP model empirically at a later date?
KL: We don’t have anything planned at this moment. We’ll leave it up to our 2MIP competitors to provide field testing results of similar quality!
Q: Can the described ray trace software handle complex terrain applications, such as undulating terrain in Northeast?
KM: SunSolve currently assumes flat terrain.
Q: We already use GCR in characterizing POA boost. Introducing row height is an orthogonal metric that can tune for bifacial effects, but the ratio of width to height is redundant if GCR is already used.
KL: I completely agree. We just wanted to debunk the myth that higher rows are better, no matter what.
Q: The measured albedo was 30%, the ideal value is 24%. Can we trust measurement to lead to good simulations?
JC: Models predict that bifacial gain results are very linear with albedo within that type of range. So it is easy to scale them.
Q: What is the optimal platform composition for building bifacial SAT’s over in order to increase albedo & diffuse fraction?
KL: The bifacial gain will come in the range of 5-10% for albedo of 0.3. If a monofacial system’s production is 100, the bifacial system will produce 105 to 110. You can get it to ~115 with more reflective ground cover, but I’m not sure if it’s really going to be worth it.
Q: Did the 2MIP system you considered take into account a racking system with the torque tube between modules so (there would be) no direct rear shading on the back of the module?
KL: East-west gap over the torque tube on 2MIP tracker actually reduces the energy production, because you backtrack more and longer. The best strategy for 2MIP tracker is to not have any gap.
Q: Any testing done in the snow?
KL: CFV simulated snow conditions with a white reflective tarp.
Q: How much impact does albedo have on the performance of the bifacial modules?
KL: The albedo has the biggest effect on the bifacial gain. A system that generates ~5% bifacial gain on ground cover of albedo ~0.3 will generate ~10% bifacial gain on ground cover of albedo ~0.6.
Q: How much does shading affect 1MIP vs 2MIP, and does that negatively affect the capture ratio advantage 1MIP has over 2MIP?
KL: 2MIP trackers can achieve lower shading loss factor by having east-west module gap, but this will actually reduce annual production, because the gap increases the GCR, resulting in longer backtracking. The energy loss from increased backtracking is more than any gain from the lower shading factor. So the best strategy for 2MIP is to not have any gap between the modules. When there isn’t any module EW gap, the shading loss of a 2MIP tracker is not too different from that of a 1MIP tracker.
Q: What does MIP tracker mean?
KL: 1MIP = One module in portrait. 2MIP = Two modules in portrait.
Q: Would you say that bifacial modules are best suited to specific geographic locations in order to get the most energy gain?
KL: No, I think bifacial systems generally make sense in most regions.
Q: Would bifacial modules do well on dual axis trackers to pick up more albedo from “edge effects?”
KL : Yes, they would in a sense. But bifacial alone is not enough to justify two-axis trackers.
Q: For PR testing, how frequently would you place albedometers or pyranometers measuring backside irradiance? How much uncertainty do you think would be on backside irradiance measurements during PR resting compared to frontside irradiance, considering less homogeneous nature of the backside irradiance?
JC: This is a complicated question. We did not measure small area backside irradiance in this project. We used pyranometers. Small area backside irradiance is probably best handled through modeling.
Q: Any thoughts on ‘mismatch loss’ to be applied in PVSyst?
KL: The mismatch loss factors that we determined from the ray tracing and SPICE model are smaller than what has typically been suggested for PVSyst, partly because the ground-reflected rays come in at high angles, creating soft shadows.
Q: The 1MP and 2MP comparison – will this come out naturally in PVSyst, or do some losses need to be changed in the settings?
KL: This will come naturally if you use the nominal configurations for 1MIP and 2MIP trackers.
Q: How does wire management impact bifacial gain?
KL: Wiring management is important, and in the CFV testing, a lot of care was taken to make sure that the wiring shading was not a significant factor. Offering a utility-scale solution is something we plan to do in near future.
Q: Did PV Lighthouse account for the spectral response of the different PV modules tested at the CFV site?
KM: Yes. The incident spectrum from the sun was estimated using published spectral models, the spectral albedo of the ground (and torque-tube) was taken from a NASA database, and the spectral response of the modules was determined from ray tracing and device simulation.
Q: How would the backtracking algorithm change, if any, due to gains in aspect ratios?
KM: The backtracking algorithm would not change. You still want to capture the most light on the frontside directly without suffering row-to-row shading.
Q: What is the difference you experience in performance of N-PERT & P-PERC bifacial?
KL: The bifaciality of an n-PERT module is in the 70-75% range. The bifaciality of a p-PERC module is in the 65-70% range. The n-PERT module we tested had higher series resistance and lower temperature coefficient too, which also helped the energy yield.
Q: What is the impact of bifacial module self-shading on performance/product health?
KL: I’m not aware of any self-shading of bifacial modules.
Q: Did you use the same GCR for the 1P and 2P simulation, and if so, would the aspect ratio and light lost to the sky be the same?
KL: Yes, we used a GCR of 35.1% for both 1P and 2P simulations. The height of the torque tubes was set to 1.6 m for 1MIP and 2.4 m for 2MIP (which are typical of commercial systems). Since the 2MIP configuration has two times the module length but not two times the height, it has a lower aspect ratio, and hence more light is lost to the sky.
Q: How is performance ratio found with bifacial modules?
KL: We did not use performance ratio. We calculated daily specific yields of the bifacial and monofacial arrays and compared these values to calculate the bifacial gain.
Q: What programming language is SunSolve written in?
Q: Can you cover the calculation of maximum current over the luminescent solar concentrator used in monofacial modules.
KL: I’d recommend looking at hourly sums of GlobInc and GlobBak values in PVsyst to estimate the maximum effective irradiance into the module, and scale the frontside luminescent solar concentrator accordingly, with some consideration for the temperature coefficient.
Q: Are the string currents of a bifacial string in parallel to inverter block limited by current mismatch to the lowest module current on each string? So the edge effect gain on the ends a short tracker are not realized?
JC: The edge effects will add some incremental yield, but will be reduced by mismatch in a long string. In a very short string or measured individually, they are more important.
Q: What about clipping? What’s the preferred AC/DC ratio?
KL: I recommend carrying out PVsyst + economics studies to determine the optimum DC:AC ratio. My first guess would be nominal monofacial DC:AC ratio minus the expected (unclipped) bifacial gain.
Q: Were any 2MIP arrangements physically tested, or was that solely modeled?
KM: In this study, the 2MIP arrangements were solely modeled. PV Lighthouse will be presenting a paper on measurements of 2MIP configurations at the IEEE PVSC in July 2020.
Q: Why was the albedo parameter different for 1 MIP compared to 2 MIP in Keith’s study? I believe it was something like 0.240 vs 0.236.
KM: The same albedo was used in the ray tracing of 1MIP and 2MIP configurations. It was a function of wavelength that represents brown, loamy soil. The results of the ray tracing were then used to determine the best value to use as the albedo factor in PVSyst to represent the same soil, and you’re right, it was indeed slightly lower for 2MIP (0.236) than for 1MIP (0.240). This indicates that the 2MIP configuration made slightly less use of the ground-reflected light than the 1MIP configuration over the course of the year (accounting for wavelength, diffuse/direct ratio, and the assumed weather conditions). We expect those values to change for different albedos, module types and conditions.
Q: Is SunSolve already available to simulate bifacial systems? What are the exact outputs of the simulation?
KM: Yes. There are two ways to simulate bifacial systems with SunSolve.
(1) On a contractual basis, where PV Lighthouse performs the simulations (as conducted in this study).
(2) On a subscription basis, where you perform the simulations to determine (i) output power and detailed losses at a single point in time, and (ii) yield, losses and PVSyst inputs over any period of time (available July 2020).
Email email@example.com for more detail.
Q: Did PV Lighthouse simulate a 2 MIP configuration to generate similar PVsyst inputs as 1 MIP?
KM: Yes. For the results of the 2MIP configuration, please refer to slide 26 of the webinar, titled “SunSolve – Determine PVSyst inputs,” and the white paper associated with the webinar.
Q: Can you quickly explain again if torque tube shading is significant on a 1 MIP configuration vs a 2 MIP?
KL: PV Lighthouse found that the structural shading loss is not so different for 2MIP and 1MIP. We studied the no-gap case, because we found that a 2MIP configuration with a module gap suffered more losses from the prolonged backtracking than any energy gain from the reduced structural shading.
Q: Is there any method or published article to measure bifacial gain at site?
KL: There is no standardized method. We calculated the bifacial gain by first calculating the daily specific yields of bifacial arrays and a monofacial array and by looking at their ratios.
Q: Is there a need to include back side irradiance in Performance Ratio calculation for the plant?
KL: Cypress Creek, along with NREL, suggested relevant modifications for the capacity testing of bifacial systems in 2019. The publication can be found here.
Q: What is a good resource for determining monthly albedo values?
KL: The best practice would be to make an actual albedo measurement at the site over a year. A fallback solution would be to use albedo values as estimated by various satellite-based services such as PVGIS. NSRDB also offers albedo values.
Q: Is ground-reflected light assumed to have a Lambertian distribution?
KM: Yes, in this study the ground-reflected light was assumed Lambertian. (SunSolve can simulate alternative scattering functions if that is of interest).
Q: Where can I get these ray-tracing derived inputs for running PVsyst simulations?
KM: The results for the PVSyst inputs are on Slide 26 of the webinar and they are also given in the white paper. These results are specific to the particular site (in West Texas) and system configuration. PVSyst inputs for other sites and configurations can be determined with SunSolve.
Q: Where do you all recommend locating an albedometer and backside POA sensor to most accurately represent an entire array?
JC: This depends on exact deployment scenario and how variable the albedo is on the site.
Q: Many bifacial module manufacturers specs or datasheets include a backside gain of up to 25 or 30%. It sounds like realistically the actual “average” or real-world backside gain will be closer to 10%?
KL: It will be closer to 5%.
Q: What is the average production gain over mono-facial modules?
KL: 5-10% would be a good estimate.
Q: Why does an event of high diffuse fraction not translate to increased power production?
KM: An increase in the diffuse fraction leads to an increase in irradiance incident to the rear side of the modules but a decrease in irradiance incident to the front side. That’s why an increase in diffuse fraction leads to an increase in bifacial gain. However, the advantage from the rear side is outweighed by the disadvantage from the front side, except for some rare exceptions like during backtracking very late in the day.
Q: Wouldn’t the gains of 1MIP relative to 2MIP be a wash or negative case considering the additional racking and construction costs associated with 1MIP?
KL: We think 2MIP trackers are still too new to understand the associated costs accurately.
Q: What is the economy of MIP1 v MIP2 in the context of bifacial increase in production using MIP1 v. overall tracker costs (which is presumed less when using MIP2 designs)?
KL: We think 2MIP trackers are still too new to understand the associated costs accurately.
Q: Any work on commercial low slope roof or residential metal roof rooftop applications?
KL: No, this is not an application our product serves.
Q: Would CFV labs be able to release the PVSyst models used for their analysis presented on slide 15?
Q: How dependent are the bifacial gain drivers on the site location? Do these relationships hold equally true in the Northeast and the Southwest U.S.?
KL: It’s complicated. Yes, you’ll see higher bifacial gains in the Northeast compared to the Southwest, but the actual additional energy production in kWh due to bifacial will still be higher in the Southwest.
Q: Is the recommended PVsyst ‘height above ground’ a fit based on the aspect ratio and measured bifacial gain? Would that hold true for all locations?
KL: The study was carried out for a West Texas location. I think you’ll find similar results on other locations, too, but the sensitivity might be different.
Q: Does your testing take into account cell temperature?
JC: We assumed modules were at the same temperature.
Q: Does PV Lighthouse only simulate silicon cells, or can it be used for thin film and/or multijunction cells?
KM: Theoretically, SunSolve can be used to simulate thin-film and multijunction cells, but it is currently oriented toward silicon cells.
Q: Did tested modules have clear gap in the middle of the module so light hits the torque tube and bounces off? Or glazing between cells?
KL: The n-type modules had clear gaps between the cells, but these were full-cell modules. The p-PERC half-cell modules did not have clear gaps.
For further information:
Array has also published a PVsyst input data sheet, which you can request here.
For more information about solar trackers from Array Technologies, contact us.