1. Introduction

Concentrator Photovoltaics (CPV) maximizes the electrical generation per module area by concentrating sunlight into small high-efficient photovoltaic solar cells [1]. In this line, micro-concentrator photovoltaics (micro-CPV) downsizes module components to increase performance and potentially reduce the cost of electricity [2]. This also allows exploring unconventional approaches that would not be neither cost effective nor efficient in traditional macro-scale CPV due to large material consumption or high bulk absorption.

While concentrator modules must be aligned to the sun, the scale of micro-CPV allows the sun-tracking mechanism to be embedded in the module, instead of using a bulky external tracker structures. This is called integrated tracking [3], where relative displacements of a few millimeters between optics and cells keeps the light spot focused on the cell at different angles of incidence (AOI) [4–6]. Consequently, it is possible to mount micro-CPV modules featuring integrated tracking in fixed tilt position (e.g., in rooftops as conventional PV), while preserving the superior efficiency of the CPV technology. The major advantage of the embedded tracker is that provides unique capabilities and market differentiation to conventional silicon flat modules [4]. Hybrid architectures, where the backplane is covered with monofacial or bifacial silicon cells, harvest separately Direct Normal Irradiation (DNI) and diffuse irradiance components providing superior efficiencies for area-constrained applications [7]; translucent architectures, where diffuse light is transmitted through the module and DNI can be either converted to electricity or transmitted through, provides added value for dual-land-use applications such as agrivoltaics, or for daylight control in building integrated photovoltaics.

The micro-scale implies a huge increase in the number of optics-cell units per square meter, and in turn a need for parallel assembly of the components in the module. Moreover, the mechanical tolerances in the manufacturing of micro-CPV modules are very tight, so high precision manufacturing processes are necessary to ensure accurate positioning of the optical components and the cells [8]. Misalignments among optics-cell units cause light leakages, reduce module fill-factor and degrade performance. Thus, inspecting and characterizing the misalignments between the module components may provide a valuable feedback and serve as a quality control tool on the production line [9,10].

Misalignments characterization of CPV modules has been widely described [11,12]. One such device, called Module Optical Analyzer [13,14] was conceived to be integrated in a solar simulator for quality control into a production line. However, this procedure does not have sufficient resolution for micro-CPV scale. Thus, an alternative procedure was proposed to be applicable to micro-CPV modules and described in [15,16]. The method consists of taking a photograph of each solar cell magnified through its corresponding optics. Then, the set of acquired images is processed to determine the relative position between each cell and its optics in pixels, which are converted to values of length or angle by applying a transfer function previously determined. The method has been successfully validated previously for conventional CPV by measurements and ray tracing simulations of a single lens-receiver unit [17].

In this work the authors take a step forward and apply the method to an entire micro-CPV Insolight module [18] that is composed by 572 optics-cell units, each one consisting of a biconvex hexagonal lens and a semi-dome optical secondary element attached to a square solar cell. The quantification of misplacements between cell and optics is particularly challenging because the image formation of the receiver through the anidolic biconvex lenses is very poor, particularly for misaligned units. We also share a new procedure developed for the validation of the method that takes advantage of the integrated-tracker system that displaces the receivers’ backplane in X,Y,Z directions following a spherical trajectory with respect to the lenses plane. Two sets of measurements are carried out, one with optimum alignment and a second with the backplane strongly misplaced, determining the statistical distribution of residuals for the uncertainty quantification.

The paper is organized as follows: first, the application of the method to micro-CPV is explained, using the Insolight module as a particular case study. Second, the misalignments among the 572 units are determined for two positions of the internal tracking. Finally, as a result of the comparison of the two sets of measurements, the uncertainty of the method is obtained for the case example and the misalignments distributions impact on the IV curve of the module is discussed.

2. Methodology for misalignment characterization in micro-CPV modules

The first generation of Insolight module (Gen0), composed of 572 lens-receiver units (see Fig. 1), is used for the experimental validation of this study. The module integrates an internal tracking system composed by three actuators: two of them displace the backplane in XY direction (i.e., coplanar to the lenses) with maximum travel of ±12mm, and the third one performs an angular XY rotation. The movement in Z direction is controlled by “guiding elements”, which force the backplane to follow a spherical trajectory that is a function of the X and Y coordinates. The focal distance of the lens is 12.7mm, and therefore the relation between the AOI and the receiver displacement (performed by the integrated tracking) is about 5°/mm.

 

Fig. 1. Gen0 Insolight modules installed onto the rooftop of the Institute of Solar Energy in Madrid.

Download Full Size | PDF

The array of lenses is composed by 64 pieces molded in PMMA and glued to a glass substrate. Each piece has 9 biconvex lenses shaped as regular hexagons of 8.34mm side, arranged in a 3 × 3 honeycomb structure. The receivers are formed by a 1mm side square multi-junction solar cell with a 1mm radius glass semi-dome attached to it. The array of lenses and the backplane of receivers are manufactured separately, misalignments among units will rely only on mechanical tolerances.

The manufacturing process was not fully automated for Gen0, so the first batches of modules suffered from significant misalignments that were reduced in subsequent generations. The result is that the hundreds of optics-cell units composing the module do not have their optical axis perfectly aligned due to the mechanical tolerances in the manufacturing process. The misalignment in an optics-cell unit between its elements depends on the relative position between the center of a cell and its lens (Δρ). A spatial misplacement between cell and optics is equivalent to a shift in the angular transmission function equal to the misalignment Δφ (see Fig. 2); their relation is explained in [17].

 

Fig. 2. Simulated angular transmission function of a single optics-cell unit of Insolight Gen0 for different incident angles (0°, 20°, 40° and 55°). Dotted line represents cosine response of the incident angle, and solid line the output power of the module at different angles of incidence. The acceptance angles AA for 90% power transmission degrades with the tracking position, particularly at large Sun incidence angles.

Download Full Size | PDF

The angular acceptance (AA_90%), defined as the angle at which the angular transmission decreases to X % of its maximum (where typically X = 90%), is a valuable figure of merit for CPV systems, because it describes the allowed tolerance for the tracking system and the stiffness requirement of the installation. In this regard, it is necessary to keep the AA as large as possible at the module level, ideally as the AA of a single optics-cell unit. Misalignments among units degrade the acceptance angle of the whole [19] and IV curve fill factor. In this regard power losses in the range of 10% are unacceptable, so misalignments should be kept well below AA_90%, for instance to AA_98% (accounting only for 2% power losses) [20].

The Insolight optical system provides a wide acceptance angle (AA_90%= ±1.7°) at normal incidence of the light, and keeps the light focused at a wide range of incident angles. The internal tracking locates the receivers at the optimum position, under the concentrated light spot, at each sun angle of incidence. Figure 2 shows the angular transmission function of a single optics-cell unit, simulated by ray tracing, considering:

The measurement set up for the misalignments characterization is composed by a motorized XYZ stage where a camera with a macro lens is mounted, details can be found in [21]. The camera is moved in XY directions (i.e., parallel to the lenses plane) to be placed in front of the center of each of the 572 optics-cell units; then a picture of the CPV lens and another of the cell magnified through the CPV lens are taken. Both cell and lens centers are determined through image processing, being the difference between centers equal to the spatial misalignment Δρ in pixels (see Fig. 3.c). This spatial misalignment (Δρ) can be translated into an angular value through a proportional pixels-to-degrees constant. This constant (°/px) may be calculated through simulations or measurements, assigning an angular size to the cell size [17].

 

Fig. 3. a) Image in EL of the magnified cell (and the result of its processing through the square recognition and center detection); b) Image in EL of the lens (and the result of its processing through the hexagon recognition and center detection; c) Definition of the spatial misalignment Δρ as the difference between the lens and cell centers.

Download Full Size | PDF

The following conditions are relevant for a correct image acquisition and successful determination of the misalignments. First, the macro lens used along with the camera magnifies both lens and cell in the image but narrows the depth of field. Thus, it is not possible to focus simultaneously both and identify their positions in the same photograph with the needed accuracy. Therefore, two pictures per lens-cell unit are taken varying only the camera focus to observe sharply the cell and the lens edges (Fig. 3.a and b respectively). Second, the anidolic design of the CPV lens of Insolight blurs the cell edges and deforms cell shape differently depending on the misalignment of the unit. So, current is injected into the cells to produce electroluminescence (EL) and enhance image contrast and ease the edge detection of the different elements.

A proper image processing is essential to minimize the uncertainty in the determination of the position of lenses and cells [21]. For the example case, a cell misplacement of 200 µm is equivalent to 1° misalignment in the angular transmission function and corresponds to hardly 100 pixels in the captured high-definition images (4904 × 3280px), where the cell has a size of nearly 800px. The processing aims, first, to enhance the hexagonal shape of the lens aperture and the square shape of the cell in the photographs (see Fig. 3), second, to determine their positions, and finally, to calculate the distance between both objects. The identification of the cell position requires different approaches depending on the observed deformation, which is in turn proportional to the cell displacement with respect to the center of its lens. Figure 4 presents examples of acquired images and its corresponding processing to show how the cell image segmentation (pink areas discerned from the background) and squared-shape features detection (i.e., corners and edges marked as dots and dashed lines) works. The center of the cells positioned within tolerance, i.e., misalignments limited to 1°, is perfectly determined (4 corners, 4 sides), but for those particular units with very large misalignments (i.e., higher than 2°), only 2 sides are identified (Fig. 4.b) which leads to a higher uncertainty in the cell position. In a few other cases, the automated process was able to identify only one edge (Fig. 4.c), so cell position could not be determined. The numerical value of the misalignment for these last cases is not relevant as the cell position is far beyond tolerance (1.7°), but, for the sake of completeness, the cell position has been manually detected in these cases.

 

Fig. 4. Upper row: original images of cells magnified through the CPV lenses. Lower row: processed images. a) Cell image correctly processed, where the cell area (in pink) is properly segmented, the 4 corners and the 4 edges are well detected. b) Cell image processing in which the segmentation does not obtain straight edges and 2 corners are missing. The center is calculated considering the two detected corners and their projected distance. c) Fail in the cell image processing as the cell position is far beyond tolerance limits.

Download Full Size | PDF

3. Validation of the method through repeatability study–uncertainty associated to camera position

For this measurement, the accuracy of the XYZ stage to position the camera centered with respect to the micro-CPV lens does not have a direct impact on the uncertainty of the method, since the cell position is determined with respect to the lens position (based on two different pictures taken with the same camera position). However, the image of the cell through the CPV lens is distorted depending on the camera position with respect to the center of the lens. Thus, it is desirable to take the picture with the camera as centered as possible with respect to the unit, avoiding important deformations of the cell shape. In order to account for this effect, an experiment has been performed in which the camera has been displaced in front of a single lens-receiver unit with steps of 0.25mm and a travel of ±0.75mm in X and Y axes. The standard deviation of the residuals (differences between the misalignment value for each camera position with respect to the one related to the camera centered) is 0.007° (see Fig. 5) which provides an idea of the corresponding uncertainty (i.e., the method repeatability).

 

Fig. 5. Scatter plot of the residuals of the misalignments obtained placing the camera at 49 positions (scan range of ± 0.75mm with steps of 0.25mm). The standard deviation (circle) of the distribution is 0.007°.

Download Full Size | PDF

4. Validation of the method through measurements of misalignments at 2 different integrated tracking positions

The main source of uncertainty of the method is determining the center of the cell by the processing above described, particularly when the image of the cell is strongly deformed. In order to account for this effect and to validate the proposed method, two set of measurements have been carried out as follows:

To show the impact of the misalignments distribution in the electrical performance, the Insolight module has been electrically characterized in the solar simulator under collimated light [22,23]. The IV curve for normal illumination and optimum placement of the integrated tracking has been measured. In this case, the optimum position for the integrated tracking is determined as the maximum power output obtained through an iterative scan in XY directions of the module backplane. Moreover, the IV curve of the module has also been measured by modifying the integrated tracking 200 µm in Y axis with respect to the optimum position.

5. Results and discussion

Figure 6 shows the X and Y spatial misalignments for both sets of experiments, only 1 quarter of the module is represented for the sake of comprehension (143 units), the other 3 exhibit similar results. The results are represented grouped by colors corresponding to pieces of 3 × 3 units of the lenses array (as shown in Fig. 6.a). The differences between lens and cell centers detected in the first set of measurements (optimum tracking position) are represented with filled circles, while for the case of the second set (200 µm displacement in Y direction) are represented as unfilled squares, both expressed in pixels. As it can be observed, the spatial misalignments (see Fig. 3.c) practically coincide for the X axis (Fig. 6.b up) for both sets, whereas a positive offset is observed in Y axis for the second measurement (Fig. 6.b low).

 

Fig. 6. a) Layout of the lens array pieces for one module quarter. b) X(upper) and Y(lower) spatial misalignment detected in both measurements for the first module quarter.

Download Full Size | PDF

As stated, a spatial misalignment between a single cell and its optics is equivalent to an angular shift in the angular transmission function of that unit with respect to the others, degrading the overall performance of the module, more details can be found in [1,9,14 and 19].

Figure 7 shows the misalignment values of the whole module, in angular units, represented as a scatter plot, Fig. 7.b for the first set (optimum tracking position) and Fig. 7.c for the second (200 µm displacement in the Y axis from optimum tracking position, corresponding to 1°). In both graphs the circles indicate the AA_90% and AA_98% of a single optics-cell unit. The units are colored accordingly to the 4 quarters of the module, as detailed in Fig. 7.a. The applied image processing performs automatic corrected segmentation (as shown in Fig. 4) to detect the lens and cell positions for most but not all units. Those that are manually processed are represented as unfilled markers in Fig. 7.

 

Fig. 7. a) Module is divided in 4 quarters (one color for each) with 572 lens-cell units in total b) Measured misalignments for a fixed optimum tracking position. c) Measured misalignments with introduced Y offset of a 200 µm (equivalent to 1°).

Download Full Size | PDF

To evaluate the uncertainty of the method, the difference between misalignments for the 572 lens-cell units, in both set of measurements, is analyzed. Firstly, the 1° offset in Y of the second set is subtracted and then, the difference of each pair of values for each unit misalignment from the two set of measurements is calculated. The result is a set of 572 deviations in X and Y axes, which are fitted to Gaussian distributions (Fig. 8.a). Both X and Y distributions have a standard deviation of 0.1° (which corresponds to a cell displacement of 20 µm) when all units are considered, and 0.08° when only the units with automatic segmentation are included. These values represent the uncertainty associated to the method, and they should be considered the worst case values, as a large displacement, such as 200 µm, causes cell photographs to be worse segmented due to deformations in the magnified cell, which increases the error in the determination of the cell position.

 

Fig. 8. a) Best fit Gaussian distribution of the differences between misalignments values determined with the two sets of measurements, taking into account all units (dotted line) and only the automatically processed units (solid line). b) IV curves measured for the integrated tracking position that maximizes the power output (optimum) and with the integrated tracking displaced 200µm (equivalent to 1°) in Y direction with respect to that optimum position.

Download Full Size | PDF

An important consideration must be done regarding the alignment between the XY axes of the module integrated tracking system and the XY axes of the camera moving system. A perfect alignment between both Cartesian XY systems is not necessary for the correct misalignment characterization, but it has a relevant impact on the presented experiment. The two sets of measurements have a displacement of 200 µm in the Y direction (accordingly to the XY axes of the internal tracker) and should be detected as a displacement only in the Y direction by the processed images. But if there is a slight rotation among both Cartesian XY systems, the images would detect it in the differences among both set of measurements. For the experiment, the XY system of the camera movement has been aligned with the module frame, which could be different than the XY system of the internal tracker. This rotation in both Cartesian systems emerges in the statistical distributions of the differences of measured values in both sets, which does not have zero mean, but is positive in Y (0.03°) and negative in X (-0.02°), as shown in Fig. 8.a.

The misalignment values with the internal tracking aligned (Fig. 7.b) are mostly within the acceptance angle AA_90%, but a significant number of units exceed AA_98%. The impact of the misalignments in the IV curve (Fig. 8.b, blue trace) is a significant slope (current ranges from 0.84A to 0.79A) and a power loss of 5.5%. It must be pointed out that the module has a first level of electrical connection of cells in parallel that reduces the impact of misalignments. If all cells were connected in series, power losses exceeding 10% would be expected accordingly to Fig. 7.b.

For the case of the backplane misplaced 200 µm (or 1°), there is a high number of cells with a misalignment above the AA_90% (Fig. 7.c), which leads to a higher slope in the IV curve (Fig. 8.b, red trace) so current drops to 0.73 A which corresponds to a power loss of 13%. It is noteworthy that, despite the acceptance angle AA_90% of the optical system is 1.7°, a displacement of the tracker as low as 1° causes a power degradation exceeding 10%, which means that there is a degradation of the AA_90% of the module compared to a single optical unit. It would be desirable to delimit the maximum misalignments to values below AA_98% to keep the high acceptance angle of the optics-cell unit at the module level. And what is more important for this technology, as the AA_90% decreases with the angle of incidence (see Fig. 2), the power losses associated to misaligned cells will increase for higher sun incident angles, which will in turn lead to a reduction in the energy yield of the module.

6. Conclusions

A method based on image processing for misalignments characterization of optics-cell units, which was conceived for conventional CPV, has been successfully applied to a micro-CPV module based on Insolight technology. The measurement set-up has been implemented based on a XYZ stage and a camera with a macro lens. Two images are captured, one focusing the CPV lens aperture and a second focusing the cells magnified by its CPV lens. The image processing for a correct automated detection of the lens and cell position has been developed.

The uncertainty of the method has been evaluated with an experiment where the misalignments of 572 optics-cell units where determined, taking advantage of the integrated tracking of the module. In a first set of measurements, the internal tracking was placed at the optimum position, in a second it was displaced 200 µm in the Y axis. The differences in the misalignment values obtained for both sets were evaluated, providing a standard deviation of 0.08° (less than 20 µm in the cell position), being an acceptable value to validate the method for quality control in the manufacturing of micro-CPV modules.

Finally, the IV curves for the two cases, cells at optimum position and with offset, were measured and compared to the misalignments distributions, showing the impact of the misalignments on the fill factor of the IV curve and the degradation of the angular tolerance of the module compared to the acceptance angle of a single optics-cell unit.