Innovative floating bifacial photovoltaic solutions for inland water areas

Photovoltaic (PV) technology has the potential to be integrated on many surfaces in various environments, even on water. Modeling, design, and realization of a floating PV system have more challenges than conventional rooftop or freestanding PV system. In this work, we introduce two innovative concepts for floating bifacial PV systems, describing their modeling, design, and performance monitoring. The developed concepts are retractable and enable maximum energy production through tracking the Sun. Various floating PV systems (monofacial, bifacial with and without reflectors) with different tilts and tracking capabilities are installed on a Dutch pond and are being monitored. Results of the thermal study showed that partially soaking the frame of PV modules into water does not bring a considerable additional yield (+0.17%) and revealed that floating PV modules experience higher temperature special variance compared with land‐based systems. Observations showed that the birds' presence has a severe effect on floating PV performance in the short term. Electrical yield investigation concluded that due to low albedo of inland water areas (~6.5%), bifacial PV systems must have reflectors. One‐year monitoring showed that a bifacial PV system with reflector and horizontal tracking delivers ~17.3% more specific yield (up to 29% in a clear‐sky month) compared with a monofacial PV system installed on land. Ecological monitoring showed no discernable impacts on the water quality in weekly samplings but did show significant impacts on the aquatic plant biomass and periods of low oxygen concentrations.

cost have increased the bifacial PV installation rapidly from 97 MWp in 2016 to over 5 GWp in 2019. 2 The market share of bifacial technology is expected to grow from 15% in 2019 to 70% by 2030. 3 On the other hand, the expected increasing trend for the world population, 4 and subsequently the need for food and energy, signifies the importance of the land. In such a situation, floating PV (FPV) systems either on off-shore or on-shore water areas seems like an interesting option to reduce the LCOE and the conflict with other landuser sectors such as agriculture and housing. Installation of PV systems on ocean water brings several challenges such as harsh dynamics, storms, and difficult accessibility. Inland water areas could be instead exploited for clean solar energy production in several countries, such as Germany, Finland, and the Netherlands. For the case of the Netherlands, 17% of the land is covered by waterways, lakes, and ponds. 5 Moreover, in countries with high insolation rate and drinkable water scarcity, FPV can produce electricity and prevent water evaporation as well. 6 For example, in Morocco, 3 m 3 of water is evaporated yearly per each square meter of water surface behind the dams. 7 The market share of the FPV system is expected to be >10% by 2030. 3 Combining these two trends in the PV industry, floating and bifacial, could be a promising way forward. However, floating bifacial PV has rather unique requirements, challenges, and opportunities. There is general understanding in the literature that FPV systems face wave and wind forces with various frequency and domains, 8 experience stronger aging mechanisms caused by moisture and harsh environment 9,10 while they might benefit from the cooler working environment and sunlight reflection from the water. 11 They should also meet the requirement for water ecology, release minimum or zero toxic material to the water, and enable mowing activities. Moreover, as with any novel technology, bifacial and FPV is still held back by a lack or inaccessibility of long-term field data to demonstrate its real-world performance under various conditions. 12  (1) location survey and accurate PV yield modeling, (2) innovative, applicable, and modular design and realization of the floating construction and PV systems, and (3) monitoring for the systems. These three main steps are discussed in detail in Sections 2 to 4 of this paper. In chapter 5, key messages of this research are highlighted.
Throughout the study, several interesting facts about floating bifacial PV were proven and/or observed that some are in contrast with initial expectation.

| Location
A storm water pond, located in Weurt, Eastern Netherlands, was selected for this research (see Figure 1). Based on the measurement, the basin has a minimum depth of 0.9 m and up to 1.9 m in the deeper parts. Depending on the season and water retention plans, the basin can occupy an area of 18 524 to 22 639 m. 2

| Horizon and sky view factor
Skyline-related information is an important input for PV yield modeling that influences the direct and diffuse components of the sunlight. 16 One key indicator is sky view factor (SVF), which is a unit-less quantity that represents the ratio between the visible sky and a hemisphere centered over the studied location. 17,18 Skyline information is extracted using a hori-catcher (see Figure 2A) 19 in and around the pond at several spots with different heights. Then, the captured horizon was processed using Meteonorm software package 20 using the approach described in Stein 21 and Keijzer. 22 Further, by applying linear interpolation, a map of SVF was obtained, as can be seen in

| Albedo
Albedo is measured by dividing the incoming global radiant fluxes reaching on the down-facing and up-facing parts of a surface. Location, time, geometry, and weather conditions influence the value of albedo. 23 Albedo is spectrally dependent, and for PV applications, the F I G U R E 1 Panoramic view of the artificial pond located in Weurt, the Netherlands (51.8514 N, 5.7950 E). The pond is developed as a water retention basin. The pond has a rectangular shape, not perfectly aligned with the south. The longer side of the pond is 42 deviated toward the west (222 ) [Colour figure can be viewed at wileyonlinelibrary.com] effective albedo, which considers the spectral response of the PV cell, should be taken into account. 24 Accurate assessment of albedo becomes more important for bifacial PV installations. Broadband and spectrally resolved albedos were measured both for the water in and the soil around the pond. Figure 3 shows the Kipp&Zonen albedometer (sensitive from 285 to 2800 nm, measures in W/m 2 ) and Avantes spectrometer (sensitive from 278 to 1098 nm, measures in μW/cm 2 /nm), which were used, respectively, to measure the broadband and spectrally resolved albedos at various spots at the pond. Then, by applying the model described in Ziar et al., 23 which bounds the geometrical and spectral features of albedo, maximum value of albedo was calculated for various levels above the water surface at the pond.
The mapped albedos of the pond at 1 and 2 m above water level are shown in Figure 4. As it can be seen, the closer to the shore and the higher from the water surface, the higher the albedo. Figure 5A shows the spectrally resolved reflected irradiance of the soil and the water at the pond, whereas Figure 5B shows two sky spectra: measured at the site and standard ASTMG173. Figure 5C shows the typical response of a monocrystalline silicon solar cell. 25 Using the information given in Figure 5, the effective albedo of the soil and water are calculated as follows: 15.64% (soil), 7.71% (water 0.5 m depth), and 8.11% (water 1 m depth). Very low albedo of water predicts a low contribution of the reflected light energy for bifacial PV installation and, therefore, suggests the necessity for using reflectors.

| Yearly irradiation modeling
Monofacial PV brings the advantage of simplicity and lower costs, whereas bifacial FPV are expected to yield more. 26,27 Therefore, in our preinstallation study, we considered both monofacial and bifacial technologies. Monofacial PV modules can be either placed aligned with the pond orientation for better area usage (higher kWh/m 2 ) or toward the south for better performance (higher kWh/kWp). For bifacial PV, reflector type and orientation and its distance from the PV module are also becoming important.
During the preinstallation study, several design parameters (type and technology of the PV modules, BoS components, etc.) had not been fixed; therefore, the focus was put on the irradiation modeling.
Since irradiation is the key component in PV yield analysis 28 the design that receives the highest yearly irradiation will yield higher electrical output.
F I G U R E 2 (A) A hori-catcher image at one spot in the pond taken using an upside-down camera and a mirror-like spherical cap. (B) Interpolated map of the sky view factor (%) for the pond area. The sky view factor (SVF) changes between 97.9% and 99.3%, showing an almost free horizon suitable for PV installation. One meter above the water surface was considered for SVF calculation as SVF does not significantly depend on few meters of change in the altitude [Colour figure can be viewed at wileyonlinelibrary. com] The pond is located between five meteorological stations of the KNMI (Koninklijk Nederlands Meteorologisch Instituut) network 31 (see Figure 6). By retrieving the 1-h resolution global horizontal irradiance (GHI) data from the stations and applying inverse distance weighted (IDW) method, 32 the GHI data at Weurt was interpolated and then processed in Meteonorm (horizon was applied) and further broken into direct normal and diffuse horizontal irradiances (DNI and DHI) through BRL decomposition model, 33 which was shown to be the most accurate irradiation decomposition model for the morphology of the Netherlands. 34 The location receives the irradiation of 989 kWh/m 2 /year. Table 1 shows the calculated yearly irradiation for various fixed and tracking monofacial PV modules. The simulated tracking system was astronomically tracking the Sun position.
Further, to assess the total yearly irradiation on bifacial PV modules, the PVMD toolbox was used. 36 Several cases were studied with various tilt and orientation for the PV module and the reflector, from which a few examples are shown in Figure 7. Simulations showed that putting a reflector very close to the bifacial PV module and relying on the light passing through the cells will not significantly contribute to the total irradiation and should be placed with a distance underneath the bifacial PV module.
The only monofacial option (see Table 1) that can outperform the tilted bifacial with a horizontal reflector ( Figure 7F) is when a freeangle dual-axis tracking is done. However, this may not be mechanically feasible that tumbling a floating system might escalate the wind force. Therefore, tracking should be done within a limited range of safe angles. This suggests that maybe a tracking bifacial PV system with limited angles can bring both safe operation, long life-time, high yield, and consequently low LCOE. Therefore, in the design phase, we F I G U R E 4 Mapped broadband albedo (%) for the pond area at (A) 1 m above water level and (B) 2 m above water level [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 5 Spectrally dependent data for effective albedo calculations. (A) Reflected spectrum measured at the pond for the soil and two different depth of water. (B) Measured sky spectral irradiance and the ASTM G173 standard global irradiance. 29 (C) Spectral response of HOQ mono-cSi reference cell reported by Fraunhofer ISE. 25 Measurements were done on November 29, 2017, under overcast sky condition, no rain, ambient temperature 2 C, and the wind speed of 0.5 to 1 m/s. Effective albedo values for the soil, 5 m, and 8 m in the pond are, respectively, calculated as 15.64%, 7.71%, and 8.11%. Considering ASTMG173 sky spectrum will change the values slightly (15.34%, 7.91%, and 8.32%). It is worth noting that for assessing the effective albedo, when the target PV technology is known, relative response of that technology should be used. Note that the front side and rear side responses of bifacial PV modules are different 30 [Colour figure can be viewed at wileyonlinelibrary.com] aimed for a tracking bifacial PV system with reflector, among several other design options, which will give conclusive results about the best solution for floating bifacial PV.

| Partial water soaking
Soaking a PV module in water changes the received sunlight spectrum on the PV module surface (negative effect) but reduces the working temperature (positive effect). 37 Therefore, for water soaking applications, it is always important to find the sweet spot that keeps the temperature low but does not drastically reduce the impinged irradiance. This is normally dictated by the soaking depth in the water. Therefore, for a FPV, which is in the vicinity of the water, it is worth investigating water cooling as a design option. However, as shown in Section 2.4, using a reflector is essential to make the most out of a bifacial PV installation. This implies that soaking a bifacial PV module fully in water is not a logical approach because it would lead to a very small contribution from the rear side. Hence, to study such F I G U R E 6 Geographical representation of the five meteorological stations around the target area. Inverse distance weighted (IDW) method that is applicable to geographical areas with uniform morphology (such as the Netherlands) was used to interpolate the irradiance data. IDW method assumes that objects that are close to one another are more similar than the ones that are farther apart   COMSOL Multiphysics was used to make a comparison between a bifacial PV module placed above the water and a module with direct water contact of the lower frame (see Figure 8). Water temperature, as an input for the COMSOL model, was calculated using the empirical model suggested in Harvey et al., 39 whereas the rest of the weather inputs were obtained as described in Section 2.4. Material properties of the PV module layers were obtained from the literature 38 and are shown in Table 2.

| Previous FPV concepts
After the first FPV system built in 2007 in Aichi, Japan, several FPV concepts were developed. 41 A FPV system normally consists of floats, mooring, and anchoring system, PV modules, and balance of system (BoS) components. The most common float types are (1) pure floats with high-density polyethylene material and (2) pontoons (or hollow cubes) with metal trusses. [41][42][43] The first type is lightweight and thin with large water-plastic contact area (at least 50% of the plant size), which increases the chance for plastic defoliation.

| Introduced FPV concepts
The available knowledge in the literature, the findings reported in Section 2, mechanical restrictions, and the requirements for water mowing activities served as inputs to converge into two new FPV concepts: (i) retractable system, and (ii) tumbler floating island. In the design concept (i) PV modules are placed between four fixed pillars (as anchoring spots) and can be lined up and spread out (using two winches) when needed. This concept resembles shopping carts where the PV panels can be folded one against the other on a similar way as the shopping carts do in a supermarket. In the design concept (ii), however, the modules are installed on a floating island that is anchored at one spot to the bottom of the lake. The island is occupied with two tanks underneath to track the sun (in the horizontal axis) by pumping water from one tank to another. Both design options enable regular mowing activities and access to the water surface. Both concepts are able to cope with water level variation even in extremely low-water-level seasons. The designed concepts are shown in

| Realization
In summer 2019, realization of the introduced FPV concepts was accomplished. Figure 11A-E shows a few snapshots of the realization procedure. In order to monitor the performance of the FPV concepts and compare it with conventional PV systems, two landbased PV systems were also installed by the pond. In total, nine pilot systems with different features were installed, as can be seen in Figure 12.

| Components and sensors
Installed monofacial and bifacial PV modules were respectively 60-cell LG Neon 2Black (LG330N1K-V5) and 72-cell LG Neon 2Bifacial Due to a lack of immediate supply for the desired white color reflector, orange color reflector with aluminum reflecting coating was utilized. Further, a series of indoor reflectance tests were done on the reflector sample using LAMBDA 950 spectrophotometer. The effective albedo for the reflector was calculated as 68.5% (see Figure 13), which shows that the orange reflector has a comparable albedo with respect to a weathered white reflector.
The winch system is powered up by two PV + battery units located at the two ends of the retractable system (see Figure 14). In this way, the power from the pilot system is not being used for retracting maneuver. However, the power needed for tumbling the floating island is driven from the grid, and to account for this, the pump energy consumption was calculated. The tumbling system uses four pumps to distribute water between two tanks on the two sides of the floater (each tank has two compartments). Each compartment has 1.3 m 3 of space.
The tracking is based on the position of the sun and is done every 15 min with the resolution of 2 using a predefined look-up

| Storm events
The pilot systems experienced two storm events on February

| Visual inspection
In May 2020, after almost 8 months of operation, a series of visual and thermal checks were done on the pilot systems. Visual checks were performed using the NREL guidelines for visual inspection. 63 (1) Slight azimuth misalignment was observed between the landbased pilot systems. It could be because of the storm event and/or inaccurate installation. Also, tilts of the land-based system deviated from the design (7 and 12.7 instead of 15 ). However, the tilt and orientation of the FPV concepts were correct.
(2) Bird droppings were observed at several spots on the FPV modules. The influence of the bird presence was more intense for the surfaces that were closer to the water and less tilted. Remarkably, the reflectors of the retractable system (pilot system no. 5) and the horizontal monofacial PV modules on the tumbler floating island (pilot system no. 6) were heavily covered by bird droppings and even a bird nest (see Figure 16). It is worth noting that the other pilot systems were slightly soiled with no heavy shading.
A few passive actions can be done to reduce the birds' presence effects such as increasing the distance of the reflectors from the water level and placing them slightly tilted in the retractable system and keeping the floating island tumbled during the night time. However, this observation suggests more active bird control techniques (such as laser-based bird control 64 ) for inland FPV concepts.
(3) Vicinity to water: It was observed that when the floating island tumbles with high degrees, the monofacial modules on the (pilot systems nos. 6 and 7) almost touch the water surface. In the long term, this might boost the possibility for PID effect. However, this cannot be confirmed as long as electroluminescence or separate I-V measurements are done on the modules. This is out of the scope of the current study and is planned to be done after the full monitoring period.

| Thermal inspection
The thermal inspection was performed on the nine pilot systems on May 15, 2020, using a Fluke Ti32 thermal imager. Several points were observed that are mentioned here.
The 1-day thermal inspection does not bring conclusive remarks about the temperature of the floating and land-based PV system.
However, it is postulated that there should not be an overall significant temperature difference between the floating and land-based PV systems when they are placed close to each other, as long as they use the same PV technology. The reason is that the main cooling mechanism in PV systems is convective cooling mainly driven by the wind flow, 65 and both systems experience similar wind speeds. Moreover, due to the fact that the FPV modules are more exposed to the bird droppings, which cause shading 66 and hotspot, 67 in general, FPV modules compared with close-located land-based PV systems can even experience higher temperature spatial variance. Figure 17A,B shows the thermal images of the monofacial PV modules on the floating island and the land. As it can be seen, heavily fouled PV modules on the floating island experience higher spatial temperatures variance.
Another interesting aspect is the temperature difference between the front and rear sides of the floating bifacial PV modules. Figure 17C,D shows the thermal images of the front and rear sides of one module on the pilot system no. 9. Although the front side is experiencing a minor hotspot, the rear side has slightly (1.2 C) higher average temperature with more temperature nonuniformity.

| Irradiance monitoring
The irradiance sensors were used to monitor the real-time water albedo and the irradiance reflected on the rear side of the floating bifacial modules (to evaluate the contribution of the reflectors). The sensors installed and started to operate in May 2020. Results of the irradiance monitoring from May 23, 2020, to May 29, 2020, are shown in Figure 18A. For this duration, the daily average water albedo is about 11.6% with rises in the mornings and evenings because of the high angular dependency (low Lambertian behavior) caused by the low altitude of the sun. Also, it is interesting to observe that the water albedo increases during a cloudy day (May 24) with respect to a clear day (May 28) for about 4% absolute.
On daily average, the reflected irradiance on the rear side has a share of 23.4%. This ratio slightly decreases during the cloudy day with respect to a sunny day as a result of two opposite effects: the diffuse light in cloudy days increases the albedo of contributing surfaces (by casting less shadow on the ground 23 ), but on the other hand, the sunlight spectrum is less favorable for the orange-colored reflector (a cloudy day spectrum has more share in the high-frequency region 69 ).
In the mornings and evenings, PV systems receive low amount of sunlight; thus, the contribution of high morning and evening reflections in yield is very low; therefore, irradiance weighted albedo is a better parameter (than the average daily albedo), which was suggested by the literature. 59 Figure 18B shows the measured irradiance weighted albedo of the water and calculated irradiance weighted albedo of the orange reflector over a period of 4 months. Knowing the fact that the sensor at the rear side of bifacial PV panel sees both water and reflector, we used view factors 23,70 (assuming Lambertian behavior) to obtain the irradiance weighted albedo of the reflector: Reflected light ratio on the rear side = α reflector × VF reflector + α water × VF water , where α is the albedo (no unit) and the VF is the view factor (no unit). View factors from the rear side of the bifacial PV panel to the reflector and water were calculated as VF reflector = 0.3313 and VF water = 0.6687, which show that for the retractable systems, the rear side of the bifacial modules see two thirds the water and one third the reflector. Now we can compare the effective albedo of the reflector before and after the heavy biofouling, which shows a drop from 68.49% to 24.21%, almost three times smaller.

| Electrical yield comparison
The specific DC yield (Wh/Wp) of the nine pilot systems were monitored 1 year, October 2019 until September 2020. The overall performance of the systems is shown in Figure 19 and further broken down into the monthly comparison in Figure 20. The comparison shows that the tracking bifacial modules with reflector (system no. 9) outperform all the other systems by yielding 17.3% more than the reference land-based monofacial modules (system no. 1, as a reference). In a month with more clear sky days (May 2020), this value

| Ecological impacts
Knowledge on the ecological impacts of FPV systems is limited. Currently, most scientific work has focused on the technological advancements, rather than the impacts to the environment. 71,72 When environmental impacts are assessed, the focus is either on temperature 73,74 or fish production, 75 rather than impacts on lake ecological functioning. Studies addressing the potential impacts of FPV on ecosystem are, to-date, theoretical in their scope. 76  Camara traps placed on and around the FPV structures revealed that the FPV systems were repeatedly visited by (water) birds (see Figure 21B). Although birds do visit the FPV systems, the number of birds and the frequency of their visits seemed to be low. However, even these low number of visits had consequences for the performance of the PV systems in short term (see Sections 4.3 to 4.5).
Submerged aquatic plants are a vital part of a healthy shallow lake ecosystem, hampering algal blooms, providing habitat for other organisms, and capturing suspended particles. 78 Hence, the impact of the pilot setup on this biotic group is of great importance to understand.
A significantly higher biomass of aquatic plants was observed in between the FPV systems than under either of the two setups (see Figure 21C). Whereas biomass was comparable to spot 5 ( Figure 21C) Figure 21D).
F I G U R E 2 0 Monthly breakdown of the specific yield for the monitored pilot systems. The green bars represent the ground-based system, the red bars the retractable system, and the blue bars the tracker system. The gray error bars depict the range of the minimum and the maximum energy yield of each module in the respective system. In 1 year, the pilot systems on the floating island have produced 7.26 MWh in total. This means sun-tracking consumes 0.45% of the production (see pumps consumption calculation in Section 3.4). GBS: ground-based system, TS: tracking system, RS For each spot, four subplots were harvested. Further, harvested material was dried for 96 h at 60 C and then weighted. The control site (spot 5), also showed a lower plant biomass than the site between FPV structures (spot 3). However, the species growing at the control site was identified as Stuckenia pectinata whereas aquatic plant species found both between and under the FPV structures were identified as waterweed (Elodea nuttallii). While both plant types are fast growers, Elodea is known to be a very sturdy plant that does well even under disturbed conditions. (D) Temperature at the bottom and dissolved oxygen in mg/L under the retractable PV system (spot 1) and at a control site (spot 5) averaged for both day and night [Colour figure can be viewed at wileyonlinelibrary.com] Such an impact of FPV on oxygen concentrations in the water column can be attributed either to increased consumption of oxygen or decreased production. Plant growth was diminished underneath the panels, which suggests that primary production through photosynthesis will be lower under the PV setup. However, increase in microbial processes consuming oxygen cannot be disregarded, as changes in wind action around the panels may impact sediment formation processes. Regardless of the cause of the increased hypoxia underneath the PV setups, its impact to the ecosystem (e.g., internal loading, impact on fish, and greenhouse gas emissions) can be far reaching. Considering the small scale of the pilot setup and its open nature, these results show that increased attention for the environmental impacts of FPV systems is necessary.

| CONCLUSION
This paper introduced two novel FPV concepts for inland water areas: the retractable and the tumbler island. The introduced concepts can accommodate bifacial PV modules equipped with reflectors and horizontal sun tracking. They also do not disturb mowing activities and can be moved around to enable light penetration into the water. The location survey, modeling, realization, and the results of 1-year monitoring were reported. Through the study, several facts about floating bifacial PV were revealed that oppose the traditional claims. A summary of outcomes is mentioned below.
1. Despite the immediate expectation, inland water areas have low effective albedo (6.5%) which makes water less favorable for bifacial PV installation. Therefore, including reflectors is highly recommended.
2. Frame soaking of the FPV modules has a very minor influence (<0.2%) on the performance and, therefore, does not bring added values.
3. Horizontal sun tracking is possible for FPV structures by adjusting the water content within the floater compartments by the pumps.
This approach consumes low amount of tracking energy (<0.5% of the produced energy).

The birds' presence on the floaters show its effect in the short
term (e.g., bio-fouling reduced the effective albedo of the reflectors from 68% to 24% in 8 months). The surfaces that directly (PV) and indirectly (reflector) play a role in the FPV irradiance capturing should be kept tilted and at a higher level than that of the water. This reduces the birds' presence effects. Active bird control techniques are also recommended. 5. FPV system experience higher bio-fouling rate and will suffer more from temperature spatial variance compared with land-based PV systems located close by. This will boost the aging procedure. 8. Under the FPV systems, periods of hypoxia (<6 mg/L O 2 ) are much more frequent, and anoxia was not found to occur more frequently.
The reported findings and the lessons learned will help the FPV community to identify knowledge gaps, tackle challenges, and improve designs to achieve technological readiness of such setups for the water sector.