Abstract
We present a detailed illumination model for bifacial photovoltaic modules in a large PV field. The model considers direct light and diffuse light from the sky and treats the illumination of the ground in detail, where it discriminates between illumination of the ground arising from diffuse and direct light. The model calculates the irradiance components on arbitrarily many positions along the module. This is relevant for finding the minimal irradiance, which determines the PV module performance for many PV modules. Finally, we discuss several examples. The code for the model is available online (DOI: 10.5281/zenodo.3543570).
Published by The Optical Society under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.
1. Introduction
Solar energy is the most abundant of all sustainable energy technologies [1]. Further, the cost of photovoltaics (PV) has decreased rapidly in the last decade [2]. Therefore, many studies suggest that the largest fraction of global energy needs in a sustainable carbon-free economy can be covered by photovoltaic solar energy [3,4].
Currently, most photovoltaic modules use monofacial solar cells, which only can utilize light that hits the cell at the front. In contrast to that, bifacial solar cells can also convert light impinging at the back and therefore increase the energy yield significantly. Because of novel silicon solar cell technologies using passivated emitter rear contacts (PERC/PERT/PERL) or IBC contacting schemes, bifacial solar cell operation becomes feasible and indeed, they are expected to have 60% market-share by 2029 [5–7].
To estimate the energy yield of bifacial modules it is vital to develop detailed illumination models, especially because the largest fraction of light that hits the module at the back arises from the ground. Several different illumination models were developed in recent years [8]. Sometimes, modules are treated as isolated and shadowing of module rows in front or behind is not considered [9], which overestimates the irradiation onto the ground. Already several authors developed detailed shadowing models of the ground [10–12], however, often variations of the irradiance along the module are neglected [12]. Calcabrini and coworkers developed a simplified model for front illumination using a sky view factor, which quantifies the landscape around the PV module, and a sun coverage factor, which they define ‘at a location with a raised horizon as the ratio between time that the sun is behind the module or blocked by the skyline per year and the annual sunshine duration at the same location with a clear horizon’ [13].
In a recent publication we minimized the levelized cost of electricity (LCOE) of large PV fields of bifacial modules using a Bayesian optimization algorithm [14]. In the present manuscript, we detail the illumination model, which we used in the preceding manuscript. The model, which is described in section 2 considers direct light and diffuse light from the sky and contains a detailed illumination model of the ground, where the model discriminates between illumination of the ground arising from diffuse and direct light. The model calculates the irradiance components on arbitrarily many positions along the module, which is relevant for finding the minimal irradiance, which determines the PV module performance for many PV modules. Finally, we discuss several examples in section 3.
2. Illumination model
With the illumination model we calculate the irradiance onto a solar module, which is placed somewhere in a big PV-field. We assume this field to be so big that effects from the boundaries can be neglected. Further, we assume the modules to be homogeneous: we neglect effects from the module boundaries or module space in between solar cells. Hence, we can treat this problem as 2-dimensional with periodic boundary conditions, as illustrated in Fig. 1. As input we use the direct normal irradiance (DNI) and the diffuse horizontal irradiance (DHI), which we obtain from freely available weather databases, and the solar position. In [14], DNI and DHI were adapted according to the Perez model to account for the diffuse irradiance more accurately [15].
For the front and back sides of the module, we have to consider four irradiance components each – hence eight components in total. These components are summed up in Table 1 and depicted in Fig. 1. In this model we assume the solar modules to be completely black, meaning they do not reflect any light that could reach another module.
The total irradiance (or intensity; given in watt per m$^2$) on front is given by
In section 2.1 we will derive expressions for the four components impinging onto the solar module from the sky. Expressions for the components impinging from the ground are derived in section 2.2. Table 2 shows the input parameters required for calculating the components.
The irradiance $I$ is based on the more fundamental notion radiance $L$ (in watt per steradian per m$^2$), which is connected to the irradiance via
where $\mathbf {r}$ defines the coordinates on the surface on which $I$ and $L$ are evaluated [16]. $\theta$ and $\phi$ denote the polar angle and azimuth, where $\theta$ is defined such that it denotes the angle between the evaluated ray and the normal to the surface at $\mathbf {r}$. The term $\cos \theta$ arises from Lambert’s cosine law, which accounts for the decreasing irradiance of a beam of light which strikes a surface under increasing angle, because the radiant power is distributed across a larger area. The solid angle element is given by $\,\textrm {d}\Omega = \sin \theta \,\,\textrm {d}\theta \,\,\textrm {d}\phi$.For the following derivation we define dimensionless geometrical distribution functions as
2.1 Irradiance components from the sky
The direct irradiance components from the sky are given by
For the diffuse irradiance from the sky $I_{\textrm {diff, }f}^{\textrm {sky}}(s)$ on the module front we have to consider the angular range from which light can reach the module. This concept is also known as view factors (VF) in literature [8]. Generally speaking, diffuse light can reach the module front from an opening that spans from the top of the front row $D_{-1}$ to the plane of the module itself, marked by the topmost point $D_0$. In the example in Fig. 1, point $P_m$ on module $\#0$ is illuminated by light from directions within $\sphericalangle D_{-1}P_mD_0$. With respect to the module normal, this range is constrained by the angles $\alpha _{s1} = -\pi /2$ and $\alpha _{s2}(s)$. However, diffuse light does not only reach the module from directions within the $xz$-plane but from a spherical wedge as explained in the appendix. Under the assumption that diffuse light arrives isotropically from the upper hemisphere and using Eqs. (22) and (24) we find
2.2 Irradiance components from the ground
As illustrated in Fig. 2, the ground receives direct and diffuse light from the sky, where we introduce the irradiances $G_{\textrm {dir}}(x_g)$ and $G_{\textrm {diff}}(x_g)$. We assume the ground to be a Lambertian reflector with albedo $A$. Under this assumption the ground emits isotropically with the radiance $L^{\textrm {gr.}}(x_g)$, which is connected to the irradiance $G(x_g)$ via
where we omitted the subscripts “diff” and “dir”. Similar to Eq. (3) we define geometrical distribution functions on the ground:The direct geometrical distribution function is position dependent and takes values of 0 for shaded areas (no direct sunlight) and the projection of the direct irradiance on the ground otherwise:
The diffuse geometrical distribution function describes the fraction of the sky that is visible at a certain location on the ground. For the example shown in Fig. 2, light can reach $P_g$ from three ranges: $\sphericalangle D_{-1}P_gB_0$, $\sphericalangle D_0P_gD_1$ and $\sphericalangle B_1P_gD_2$. Using Eq. (22), we find
Now we derive expressions for the irradiance components, which illuminate the module from the ground: $I_{\textrm {dir, }f}^{\textrm {gr.}}$, $I_{\textrm {diff, }f}^{\textrm {gr.}}$, $I_{\textrm {dir, }b}^{\textrm {gr.}}$ and $I_{\textrm {diff, }b}^{\textrm {gr.}}$. The following derivation is valid for both components originating either from direct sunlight or diffuse skylight. Hence, we omit the subscripts “dir” and “diff” in the following. As illustrated in Fig. 1, the relevant angular range is given by $\sphericalangle B_{-1}P_mB_0$ for the front and $\sphericalangle B_0P_mB_1$ for the back, respectively. With respect to the module normal, this range is constrained by the angles $\alpha _{g1}(s)$ and $\alpha _{g2} = \pi /2$ for the front and by the angles $\epsilon _{g1}=-\pi /2$ and $\epsilon _{g2}(s)$ for the back. The diffuse irradiance components from the ground can be calculated using Eq. (21) in the appendix,
2.3 Implementation details
The implementation is written in Python and utilizes the NumPy scientific computing package fast tensor operations. The code was implemented with the fast evaluation of time-series data for energy yield calculations in mind. For a fixed-array geometry and changing solar positions only the geometric distribution functions for direct light $\iota _{\textrm {dir}}$ have to be calculated separately for every timestamp while the different $\iota _{\textrm {diff}}$ are time independent:
To account for reflected light from the ground the integral in Eq. (13) would need to be solved for every timestep. To avoid this, the ground is discretized into a finite amount of elements $N_g$. Within each element the radiance is assumed to be constant. This allows to reformulate Eq. (13) into a sum over all elements visible to the module. By introducing the ground view factor VF$_g$ the geometrical calculations are not depending on the timestamp anymore, which is very beneficial for the speed of numerical evaluation:
3. Some examples
In this section we demonstrate the model with a few examples. We assume a PV system which is located in Berlin, Germany (52.5°N, 13.25°E). Based on a rule-of-thumb rule we set the module tilt approximately to the geographical latitude, $\theta _m = 52^\circ$ and determine the distance such that a module does not shade a module behind on 21 December, 12 noon [11,18]. For modules with $\ell = 1.96$ m length this leads to a distance between module rows of $d=7.3$ m. A sketch of this configuration is shown in Fig. 3(a). The solar positions are calculated using the Python package pysolar [19].
Figure 3(b) shows the geometrical distribution functions $\gamma$ for direct and diffuse components on the ground for 20 June 2019, 11:52 am CEST. $\gamma _{\textrm {diff}}$ is minimal below the module where the angle covered by the module is largest; and maximal at $x'$, because here the ground sees (almost) no shadow from module #1.
Figure 4 shows how the direct geometrical distribution function on the ground $\gamma _{\textrm {dir}}$ develops during three days (20 June, 23 September and 20 November 2019). For 20 June, a large fraction of the ground is illuminated throughout the day. After sunrise and before sunset the Sun is below the module plane and direct sunlight hits the module back. During short amounts of time in the morning and evening the Sun lies in the module plane and no direct sunlight hits the module. At these times the shadow on the ground vanishes (around 7 am and 7 pm). In between these times the Sun is above the plane and direct light hits the module front. On 23 September, the length of the shadow is almost the same during the whole day: the module tilt ($52^\circ$) is almost the same as the latitude of Berlin ($52.5^\circ )$ - hence at equinox the ecliptic is practically normal to the module plane. On 20 November, only short stretches of the ground are illuminated.
Figure 5 shows an example for the eight geometrical distribution functions $\iota$ corresponding to the irradiance components hitting the PV module on its front and back sides. While the functions originating from the sky (a) are stronger on the front side, the components originating from the ground (b) are stronger on the back side. This can be understood by the opening angles: the opening angle towards the sky is larger on the front side, but the opening angle of the ground is larger at the back. The largest relative variations are observed for the ground component at the module back side originating from direct sunlight: it is strongest for short $s$-values. This can be understood when looking at Fig. 3(b): the ground is illuminated by sunlight directly below the module.
In Fig. 6 we look at two examples on how the irradiance distribution along the module varies depending on the irradiation conditions. We utilized irradiation data measured at the Hochschule für Technik und Wirtschaft Berlin on 20 June 2019. DNI data was measured with the SHP1 pyrheliometer and DHI data was obtained with the SMP1 pyranometer [20]. Figure 6(a) shows results for 11:52 am CEST. Then, no DNI (0 W/m$^2$) was measured while DHI was 144 W/m$^2$ – hence clouds covered the Sun. As discussed in [14], for many PV modules, such as silicon modules composed of several single solar cells connected in series, the overall module performance is not determined by the mean irradiance but by the lowest irradiance observed by the module. Both the front and total irradiation are lowest at the lower end of the module. The total irradiance at the upper end is 10.7% higher than at the lower end. Figure 6(b) shows the situation 46 minute later, at 12:38 pm. Then, the DNI (883 W/m$^2$) is much higher than the DHI (134 W/m$^2$) – direct sunlight hits the PV module. Here, the irradiation onto the module front is almost independent of the module position, the maximum variations in total irradiance are below 0.5%. The total irradiation, which would be relevant for bifacial modules, is minimal at $s\approx 0.57$ m
The varying position of the irradiance minimum shows that it is very relevant to determine the irradiance components as functions along the module position $s$, which is automatically done by the model presented in this manuscript. Models based on view factors often have only mean irradiances as output, which might overestimate the final module performance.
4. Conclusions and outlook
Accurate illumination models are crucial for yield estimations of PV power plants with bifacial solar cells. In this manuscript, we derived a detailed model that takes all relevant irradiation components onto PV modules into account and allows for a quick calculation. The code can be accessed online [17]. The irradiation components are calculated locally resolved along the module, which enables the user to determine the minimal irradiance, which is dependent on the illumination conditions, as illustrated in Fig. 6. This is relevant for accurately estimating the performance for many types of PV modules.
In a next step, we will expand the model such that the irradiance components onto the module are angular resolved. This is especially relevant when connecting the illumination model to optical models of solar cells. Further, it would be desirable to experimentally assess the model described in this manuscript in the future.
5. Appendix. Integration of radiance across a spherical wedge
As shown in Fig. 1, the different diffuse components reach a point $P_m$ on the solar module from certain angular ranges. However, diffuse light does not only reach the module from directions within the $xz$-plane, but from a spherical wedge, as illustrated in Fig. 7.
The irradiance can be calculated with
where $\Omega$ is the set of solid angles defined by the wedge and $\theta$ is the angle spanned between the directional unit vector $\mathbf {e}(\alpha ,\beta )$ and the normal $\mathbf {n}$. With the spherical coordinate system defined in Fig. 7, we have The cosine factor can be expressed asFunding
Helmholtz Einstein International Berlin Research School in Data Science (HEIBRiDS); Helmholtz Association via Helmholtz Excellence Network SOLARMATH (ExNet-0042-Phase-2-3).
Acknowledgments
We thank Lev Kreinin and Asher Karsenti from SolAround for fruitful discussions regarding the illumination model for bifacial solar cells. The results were obtained at the Berlin Joint Lab for Optical Simulations for Energy Research (BerOSE) of Helmholtz-Zentrum Berlin für Materialien und Energie, Zuse Institute Berlin and Freie Universität Berlin, and within the Helmholtz Excellence Network SOLARMATH, a strategic collaboration of the DFG Excellence Cluster MATH+ and Helmholtz-Zentrum Berlin für Materialien und Energie.
Disclosures
The authors declare no conflict of interest.
References
1. F. Creutzig, P. Agoston, J. C. Goldschmidt, G. Luderer, G. Nemet, and R. C. Pietzcker, “The underestimated potential of solar energy to mitigate climate change,” Nat. Energy 2(9), 17140 (2017). [CrossRef]
2. N. M. Haegel, H. Atwater, T. Barnes, C. Breyer, A. Burrell, Y.-M. Chiang, S. D. Wolf, B. Dimmler, D. Feldman, S. Glunz, J. C. Goldschmidt, D. Hochschild, R. Inzunza, I. Kaizuka, B. Kroposki, S. Kurtz, S. Leu, R. Margolis, K. Matsubara, A. Metz, W. K. Metzger, M. Morjaria, S. Niki, S. Nowak, I. M. Peters, S. Philipps, T. Reindl, A. Richter, D. Rose, K. Sakurai, R. Schlatmann, M. Shikano, W. Sinke, R. Sinton, B. Stanbery, M. Topic, W. Tumas, Y. Ueda, J. van de Lagemaat, P. Verlinden, M. Vetter, E. Warren, M. Werner, M. Yamaguchi, and A. W. Bett, “Terawatt-scale photovoltaics: Transform global energy,” Science 364(6443), 836–838 (2019). [CrossRef]
3. M. Z. Jacobson, M. A. Delucchi, Z. A. Bauer, S. C. Goodman, W. E. Chapman, M. A. Cameron, C. Bozonnat, L. Chobadi, H. A. Clonts, P. Enevoldsen, J. R. Erwin, S. N. Fobi, O. K. Goldstrom, E. M. Hennessy, J. Liu, J. Lo, C. B. Meyer, S. B. Morris, K. R. Moy, P. L. O’Neill, I. Petkov, S. Redfern, R. Schucker, M. A. Sontag, J. Wang, E. Weiner, and A. S. Yachanin, “100% Clean and Renewable Wind, Water, and Sunlight All-Sector Energy Roadmaps for 139 Countries of the World,” Joule 1(1), 108–121 (2017). [CrossRef]
4. M. Child, C. Kemfert, D. Bogdanov, and C. Breyer, “Flexible electricity generation, grid exchange and storage for the transition to a 100% renewable energy system in Europe,” Renew. Energy 139, 80–101 (2019). [CrossRef]
5. T. Dullweber, C. Kranz, R. Peibst, U. Baumann, H. Hannebauer, A. Fülle, S. Steckemetz, T. Weber, M. Kutzer, M. Müller, G. Fischer, P. Palinginis, and H. Neuhaus, “PERC+: industrial PERC solar cells with rear Al grid enabling bifaciality and reduced Al paste consumption,” Prog. Photovoltaics 24(12), 1487–1498 (2016). [CrossRef]
6. S. Chunduri and M. Schmela, “Bifacial Solar Technology Report 2018 edition,” Tech. rep., TaiYang News (2018).
7. “International Technology Roadmap for Photovoltaic (ITRPV), Tenth edition,” Tech. rep., VDMA (2019).
8. T. S. Liang, M. Pravettoni, C. Deline, J. S. Stein, R. Kopecek, J. P. Singh, W. Luo, Y. Wang, A. G. Aberle, and Y. S. Khoo, “A review of crystalline silicon bifacial photovoltaic performance characterisation and simulation,” Energy Environ. Sci. 12(1), 116–148 (2019). [CrossRef]
9. R. Schmager, M. Langenhorst, J. Lehr, U. Lemmer, B. S. Richards, and U. W. Paetzold, “Methodology of energy yield modelling of perovskite-based multi-junction photovoltaics,” Opt. Express 27(8), A507–A523 (2019). [CrossRef]
10. U. A. Yusufoglu, T. M. Pletzer, L. J. Koduvelikulathu, C. Comparotto, R. Kopecek, and H. Kurz, “Analysis of the Annual Performance of Bifacial Modules and Optimization Methods,” IEEE J. Photovolt. 5(1), 320–328 (2015). [CrossRef]
11. L. Kreinin, A. Karsenty, D. Grobgeld, and N. Eisenberg, “PV systems based on bifacial modules: Performance simulation vs. design factors,” in Proceedings of the 43rd IEEE Photovoltaic Specialists Conference (PVSC), (IEEE, 2016), pp. 2688–2691.
12. B. Marion, S. MacAlpine, C. Deline, A. Asgharzadeh, F. Toor, D. Riley, J. Stein, and C. Hansen, “A practical irradiance model for bifacial PV modules,” in Proceedings of the 44th IEEE Photovoltaic Specialist Conference (PVSC), (IEEE, 2017), pp. 1537–1542.
13. A. Calcabrini, H. Ziar, O. Isabella, and M. Zeman, “A simplified skyline-based method for estimating the annual solar energy potential in urban environments,” Nat. Energy 4(3), 206–215 (2019). [CrossRef]
14. P. Tillmann, K. Jäger, and C. Becker, “Minimising levelised cost of electricity of bifacial solar panel arrays using bayesian optimisation,” Sustain. Energy Fuels 4(1), 254–264 (2020). [CrossRef]
15. R. Perez, R. Stewart, R. Seals, and T. Guertin, “The development and verification of the Perez diffuse radiation model,” Tech. rep., Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA (United States) (1988).
16. J. E. Harvey, C. L. Vernold, A. Krywonos, and P. L. Thompson, “Diffracted radiance: A fundamental quantity in nonparaxial scalar diffraction theory,” Appl. Opt. 38(31), 6469–6481 (1999). [CrossRef]
17. P. Tillmann and K. Jäger, Bifacial illumination model, (2019). DOI: 10.5281/zenodo.3543570.
18. P. Grana, “The new rules for latitude and solar system design,” (2018). https://www.solarpowerworldonline.com/2018/08/new-rules-for-latitude-and-solar-system-design/.
19. B. Stafford, “Pysolar,” (2018). DOI: 10.5281/zenodo.1461066.
20. V. Quaschning, T. Tjaden, J. Weniger, and J. Bergner, “HTW weather station,” (2019). https://wetter.htw-berlin.de/History/DIF_SMP5,DIR_SHP1/2019-06-20/2019-06-21.