Available online at www.sciencedirect.com

ARTICLE IN PRESS

www.elsevier.com/locate/solener

Solar Energy xxx (2010) xxx–xxx

Automated high resolution measurement of heliostat slope errors

Steffen Ulmer a,*, Tobias Marz a, Christoph Prahl a, Wolfgang Reinalter a, Boris Belhomme b

a German Aerospace Center (DLR), Institute of Technical Thermodynamics, Solar Research, Plataforma Solar de Almerıa (PSA), 04200 Tabernas, Spainb German Aerospace Center (DLR), Institute of Technical Thermodynamics, Solar Research, 51170 Cologne, Germany

Received 26 November 2009; accepted 6 January 2010

Communicated by: Associate Editor Gregory Kolb

Abstract

A new optical measurement method that simplifies and optimizes the mounting and canting of heliostats and helps to assure theiroptical quality before commissioning of the solar field was developed. This method is based on the reflection of regular patterns inthe mirror surface and their distortions due to mirror surface errors. The measurement has a resolution of about 1 million points perheliostat with a measurement uncertainty of less than 0.2 mrad and a measurement time of about 1 min per heliostat. The system is com-pletely automated and allows the automatic measurement of an entire heliostat field during one night. It was extensively tested at theCESA-1 heliostat field at the Plataforma Solar de Almerıa. Comparisons of flux simulations based on the measurement results with realflux density measurements were performed. They showed an excellent agreement and demonstrated in a striking manner the high mea-surement accuracy and high grade of detail in the simulation achieved by this technique.� 2010 Elsevier Ltd. All rights reserved.

Keywords: Heliostat; Optical quality; Optical measurement; Deflectometry; Ray tracing; Flux measurement

1. Introduction

The heliostat field is the major cost component of a solarthermal tower power plant and the optical quality of theheliostats has a significant impact on the field efficiencyand thus on the performance of the power plant. It is there-fore a permanent goal to decrease their manufacturing costbut at the same time maintain or even improve their opticalquality. To accomplish that, economical and accurate mea-surement systems are needed to detect optical deficienciesin facets and support structures, to allow precise and opti-mized canting without sun and to determine the overalloptical quality of single heliostats and entire fields foracceptance tests.

Since the early eighties canting and quality assessmentof heliostats was mainly done by observation of the focal

0038-092X/$ - see front matter � 2010 Elsevier Ltd. All rights reserved.

doi:10.1016/j.solener.2010.01.010

* Corresponding author. Tel.: +34 950 265320; fax: +34 950 260315.E-mail address: steffen.ulmer@dlr.de (S. Ulmer).

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

spot produced by the heliostat on a flat white surface closeto the receiver area. Specialized measurement systems withdigital cameras and specific evaluation algorithms weredeveloped to quantify and automate this task (Monterreal,1997a; Imenes et al., 2006). Although this method deliversa graphic and reliable result it has two main drawbacks: itdepends on clear-sky conditions and it only allows to deter-mine an overall quality parameter (beam quality), but it isnot possible to resolve and assign local surface errors.

For facet canting without sun a total station can beused. All four corners of each facet are measured with aprism and from these points in space the facet normaland its necessary correction is calculated. This method isused in recent years at the Plataforma Solar de Almerıaand in several collector assembly lines. However, thismethod requires considerable manual work and considersthe information of only four points per facet. Any localerrors at these points are translated to the orientation ofthe entire facet. Manufacturing errors of the facets them-

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/

2 S. Ulmer et al. / Solar Energy xxx (2010) xxx–xxx

ARTICLE IN PRESS

selves cannot be measured. The same applies to the cantingmethod with a rotating laser proposed by Monterreal(1997b).

For the measure of local shape deviation of solar con-centrators photogrammetry can be used (Pottler et al.,2005). However, as the measured surface has to beequipped with a large number of target points which areused as individual surface measurement points, the mea-surement is a time consuming task which is not appropriatefor the measurement of large numbers of mirrors. A furtherdrawback of this method is that the slopes of the reflectivesurface are calculated from coordinates of points in space.High-resolution measurements imply short distancesbetween the measured points and tiny but inevitable mea-surement deviations can lead to significant errors in thelocal slope calculated. To overcome this dilemma severalmethods that directly measure surface slopes instead of sur-face coordinates were developed:

CIEMAT proposed to use the reflection of a star in themirror at night and move the heliostat to get enough infor-mation to calculate the local surface normals at variouspoints on the reflector (Arqueros et al., 2003). Again, thismethod depends on clear-sky conditions, although at night,and its use is a time consuming task, especially when highspatial resolution is desired.

The so-called V-shot measurement system developed bySandia and NREL measures the local slopes of a mirror byscanning it with a laser beam, detecting the point of inci-dence of the reflected beam and calculating the resultingsurface normal (Jones et al., 1996). So far this systemwas only applied for dishes and parabolic troughs whereit is sufficiently accurate. However, it might be difficult to

Fig. 1. Measurement set-up used for h

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

adapt it to heliostat measurements due to the large dis-tances and the necessary extremely high pointing precisionof the laser that comes along with it. Also, scanning at highresolution is time consuming.

A measurement principle called deflectometry or fringereflection has been coming up in the last decade in variousapplications where specular surfaces need to be qualifiedsuch as windscreens, glossy automobile parts, correctivelenses for glasses and so on (Kammel, 2003; Knaueret al., 2004). This method uses known regular stripe pat-terns whose reflection in the mirror is observed by a cam-era. The deformations of the stripe pattern in thereflection are then used to evaluate the local slopes of themirror. This method was first introduced in the qualifica-tion of concentrating solar collectors by the German Aero-space Center (DLR) (Ulmer et al., 2006, 2007) and is morerecently also used by other groups, although in the qualifi-cation of other concentrators than heliostats (Heimsathet al., 2008; Andraka et al., 2009).

Based on this basic principle, DLR has developed a newmeasurement system for the special case of heliostats. Thedeclared aim was to develop a system that allows measur-ing surface slopes with high resolution and with high accu-racy, and is suitable for large surfaces, rapid and easy toset-up.

2. Measurement

2.1. Measurement system

From daily observations it is known that smallest irreg-ularities in a mirror surface cause the reflected image to dis-

eliostat measurements in the field.

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/

Fig. 2. Example of a regular horizontal stripe pattern reflected in aheliostat.

S. Ulmer et al. / Solar Energy xxx (2010) xxx–xxx 3

ARTICLE IN PRESS

tort considerably. Hence, an observer that studies thereflection of a regular pattern can detect irregularities inthe slope of a specular surface that would often be imper-ceptible on a dull surface. This principle can be used toaccurately measure local slopes for mirror surfaces.

The chosen set-up for measuring heliostats in a solartower power plant is shown in Fig. 1. A projector situatedin the heliostat field projects the stripe patterns on a whitetarget surface at the tower at night. A digital cameralocated on top of the tower takes images of the heliostatswhich are oriented the way that the stripe patterns can beseen in the reflection. If the camera position, the heliostatposition and the target position are known and if it isachieved to clearly identify each point M(i, j) of the dis-torted pattern seen in the heliostat and assign its corre-sponding position D(i, j) on the target, the local normalvector can be calculated from the law of reflection:

n ¼ r � ijr � ij

With the normalized vectors i for the incident ray, r forthe reflected ray and n for the surface normal.

The unambiguous identification of the reflected targetpoints is achieved by codifying the target surface with a ser-ies of stripe patterns. Vertical stripes are used to codify thetarget in x-direction and horizontal stripes are used to codeit in y-direction. Stripe patterns with sinusoidal brightnessvariations are used to codify the position in each direction.Fig. 2 shows an example picture of a projected horizontalstripe pattern on the target surface (left) and the corre-sponding pattern as it is seen from the camera in the reflec-tion of the heliostat.

From the brightness (grey level) seen at a certain point,the corresponding phase of the sinus wave can be assigned.However, this is ambiguous for phase angles larger than±p/2 as the same grey values appear repeatedly. Withinthe period of 0–2p, this can be resolved by using four pat-terns with a phase shift of p/2 between each.

phase ði; jÞ ¼ atanGV 4ði; jÞ � GV 2ði; jÞGV 1ði; jÞ � GV 3ði; jÞ

� �

This also increases the accuracy of the determined phaseand makes it independent of local brightness alterations.The ambiguity between multiples of 2p can be solved byusing additional patterns with longer wavelengths (Fig. 3).The phase information of the longer wavelength is used toidentify the phase interval of the shorter wavelength pattern.The accuracy in the determination of the phases in the longwavelength series must be higher than the phase interval ofthe shorter wavelength series to be identified. Additionalstripe pattern series with shorter wavelengths can be usedto increase final precision. This method of codification iscalled hierarchical phase shift (HPS).

A computer algorithm developed in Matlab� evaluatesthe images, calculates the surface normal vectors and visu-alizes the results. It takes into account and corrects allknown systematic error influences such as distorted projec-

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

tion of stripe patterns due to oblique projection angle andprojector lens distortions, non-linearity of projector andcamera, background lighting, inhomogeneous reflectionproperties of mirror and target, perspective rectification,lens aberrations, etc.

3. Results

(Fig. 4 shows an example measurement result of a helio-stat at the Plataforma Solar de Almerıa (PSA). The result isgiven as deviation of local slope angle from ideal slopeangle in milliradians (mrad) in x-direction (azimuth) inthe left graph and in y-direction (elevation) in the rightgraph. The deviations are in the range of ±5 mrad. Positivedeviations in x/y-direction correspond to tilt of the normalvector towards the negative x/y-direction. Certain system-atic deviations can be observed: in the facets generallythe x-deviation is positive in the left half and negative in

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/

Fig. 4. Measurement result, displayed as slope deviations from idealsurface in azimuth (top) and elevation (bottom) in milliradians.

0

127.5

255

0pi 2pi 4pi 6pi 8pi 10pi 12pi

Phase

Gre

y le

vel

long wavelength patternshort wavelength patternphase angles from short wavelengthphase angle from long wavelengthidentified phase angle

Fig. 3. Decoding of target position coded with sinusoidal stripe patterns with different wavelengths.

4 S. Ulmer et al. / Solar Energy xxx (2010) xxx–xxx

ARTICLE IN PRESS

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

the right half. This corresponds to a longer focal length in x

than the ideal one. The same behavior can be observed forthe entire heliostat. In y-direction, especially within theright half, there are some facets with positive or negativedeviations in the entire facet area which corresponds todeviations in the facet alignment. This misalignment couldbe corrected by re-canting the facets. There are also somelocal deviations visible (most visibly in the facet in theupper left corner), probably due to stresses from mirrormounting and bending ).

The high resolution data are further processed forinterpretation. The standard deviation of the slope devia-tion within an individual facet corresponds to the defor-mation or shape deviation of the facet itself. The meandeviation of the slope deviation within an individual facetcorresponds to the deviation in facet alignment. The mea-sured focal lengths of the facets are determined by a par-abolic fit to the data within one facet. Table 1 shows anexample of these values for the measurement shownabove. These values can be used for adjustments of facetcanting and for control of facet focal lengths. Heliostatfacets are often made from flat glass that is slightly bentby a center screw that draws the facet back. For facetsthat allow adjustment of the focal length by such a centerscrew, the corresponding center displacement for adjust-ment can be calculated.

The same values can be determined for the entire helio-stat. The shown heliostat has a standard slope deviation of1.35 mrad in x-direction and 1.44 mrad in y-direction,which corresponds to a total RMS-value of 1.98 mrad.The measured focal length of the heliostat is 116.2 m,design focal length is 104.1 m.

3.1. Uncertainty analysis

The deflectometric measurement principle has theadvantage that the reflexion information is used as direct

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/

Table 1Measurement results of individual facets derived from high resolutiondata.

Row/col 1 2 3 4

Tilt in x-direction (mrad)

1 1.96 1.32 �0.46 �1.112 0.46 �0.18 �0.16 �1.273 1.32 0.68 �0.17 �1.154 1.13 0.15 0.06 �0.545 1.23 0.01 �0.56 �1.556 0.57 �0.12 �0.04 �1.00

Tilt in y-direction (mrad)

1 �1.25 �1.72 0.32 1.252 1.33 �0.58 1.00 2.083 �0.29 �0.68 �1.88 1.134 1.02 �0.57 1.18 2.435 0.87 �0.40 �1.34 0.386 0.45 �0.83 �2.45 �1.06

Measured focal length (m)

1 116.8 141.5 145.0 126.72 242.0 97.8 131.5 155.83 152.2 135.3 138.6 137.74 189.0 135.1 142.0 134.85 158.4 143.9 145.5 146.46 143.2 129.0 133.3 143.6

Fig. 5. Sketch of the automatic measurement system with wireless

S. Ulmer et al. / Solar Energy xxx (2010) xxx–xxx 5

ARTICLE IN PRESS

input of the measurement which leads to high accuracies inthe measured surface slopes at high spatial resolution. Dueto the numerous transformations and corrections duringthe evaluation procedure the uncertainty analysis of themeasurement method is a complex task. To cope with thisa mathematical model that considers all known uncertain-ties and that completely reproduces the evaluation processwas developed. It allows the prediction of the expectederror for a certain measurement set-up, the calculation ofthe local error of performed measurements and is a valu-able tool for the system improvement as error budgetscan be analyzed in detail. At present, a resolution of about1000 � 1000 measurement points per heliostat with a localmeasurement error of below 0.2 mrad is reached.

communication (a) and camera system with motor zoom on pan/tilt headwith removed housing (b).

3.2. Automation

The measurement procedure and evaluation was com-pletely automated. An industrial type CCD camera fromLumenera� together with a computer controlled pan/tilthead and motor zoom lens allow completely automaticscans of entire heliostat fields. The projected stripe patternsare remotely controlled and the camera shutter triggeredaccordingly. The different components of the measurementsystem communicate over a wireless network, so no cablingexcept electric power is necessary (Fig. 5). The measure-ment time for one heliostat is about 1 min, so several hun-dred heliostats can be measured during one night.

The evaluation process is independent from the imageacquisition process. A configuration file with all necessaryevaluation parameters has to be created by the user. Somestart values for automatic edge detection and thresholdshave to be determined which is usually done during the

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

evaluation of the first heliostat. The rest of the imagescan be processed in batch mode without further user inter-action. The evaluation time for each heliostat depends onthe selected resolution and the options chosen, but is usu-ally in the range of about 1 min. That means that the mea-surements taken during one night can be evaluated in1 day. Compared to existing measurement methodsdescribed in the introduction this method offers significantgain in speed and handling.

One limitation of the described measurement method isthat the facets need to be sufficiently pre-aligned to guaran-tee that the reflection of the stripe patterns on the targetcan be seen in the entire heliostat. This is not case if errorsexceed a certain limit. Yet, as described in Roger et al.(2008) these cases can be covered by a different opticalmeasurement method that uses the same hardware andset-up, although with less accuracy.

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/

Fig. 6. Screenshot of ray tracing program STRAL with heliostat set-up used for validation.

6 S. Ulmer et al. / Solar Energy xxx (2010) xxx–xxx

ARTICLE IN PRESS

4. Ray tracing simulation and validation

In recent years DLR also developed a new ray tracingprogram named STRAL (Solar Tower Raytracing Labora-tory (Belhomme et al., 2009) that efficiently processes largenumbers of high resolution data sets obtained from abovedescribed heliostat measurements. It allows fast and accu-rate simulations of the flux distributions delivered by theheliostats and can be used for assessment of the measure-ment results and for optimization of heliostat field opera-tion using dynamic aimpoint strategies.

Measurements of flux distributions of real heliostats atPSA were used to validate the deflectometry measurementaccuracy and the ray tracing model by comparison. Inorder to take into account as many influences as possible,a group of four heliostats partly blocked and shaded bythe heliostats in front was chosen (Fig. 6).

A flux density measurement with the four heliostats inthe back row pointing to four different points on the flatmeasurement target at the tower and with the heliostatsof the front row in the shown positions was performed.Also, deflectometry measurements of all four heliostats

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

were done and the corresponding data files for STRALgenerated. Fig. 7 shows the measured flux distribution ofthe four heliostats (bold lines) and the ray tracing result(thin lines) in an overlay. The agreement between both dis-tributions is very good which means that local surface devi-ations, as well as shading and blocking are correctlyreflected in measurement and simulation.

This type of highly accurate ray tracing simulationsbased on high resolution measurement data allow muchmore realistic predictions and optimizations of the heliostatfield performance than in earlier simulation models whereshape deviations of heliostats were approximated with asingle value for a Gaussian slope error distribution (Garciaet al., 2007).

5. Summary and outlook

The presented optical measurement system allows eco-nomic and automatic measurements of concentrator slopedeviations of entire heliostat fields with high resolutionand accuracy. This can be used to check and optimize con-centrator quality, canting of facets and the optical quality

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/

Fig. 7. Measured flux distribution of the four heliostats (bold lines), and simulated flux distribution of four heliostats based on measured deflectometrydata (thin lines).

S. Ulmer et al. / Solar Energy xxx (2010) xxx–xxx 7

ARTICLE IN PRESS

of entire heliostat fields. Combined with the ray tracingcode STRAL it is possible to accurately predict the helio-stat field performance and improve heliostat field operationusing dynamic aimpoint strategies. The described measure-ment method can also be used to measure other solar con-centrators such as dishes, troughs and individual facets.

Acknowledgements

Financial support from the German ministry for theenvironment, nature conservation and nuclear safety with-in the contract 03UM0068 is gratefully acknowledged.

References

Andraka, C., Sadlon, S., Myer, B., Trapeznikov, K., Liebner, C., 2009.Rapid reflective facet characterization using fringe reflection tech-niques. In: Proceedings of the Energy Sustainability 2009, SanFrancisco, California, USA, July 19–23, 2009.

Arqueros, F., Jimenez, A., Valverde, A., 2003. A novel procedure for theoptical characterization of solar concentrators. Solar Energy 75, 135–142.

Belhomme, B., Pitz-Paal, R., Schwarzbozl, P., Ulmer, S., 2009. A new fastray tracing tool for high-precision simulation of heliostat fields. J.Solar Energy Eng. 131 (3), 2.

Garcia, P., Ferriere, A., Bezian, J.-J., 2007. Codes for solar fluxcalculation dedicated to central receiver system application: a com-parative review. Solar Energy 82 (3), 189–197.

Heimsath, A., Platzer, W., Bothe, T., Wansong, L., 2008. Characterizationof optical components for linear Fresnel collectors by fringe reflectionmethod. In: Proceedings of Solar Paces Conference, Las Vegas, NV,2008.

Imenes, A., Stein, W., Hinkley, J., Benito, R., Bolling, R., Schramek, P.,Ulmer, S., 2006. Ray Tracing and Flux Mapping as a Design and

Please cite this article in press as: Ulmer, S. et al. Automated high resolutionj.solener.2010.01.010

Research Tool at the National Solar Energy Centre. Paper numberS06-81, ANZSES 2006 Conference (Australia and New Zealand SolarEnergy Society), Canberra, Australia, September 13–15, 2006.

Jones, S.A., Neal, D.R., Gruetzner, J.K., Houser, R.M., Edgar, R.M.,Kent, J., Wendelin, T.J., 1996. VSHOT: a tool for characterizing large,imprecise reflectors. In: International Symposium on Optical ScienceEngineering and Instrumentation, 1996, Denver, CO.

Kammel, S., 2003. Deflectometry for quality control of specular surfaces.Technisches Messen tm 70, April 2003, pp. 193–198.

Knauer, M.C., Kaminski, J., Hausler, G., 2004. Phase measuringdeflectometry: a new approach to measure specular free-form surfaces.Optical Metrology in Production Engineering. Proc. SPIE 5457.

Monterreal, R., 1997. Optical comparison ASM 150 and GM 100A. In:Proceedings of the 10th Task III Meeting within the IEA SolarPACESon “Solar Technology and Applications, SolarPACES TechnicalReport III – 6/97”, Sandia, Albuquerque, 15.9.1997.

Monterreal, R., 1997. A new on-axis canting method for facetedheliostats, solar thermal concentrating technologies. In: Proceedingsof the Eighth International Symposium, October 6–11, 1996, Koln,Germany, Deutsche Forschungsanstalt fur Luft- und Raumfahrte.V. Manfred Becker/Manfred Bohmer (Hrsg.), Muller, Heidelberg,1997.

Pottler, K., Lupfert, E., Johnston, G., Shortis, M., 2005. Photogramme-try: a powerful tool for geometric analysis of solar concentrators andtheir components. J. Solar Energy Eng. 127 (1), 94–101.

Roger, M., Prahl, C., Ulmer, S., 2008. Fast determination of heliostatshape and orientation by edge detection and photogrammetry. In:Proceedings of the 14th Biennial CSP SolarPACES Symposium, LasVegas, NV, March 4–7, 2008.

Ulmer, S., Heller, P., Reinalter, W., 2006. Slope measurements ofparabolic dish concentrators using color codified targets. In: Proceed-ings of the 13th SolarPACES International Symposium on Concen-trated Solar Power and Chemical Energy Technologies, Seville, Spain,June 20–23, 2006, ISBN: 84-7834-519-1.

Ulmer, S., Roger, M., 2007. Automatisierte hochaufgeloste Vermessungder Spiegelfehler von Heliostaten. 10. Kolner Sonnenkolloquium “F�Emit hochkonzentrierter Strahlung”, 21. Juni 2007, DLR, Koln-Porz.

measurement of heliostat slope errors. Sol. Energy (2010), doi:10.1016/