Methods for modelling and analysis of bendable photovoltaic modules on irregularly curved surfaces
 1.5k Downloads
 5 Citations
Abstract
Most photovoltaic modules are planar and as a result, research on panel layout for photovoltaic systems typically uses planar panels. However, the increased availability of thinfilm photovoltaic modules opens up possibilities for the application of flexible solar panels on irregularly curved surfaces, including the integration of photovoltaic panels on building roofs with double curvature. In order to efficiently arrange photovoltaic panels on such surfaces, geometric CAD tools as well as radiation analysis tools are needed. This paper introduces a method to generate geometry for flexible photovoltaic modules on curved surfaces, as well as a method to arrange multiple of such modules on a surface. By automating the generation of possible photovoltaic panel arrangements and linking the geometric tools to solar analysis software, large numbers of design options can be analysed in a relatively short time. This combination of geometry generation and solar analysis provides data that is important for electrical design of photovoltaic systems. The merits of the methods we introduce are illustrated with a case study, for which hundreds of design configurations have been explored in an automated manner. Based on analysis of the numeric data generated for each of the configurations, the effects of panel dimensions and orientation on solar insolation potential and panel curvature have been established. The quantitative and qualitative conclusions resulting from this analysis have informed the design of the photovoltaic system in the case study project.
Keywords
Flexible photovoltaic modules Geometry Double curvature Surface rationalisation Solar analysisIntroduction
Strong cost reductions and increasingly stringent government regulations combined with support schemes for renewable energy in several countries are leading to a rapid deployment of photovoltaic (PV) systems. Many studies exist on the topic of placement and orientation of PV panels on roofs, facades or the ground, but solely taking into account flat modules on planar surfaces [1, 2, 3, 4, 5]. With the increased availability of flexible thinfilm PV panels in recent years, the application of PV panels on curved surfaces becomes feasible, thus broadening the application potential of photovoltaic modules. Curvature poses some major challenges related to structural integration, PV module operation and electrical system design [6, 7]. However, so far no methodology for modeling of flexible PV modules on curved surfaces exists. In this paper, we introduce design and analysis methods for the application of flexible PV panels on irregularly curved surfaces.
The design of buildings that are optimised for structural efficiency [8] or tailored for digitally informed fabrication [9, 10] often results in complex geometry [11, 12, 13], which may include roofs and facades featuring double curvature. Due to their rigidity and shape, most PV panels are difficult to integrate in surfaces with double curvature. However, thinfilm photovoltaic modules can be applied to thin sheet metal or flexible polymer substrates, offering more geometric flexibility. Combined with the high efficiency and low weight of thinfilm PV technologies such as CIGS (copper indium gallium selenide), this offers new application possibilities of PV modules for building integrated photovoltaics (BIPV) as well as portable PV applications [14]. With this technique, thinfilm photovoltaic modules could potentially be directly integrated in the building envelope. As this would remove the need for the substructures that are typically used for PV installations, such a system could be lightweight and costeffective.
Flexible sheet metal panels can be applied to surfaces with single curvature easily and have historically been used for roofing [15]. However, when applying bendable panels to surfaces with double curvature, methods to predict the geometric behaviour of such panels are needed. When these geometric methods are linked to solar analysis, the amount of solar insolation can be predicted at any desired spatial resolution. Such solar insolation data is important for the electrical design of PV systems: taking shading conditions into account reduces electrical mismatch and thus results in significantly higher electricity generation [16, 17].
In a pilot project, the geometric methods we developed have been used to generate hundreds of configurations of flexible PV panels on an irregularly curved roof. After running solar insolation simulations for each configuration, the influence of panel dimensions and panel orientation on solar insolation were analysed.
This paper starts by introducing methods to approximate doublecurved geometry using developable strips. After assessing the validity of these methods, we then discuss how the resulting geometry can be analysed on a number of metrics: total surface area, occurrence of bending, surface approximation accuracy and solar insolation. We then show the application of these methods to a case study building and discuss the results, focusing on the influence of various design parameters on the total solar insolation.
Methods
In this section, we introduce methods to generate strips of bendable photovoltaic panels by approximating a doublecurved surface using two different triangulation approaches (2.1–2.3), to efficiently arrange multiple of these strips on a larger surface (2.4) and to analyse the resulting geometry with regard to various geometric metrics (2.5) as well as solar insolation (2.6).
Panel generation method A: congruent triangles
Our aim is to generate an approximation of a flexible panel that is bent over a doublecurved surface, following the surface as closely as possible while remaining developable and resulting in an exactly rectangular shape when unrolled.
 1.
Create two points (P0, P1) on the base surface, such that the distance between them is the desired panel width (w) and the direction is perpendicular to the desired panel direction. Connect the points with a line (L0).
 2.
Create a circle with radius R0 around point P1, using a plane perpendicular to line L0. Create a point (P2) at the intersection of this circle with the base surface.
 3.
Create two lines (L1, L2) connecting point P2 to points P0 and P1, thus creating triangle T0 with angle α at P0.
 4.
Create a point (P3) on L2 at distance \(w \cdot \cos \left( \alpha \right)\) from P2, then create a circle perpendicular to L2 around P3 using radius \(w \cdot { \sin }\left( \alpha \right)\). Intersect this circle with the base surface, creating point P4.
 5.
Connect P4 with two lines to points P0 and P2, creating triangle T1. Note that triangles T0 and T1 are congruent.
 6.
Repeat steps 2–5 as many times as necessary.
Panel generation method B: adaptive triangles
 1.
Create two points (P0, P1) on the base surface, such that the distance between them is the desired panel width (w) and the direction is perpendicular to the desired panel direction. Connect the points with a line.
 2.
Create a series of lines starting at P0 and ending at points on the surface with distance \(n \cdot d\) from P1 and \(\sqrt {\left( {n \cdot d} \right)^{2} + w^{2} }\) from P0, where n is the number of lines and d is a distance that can be chosen at will. Create a second series of lines starting at point P1. Select the line that has the smallest distance between its midpoint and the base surface and call the endpoint of this line P2.
 3.
Draw a triangle using points P0, P1 and P2.
 4.
Continue creating triangles by repeating steps 2 and 3. Note that in some cases, more than two diagonal lines meet in one point, such as point 5 in Fig. 3.
 5.
The end of the strip is a special case: if an exact strip length is desired, the value for distance d in step 2 should be set to the remaining edge length divided by an integer (which may be chosen at will). Once one of the corner points of the strip has been reached, potential end points can be created on the panel’s short edge.
Assessment of panel triangulation methods
In a test of 46 triangulated strip segments generated with method A, the median lateral deviation between strips using different diagonal directions turned out to be 0.14 % of the strip length. In the most extreme case, the lateral deviation was 2.1 %.
Comparing method A with method B, the median lateral deviation on our test geometry is 0.20 %. The most extreme deviation occurring is 3.7 %.
Two scale models were created to test the sheet metal strip behaviour: one supported by elements perpendicular to the strip direction (generated with method A), and one supported by elements oriented roughly along the main curvature direction (using a subset of edges generated using method B).
An observation relevant to the fixation method of the panels is that perpendicular supports visibly deform the metal surface in areas of concave curvature, except when the main curvature direction is aligned with the strip. To avoid this visible deformation, the supports for the panels can be aligned to the local curvature direction of the strips.
Geometric method for strip arrangement
The methods shown above can be used to generate single strips, but in order to cover a larger area, a method to arrange multiple strips is needed. To maximise solar insolation, we are looking for a solution that fits as many panels as possible. As an additional architectural constraint, we choose to only look at solutions where multiple panels are arranged in long strips.
On a surface with double curvature, strips that are parallel to each other at one position will diverge or overlap elsewhere. We aim to avoid overlapping and to keep strips roughly parallel to each other, so that the unused area between panels is minimised. However, in order to facilitate the installation process, a certain minimum distance between the panels is defined.
Calculation of bending and of surface approximation accuracy
The amount of module bending and the accuracy of surface approximation are important metrics as they strongly influence buildability, detail design and visual appearance.
In order to get an estimate of the geometric deviations between the original surface and the generated strips, we measure the distance between various points of the strip geometry and the nearest points on the original surface. As all points on both sides of the generated strips lie exactly on the base surface, the points with the largest surface deviation typically lie very close to the centre line of the strips. As with the bending radii, this surface approximation accuracy is exported both graphically (see Fig. 9, right) and numerically.
Calculation of solar insolation
The solar insolation of all generated panel configurations has been analysed using the EnergyPlus building energy simulation software [22], which we accessed through the Ladybug [23] plugin for Grasshopper [24], which is a parametric modelling and programming environment for the 3d modelling program Rhinoceros [25]. Using EnergyPlus weather files with hourly resolution, Ladybug calculates irradiance on the modules with the cumulative sky approach [26, 27]. For the analysis presented in this work, the spatial resolution of the striplike panels is ten triangular faces per stretching meter.
As weather data source, we used IWEC weather data for Geneva [28], covering one calendar year at one hour intervals. Using custom C# scripts, both graphical and numerical output were generated. Sun paths generated with the DIVA [29] plugin for Grasshopper were then used to visually study shadow occurrence on the roof, using the physically based render engine LuxRender [30].
The curved geometry leads to varying irradiance on the modules, which in turn induces electrical mismatch between cells and modules [31, 32]. Though not investigated in this paper, coupling module irradiance with an electrical model can be used to predict PV module powervoltage characteristics. This is important for efficient system design such as the development of advanced maximum power point tracking methods [31, 32] or distributed power electronics [33].
Results
Comparative analysis of various roof shell shapes
Analysis of design parameters, as applied to case study roof
 Panel width

0.30, 0.45, 0.60, 0.75 and 0.90 m
 Panel length

1.2, 1.6, 2.0, 2.4 and 2.8 m
 Strip orientation

13 angles, at 15° intervals
In total, this resulted in data for 325 design alternatives. A detailed analysis of this data applied to one specific roof shape is presented in the following sections.
Surface area as a function of panel dimensions

The available length for a strip is rarely an exact multiple of the panel length, so part of the available area will not be used. The size of the unused area depends on the angle between the panel and the edge, and on the panel length. On average, more than half a panel length of potential space is lost on each strip. Using shorter panels helps reduce these losses.

The effect of panel width on active surface area is clear: the use of wider panels results in a smaller loss of potential roof area, as there are fewer gaps between panels.
 In practice, photovoltaic panels often have an inactive edge area. The losses caused by these edges are relatively strong for narrow and/or short panels. Still, for common panel sizes, short, wide panels result in the largest active PV surface in our case study, as can be seen in Fig. 12.
Active panel surface area as a function of panel strip orientation
Strip bending as a function of strip orientation
On the topic of bending, considering the possible arrangements of strips on a cylinder shows that the strip orientation can have a large impact on the amount of bending within the strips as it can range from 0 (following the direction of the cylinder) to totally curved (following the circumference of the cylinder). The same thought experiment suggests that the width of the strips does not affect the amount of bending in the length direction of the strips.
Effect of panel dimensions on surface approximation accuracy
Solar insolation
Conclusions
In this paper, we introduce methods to design and analyse photovoltaic systems using flexible panels, which facilitates the application of photovoltaic systems on curved surfaces where other photovoltaic systems would not be suitable. Thanks to the systematic generation of flexible panel geometry, we were able to identify the influence of various geometric parameters (including panel dimensions and panel arrangement) on the potential surface area of photovoltaic panels on a doublecurved roof, as well as on the expected solar irradiation on such panels.
We have introduced geometric methods to approximate doublecurved geometry using triangulated strips, as well as methods to organise such strips efficiently on a surface. By combining these methods with solar insolation analysis software, we analysed the solar insolation potential of various roof shell geometries in a case study project. We also studied the impact of various geometric parameters on solar insolation, module curvature and the size of the gaps between the flexible panels and the roof surface.
For the roof geometry we studied, the solar insolation is almost perfectly linearly dependent on the panel surface area. Short and wide panels that are oriented mostly perpendicularly to the longest edges of the roof resulted in the largest effective PV area and in the highest solar insolation. On the other hand, narrow panels result in less geometric deviation between the flexible panels and the roof surface.
The methods we introduced proved to work reliably and efficiently in our case study, despite the geometric complexity of the roof. This suggests that the methods work for a wide range of shapes. Because the strip geometry generation and irradiance analysis are automated and take little time to calculate, the system described in this paper could potentially be linked to other computational design tools. Data generated using the presented method could be used to inform the electrical design of photovoltaic systems.
Notes
Acknowledgments
This research has been financially supported by Swiss Commission for Technology and Innovation (CTI) within the Swiss Competence Center for Energy Research (SCCER) Future Energy Efficient Buildings & Districts (FEEB&D, CTI.2014.0119) and by the Building Technologies Accelerator program of ClimateKIC. The base geometry for the case study roof has been kindly provided by Diederik Veenendaal at the Block Research Group, ETH Zurich.
References
 1.DíazDorado, E., SuárezGarcía, A., Carrillo, C.J., Cidrás, J.: Optimal distribution for photovoltaic solar trackers to minimize power losses caused by shadows. Renew. Energy 36(6), 1826–1835 (2011)CrossRefGoogle Scholar
 2.Ng, K.M., Adam, N.M., Inayatullah, O., Kadir, M.Z.A.A.: Assessment of solar radiation on diversely oriented surfaces and optimum tilts for solar absorbers in Malaysian tropical latitude. Int. J. Energy Environ. Eng. 5(1), 1–13 (2014)CrossRefGoogle Scholar
 3.Gordona, J.M., Wengera, H.J.: Centralstation solar photovoltaic systems: field layout, tracker, and array geometry sensitivity studies. Sol. Energy 46(4), 211–217 (1991)CrossRefGoogle Scholar
 4.Choo, T.S., Janssen, P.: Evolutionary optimisation of semitransparent building integrated photovoltaic facades. Int. J. Archit. Comput. 12(1), 81–100 (2014)CrossRefGoogle Scholar
 5.Sun, L., Lu, L., Yang, H.: Optimum design of shadingtype buildingintegrated photovoltaic claddings with different surface azimuth angles. Appl. Energy 90, 233–240 (2012)CrossRefGoogle Scholar
 6.Hofer, J., Nagy, Z., Schlueter, A.: Electrical design and layout optimization of flexible thinfilm photovoltaic modules. In: EU PVSEC conference, Munich (2016)Google Scholar
 7.Sharma, P., Duttagupta, S.P., Agarwal, V.: A novel approach for maximum power tracking from curved thinfilm solar photovoltaic arrays under changing environmental conditions. IEEE. Trans. Ind. Appl. 50(6), 4142–4151 (2012)CrossRefGoogle Scholar
 8.Rippmann, M., Lachauer, L., Block, P.: Interactive vault design. Int. J. Space. Struct. 27(4), 219–230 (2012)CrossRefGoogle Scholar
 9.Williams, N., Stehling, H., Scheurer, F., Oesterle, S., Kohler, M., Gramazio, F.: A case study of a collaborative digital workflow in the design and production of formwork for ‘nonstandard’ concrete structures. Int. J. Archit. Comput. 9(3), 223–240 (2011)CrossRefGoogle Scholar
 10.Schwinn, T., Krieg, O., Menges, A.: Robotically fabricated wood plate morphologies. In: BrellCokcan, S., Braumann, J. (eds.) RobArch 2012, pp. 48–61. Springer, Vienna (2013)Google Scholar
 11.Park, P., Gilbert, M., Tyas, A., PopovicLarsen, O.: Potential use of structural layout optimization at the conceptual design stage. Int. J. Archit. Comput. 10(1), 13–32 (2012)CrossRefGoogle Scholar
 12.Scheurer, F.: Materialising complexity. Archit. Design 80(4), 86–93 (2010)CrossRefGoogle Scholar
 13.Kolarevic, B. (ed.): Architecture in the digital age: design and manufacturing. Spon, New York (2003)Google Scholar
 14.Lin, Q., et al.: Flexible photovoltaic technologies. J. Mater. Chem. C. 2(7), 1233–1247 (2014)CrossRefGoogle Scholar
 15.Yalden, J.E.G.: The design and construction of a small observatory. Pop. Astron. 28, 449–458 (1920)Google Scholar
 16.Hofer, J., Groenewolt, A., Jayathissa, P., Nagy, Z., Schlueter, A.: Parametric analysis and systems design of dynamic photovoltaic shading modules. Energy Sci. Eng. 4(2), 134–152 (2016)CrossRefGoogle Scholar
 17.Dolara, A., Lazaroiu, G.C., Leva, S., Manzolini, G.: Experimental investigation of partial shading scenarios on PV (photovoltaic) modules. Energy 55, 466–475 (2013)CrossRefGoogle Scholar
 18.Kidder, F.S.: Triangulation applied to sheet metal pattern cutting. The Sheet Metal Publication Company, New York (1917)Google Scholar
 19.Frey, W.H.: Boundary triangulations approximating developable surfaces that interpolate a closed space curve. Int. J. Found. Comput. Sci. 13, 285–302 (2002)MathSciNetCrossRefzbMATHGoogle Scholar
 20.Tang, K., Wang, C.L.: Modeling developable folds on a strip. J. Comput. Inf. Sci. Eng. 5(1), 35–47 (2005)CrossRefGoogle Scholar
 21.Wallner, J., Schiftner, A., Kilian, M., Flöry, S., Höbinger, M., Deng, B., Huang, Q., Pottmann, H.: Tiling freeform shapes with straight panels: algorithmic methods. In: Hesselgren, L., Pauly, M., Pottmann, H., Wallner, J. (eds.) Advances in architectural geometry 2010, 17th edn, pp. 73–86. Springer, Vienna, New York (2010)CrossRefGoogle Scholar
 22.Crawley, D.B., Pedersen, C.O., Lawrie, L.K., Winkelmann, F.C.: EnergyPlus: energy Simulation Program. ASHRAE J. 42(4), 49–56 (2000)Google Scholar
 23.Roudsari, M.S., Pak, M., Smith, A.: Ladybug: a parametric environmental plugin for grasshopper to help designers create an environmentallyconscious design. In: Proceedings of Building Simulation 2013: 13th Conference of International Building Performance Association, IBPSA, pp. 26–28. Chambery (2013)Google Scholar
 24.Grasshopper—algorithmic modeling for Rhino. http://www.grasshopper3d.com. Accessed 23 Oct 2015
 25.Rhinoceros. https://www.rhino3d.com/. Accessed 23 Oct 2015
 26.Robinson, D., Stone, A.: Irradiation modelling made simple: the cumulative sky approach and its applications. In: PLEA Conference, 19–22 (2004)Google Scholar
 27.Ibarra, D., Reinhart, F.: Solar availability: a comparison study of six irradiation distribution methods. In: Proceedings of Building Simulation 2011: 12th Conference of International Building Performance Simulation Association. Sydney (2011)Google Scholar
 28.ASHRAE, International Weather for Energy Calculations (IWEC Weather Files) Users Manual and CDROM. ASHRAE, Atlanta (2001)Google Scholar
 29.Jakubiec, A., Reinhart, C.F.: DIVA 2.0: integrating daylight and thermal simulations using Rhinoceros 3D, DAYSIM and EnergyPlus. In: Proceedings of Building Simulation 2011, 12th Conference of International Building Performance Simulation Association, pp. 2202–2209. IBPSA, Sydney (2011)Google Scholar
 30.LuxRender. http://www.luxrender.net. Accessed 23 Oct 2015
 31.Konstantopoulos, C., Koutroulis, E.: Global maximum power point tracking of flexible photovoltaic modules. IEEE. Trans. Power Electron. 29(6), 2817–2828 (2014)CrossRefGoogle Scholar
 32.Sharma, P., Duttagupta, S.P., Agarwal, V.: A novel approach for maximum power tracking from curved thinfilm solar photovoltaic arrays under changing environmental conditions. IEEE. Trans. Ind. Appl. 50(6), 4142–4151 (2014)CrossRefGoogle Scholar
 33.Mazumdar, P., Enjeti, P. N., Balog, R. S.: Smart PV modules—design considerations. In: 2012 IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES). IEEE (2012)Google Scholar
 34.NEST HiLo: ultralightweight & superintegrated. http://hilo.arch.ethz.ch. Accessed 23 Oct 2015
 35.Veenendaal, D., Block, P.: Design process for a prototype concrete shells using a hybrid cablenet and fabric formwork. Eng. Struct. 75, 39–50 (2014)CrossRefGoogle Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.