(S)

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Energy Procedía 69 (2015) 50 - 59

International Conference on Concentrating Solar Power and Chemical Energy Systems,

SolarPACES 2014

Experimental validation of theoretical heliostat wind loads

M. Burischa*, I. Santanaa, M. Sancheza, C. Villasanteb, E. Olabarrietab

aCENER, C/Ciudad de la Innovación 7, 31621 Sarriguren (Navarra), Spain b IK4-TEKNIKER, C/ Iñaki Goenaga 5, 20600 Eibar (Gipuzkoa), Spain

Abstract

In the current concentrated solar power (CSP) context central receiver systems seems to be the most promising technology with the highest cost reduction potential. In the construction of such plants the solar field represents one of the largest investments, a fact which has made cost reduction of these elements become the main target for many actors of the solar thermal sector. Therefore, the design of cost efficient heliostats has become a major interest. One of the design challenges is to develop a heliostat which is able to withstand all foreseen mechanical loads, while at the same time have reduce costs by avoiding unnecessary safety margins. To achieve this goal a good knowledge of expected wind loads onto the support structure and drive mechanism is crucial.

In this paper a heliostat prototype, equipped with measuring devices, is deployed to a test site in order to measure the true mechanical stresses suffered due to varying wind loads. These data together with the information about wind speed and direction is then used to validate theoretical loads determined via formulations extracted from wind tunnel experiments. It is shown, that for low turbulence conditions the theoretical values correspond well to the measured values; however, for turbulent wind conditions a significant difference is detected. As those conditions are difficult to simulate in wind tunnels or CFD simulations, this shows the importance of on-site testing of new heliostat designs.

©2015The Authors.Publishedby ElsevierLtd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer review by the scientific conference committee of SolarPACES 2014 under responsibility of PSE AG Keywords: heliostat; wind load; turbulence; aerodynamic coefficients; load measurement

* Corresponding author. Tel.: +34 948 25 28 00; fax: +34 948 27 07 74. E-mail address: mburisch@cener.com

1876-6102 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer review by the scientific conference committee of SolarPACES 2014 under responsibility of PSE AG doi: 10.1016/j .egypro .2015.03.007

1. Introduction

Power towers are a cost-effective Concentrating Solar Power (CSP) technology for producing solar-generated electricity on a large scale [1]. However, investment costs for CSP systems are still high. One possible target for costs reduction is the solar field, which involves around 50% of the total capital costs [2]. That means heliostats are the most important cost element of a power tower plant, and this fact has made cost reduction of these elements become the main target for many actors of the solar thermal sector.

The design of heliostats aims at achieving the lowest possible costs for construction, operation and maintenance over the lifetime of the plant. The most important parts of such a design are the drive mechanism, structural support, reflecting surface and the foundation. Selecting appropriate parts and dimensioning them to the actual needs, allows for significant costs savings opposed to conservational designs.

Among other factors, the dimensions and, therefore, the amount of material used for the structure and support of the heliostat are defined by the loads to which it is exposed. The main sources of mechanical load are gravity and wind load, the latter being a crucial factor due to its uncontrollable nature. The somewhat random behavior with respect to direction, speed and especially turbulences make a prediction about the mechanical loads onto the heliostat structure difficult.

A common approach to calculate the loads on a heliostat is by numerical simulations using CFD and FEM codes [3, 4] or wind tunnel experiments [5], where in general the pressure onto the facets are calculated respectively measured. For wind tunnel experiments usually scaled down heliostats are used to be able to use smaller wind tunnels or test multiple heliostats at once. In order to achieve comparable results scaling has to be done appropriately, which poses difficulties for the drive mechanism as certain parts cannot be scaled easily. However, both numerical and wind tunnel simulations are valuable tools to calculate the mechanical stress on the structure and optimize the design of the heliostat and solar field for instance by adding wind fences [6], wind protection devices [7] or using different heliostat geometries [8].

However, these entire tests are only approximations and should be complemented with on real testing like for instance the one in [9] where the effect of load degradation at different rows of heliostats is analyzed. The importance of those tests is pointed out by [10] where they point out a general underestimation of the numerical solutions in comparison with measured field data. These differences can be attributed partially to the dynamic behavior of wind loads which is generally not reflected by the steady state numerical simulations or wind tunnel experiments. In [11] a method is described to try to account for those variations and achieve a more precise prediction. In [12] all three methods (numerical simulation, wind tunnel and on field monitoring) are compared also pointing out the significant impact of the turbulent nature of wind.

Zang et al. [13] estimate the dynamic response of a heliostat to wind loads by correlating deformations measured by accelerometers and strain gauges with wind data. Subsequently the frequency response of the deformations is compared with a numerical simulation showing a good match between both methods.

Having an appropriate heliostat foundation and support structure to withstand the suffered mechanical loads is important to not only avoid permanent damage of the heliostat but also to maintain the solar field efficiency. In [14, 15] forces and torques in the drive system caused by the wind loads are calculated and the resulting deviations of the tracking system analyzed. Their results show that an insufficient design of the drive mechanism can lead to a significant performance drop in the yield of the solar field, thus, pointing out the importance of a proper dimensioning of the crucial components.

In [16] an analytical method for calculating the wind loads is derived from a series of wind tunnel experiments using scaled down heliostats. These correlations allow for a quick estimation of the expected forces and torques the heliostat support structure and foundation have to withstand and to make an initial design of the heliostat. In the design stage, loads can be predicted with aero elastic models and codes. However, such models have their shortcomings and uncertainties, and they ought to be validated by measurement.

In the following sections we will present an experimental validation of the theoretical values by measuring the actual wind loads on a small sized heliostat prototype and comparing them with the calculated theoretical loads. Assuring that theoretically calculated wind loads are reasonable approximations of the actual loads will help in designing heliostats and reduce complex and expensive wind tunnel tests or CFD simulations to a minimum.

2. Measurements

2.1. Prototype

Experimental field measurements are performed using a simplified heliostat prototype as shown in Figure 1a. The structure of the prototype is made out of F-1140 steel. The heliostat itself is attached to the ground via four ground screws, preventing all translational or rotational movements of the base. The prototype is designed to facilitate the measurement of bending moments and torques with a high sensitivity level, thanks to a weakened structure at the measuring points (Figure 1b). The mirror consists of an acrylic sheet with a square surface of 1 m2 and a thickness of 5 mm. The sheet is attached to the heliostat using a triangular connection part allowing to measure at different elevation angles.

Fig. 1. (a) Measurement prototype, (b) Test piece with attachment points for strain gauges; (c) Mounted test piece.

The measuring equipment is placed near the attachment point of the mirror, allowing measuring the loads around the usual mounting point of the drive mechanism. As the drive mechanism is the most crucial part in terms of possible costs reductions, accurate load predictions around this area are of high interest to allow for an adequate dimensioning of those parts. The prototype is designed to allow measurements at different azimuth and elevation angles of the reflective area, providing a wide flexibility on the test configurations. For the following evaluation the elevation angle is fixed at 30 which is a very common angle during operation for typical plants. The azimuth angle is set to 20 ° with respect to north. This direction was chosen in accordance with the expected dominant wind direction at the test site.

2.2. Measurement equipment

For measurement purposes the prototype is equipped with a set of strain gauges attached to the heliostat stand to measure the forces and torques caused by the wind. The torque moment is measured at the tubular part of the measuring object and the bending moments are measured at the two perpendicular sections (see Figure 1b and c). Furthermore, a wind vane and an anemometer to measure wind speed and direction are installed. Both are mounted roughly at the same height as the heliostat surface, to give the most precise data. To account for the remaining height differences the measured values are corrected as explained further below.

2.3. Test site

The test site was selected due to its favorable wind conditions, allowing for a good range of wind speeds and predictable wind directions as shown in Figure 2.

Fig. 2. (a) Wind direction; (b) Wind speed at test site.

Topographical considerations of the site must be taken into account in order to avoid non-desired effects. In this case, the prototype has been set onto a completely flat sector. Moreover, there are no obstacles around and on the ground a thin layer of gravel has been deposited. However, in the north direction, the test prototype is preceded by a progressive slope (see Figure 3). This type of topographical change may cause a considerable increment of turbulence intensity by altering the boundary layer and making the effects of the wind to be different on the measurements. This effect could be seen in the following figure [17].

Fig. 3. Sketch of surface profile at test site.

Overall, in the case study the progression of the slope is continuous enough not to cause significant changes in turbulence intensity for moderate wind speeds. Although, for some specific atmospheric conditions, turbulence intensity could show a considerable increment which could lead to non-concluding measurements, as will be presented later.

2.4. Dynamic considerations

Too low natural frequencies of the heliostat system could cause dynamic effects due to wind load excitation and possibly distort the results obtained from the measurements. A dynamic force with a frequency near the structural mode frequencies lead to a resonance state increasing the amplifications of the vibrations and in consequence the deformation level on the strain gauges. Thus, it is essential to analyze the frequency content of the system.

The spectrum analysis of wind speeds shows the frequency content of suffered wind loads. Therefore, a Fourier analysis has been performed for the measured wind speeds at the prototype location (Figure 4).

WIND SPECTRUM

0.5 1 1.5

Frequency [Hz]

Fig. 4. Wind speed spectrum at test site.

As seen, frequency content of wind loads is quite low. At this stage a theoretical modal analysis of the structure

needs to be calculated, which is shown in the following scheme (Figure 5).

Fig. 5. Modal analysis of heliostat prototype used for evaluation of wind dynamics.

The lowest natural frequency is 5.36 Hz which is far from on-site wind frequency spectrum. So, for this case study in particular, dynamic effects caused by structural flexibility are not going to be taken into account.

3. Theoretical wind loads

First, the anemometer and heliostat centroid have been mounted at 1.62 m and 1.38 m respectively. Although the difference is small, it should be corrected for. In our case the power law correction method is used due to its simplicity. :

\Href)

Where V is wind speed at heliostat centroid, Vref is the wind speed measured by the wind vane, H heliostat centroid height and Href is anemometer mounting height.

Moreover, the theory of aerodynamic loads determines that the forces and moments caused by an air flow are given by the following equations:

F = Cf •-• p • A • V2

'f • 2

M = Cm • i • p • A • V2 • h

(2) (3)

Where Cf and Cm are the aerodynamic coefficients corresponding to forces and moments respectively, p is air density (1.25 kg/m3 is taken as reference value), V is the mean wind speed, A the area of reflection surface and h is a reference length known as heliostat chord. The reference value used for the last one is the 1 m.

The most challenging fact in this type of calculation is, without doubt, obtaining of the aerodynamic coefficients. There are several ways to obtain them, for instance, through elaborate CFD calculations or via innumerable highly controlled tests in wind tunnels. The experimental determination of these coefficients has been performed within a couple of studies. One of the most famous is the study conducted by Peterka et al. [16]. They have performed a large study on the aerodynamic coefficients for heliostats and Dish concentrators obtaining coefficients for all loads for different angles of elevation and azimuth. More recently, a team at Google [18] has made a similar wind tunnel. They have proven that the coefficients calculated by Peterka, follow the same trend but are slightly larger than the actual ones. Thus, the aerodynamic coefficients for the boundary layer extracted by Google will be used. These coefficients are given for a CFD coordinate system, i.e. are aligned with the fluid media and not with the heliostat. Hence, once the moments are computed for the fluid coordinate system, a linear transformation is applied to obtain the loads in the heliostat coordinate system:

Mx heliostat) = ^ (Mx CFd) My heliostat) {My CFD)

The subscript heliostat states the coordinate system of the heliostat and CFD the coordinate system of the fluid. Finally, R is the rotation matrix for mapping between the different coordinate systems. For the torque moment it is not necessary to apply any conversion since the axis is always coincident. An example of the two coordinate systems arrangement and force coefficient values is shown in Figure 6.

100 200 300

Azimuth angle [°]

Fig. 6. (a) Heliostat and aerodynamic coefficients coordinate systems, (b) Aerodynamic coefficients used for calculating theoretical wind loads [18]

Likewise, it is necessary to compensate the moments generated by the forces acting at the centroid of the heliostat with respect to the deformation measurement points. As can be seen in Figure 7, there is a distance L1 and a distance L2 to these points from the force acting point (centroid of the heliostat). These forces will result in moments that have to be added to the resulting moments in the centroid of the system. To compensate for these forces a transformation similar to the one used as above is applied.

Fig. 7. Height difference between the measurement points for the forces and the heliostat centroid.

Finally, it is essential to keep track of the turbulence intensity. The turbulence intensity is calculated by the following expression:

l = J^f0Tiv(t)-v]2dt

Where U (t) is the time series wind speed signal, U is the mean wind speed for the time series signal taken for the calculations (see Figure 8) and T is the total period of the analyzed signal. The turbulence intensity is dimensionless.

Fig. 8. Mean value of wind speed used for calculating turbulence intensity.

Peterka et al. [16] demonstrated the importance of the turbulence intensity on the effects of wind on heliostats. They noticed that high level turbulence intensities alter notably the values of the aerodynamic coefficients. In the following we show the results of an analysis of this behavior. For that purpose, 10 minute time series are going to be analyzed individually, taking into account the mean load values, elevation and azimuth angles and turbulence intensities.

4. Results

The measurement program involves collecting both a comprehensive statistical database and a set of time series, which define the behavior of the heliostat in certain specific situations of wind speed, turbulence intensity and wind direction (Figure 9). First, datasets for every 10 minutes are processed.

11.75----

0 1000 2000 3000 4000

Time [s]

Fig. 9. Excerpt from the measured heliostat mechanical loads.

Furthermore, wind turbulence intensities at the test site are estimated (Figure 10), which are used for the evaluation of the correlation between measured and calculated wind loads.

Wind Speed [m/s]

Fig. 10. Estimated wind turbulence intensity in dependency of measured wind speed.

After applying the analysis presented in the previous sections the following results are obtained. For turbulence intensities corresponding to wind tunnel tests a good correlation is achieved as could be observed in Figure 11 (for one hour time series samples).

My Correlation

___________ "------—

------------

13% Turbulence intensity -----

— Measured -- Theoretical Cale

83% Correlation level

Fig. 11. Correlation between measured and calculated heliostat forces for low turbulence time series.

For turbulence intensities below 16%, correlation between theoretical moments and measured loads are considerably high. However, for high turbulence intensities the correlation levels drop drastically as seen in Figure 12 calculated from a 10 minute time series sample.

Samples [-]

Fig. 12. Correlation between measured and calculated heliostat forces for high turbulence time series.

5. Conclusions

In this paper we have studied wind load effects on a low cost heliostat prototype. It has been observed that for wind tunnel condition turbulence intensities a good correlation is achieved between loads measured by strain gauges and loads calculated by theoretical methods using experimentally obtained aerodynamic coefficients.

However, for turbulence intensities above 16% no correlations are possible because due to turbulences the values for aerodynamic coefficients change dramatically. A possible solution for this problem could be to get derive an aerodynamic coefficient correction method via experimental or CFD applications.

References

[1] Sargent & Lundy LLC Consulting Group: Assessment of Parabolic Trough and Power Tower Solar Technology Cost and Performance Forecasts, NREL. 2003

[2] Kolb et al.: Power Tower Technology Roadmap and Cost Reduction Plan, SAND2011-2419 - April 2011

[3] Zhang Jianhan,Wang Zhifeng: Theoretical and numerical analysis of wind load on 4.46m2 small Heliostat, SolarPACES, Perpignan, France, 2010

[4] Mier-Torrecilla, et al.: Numerical Calculation of Wind Loads over Solar Collectors, Energy Procedia 49 (2014): 163-173

[5] Terrés-Nicoli, et al.: Dynamic Effects of a Heliostat to Wind Loading, Energy Procedia 49 (2014): 1728-1736

[6] Verma, Singh, Ghosh: Evaluation of Wind load on Heliostat, Proceedings of the 37th National & 4th International Conference on Fluid Mechanics and Fluid Power

[7] Pfahl, et al.: Wind Load Reduction for Light-weight Heliostats, Energy Procedia 49 (2014): 193-200

[8] Pfahl et al.: Determination of wind loads on heliostats, SolarPACES 2011 Conference, Granada, Spain, September 2011

[9] Sment and Ho: Wind patterns over a heliostat field. Energy Procedia 49 (2014): 229-238

[10] Bitsuamlak, Dagnew, Erwin: Evaluation of wind loads on solar panel modules using CFD, 5th International Symposium on Computational Wind Engineering (CWE2010), Chapel Hill, North Carolina, USA, 23th-27th. 2010

[11] Huss, S., et al., Evaluating Effects of Wind Loads in Heliostat Design, Proceedings of. SolarPACES 2011, Granada, Spain

[12] Rebolo et al.: Design of heliostats under extreme and fatigue wind loads, Proceedings of. SolarPACES 2011, Granada, Spain

[13] Zang, , et al.: Numerical Simulation of Wind Loads and Wind Induced Dynamic Response of Heliostats, Energy Procedia 49 (2014): 15821591

[14] Teufel, et al.: Dimensioning of heliostat components under wind and gravity load: the map approach, SolarPACES 2008 Conference, Las Vegas, NV, March. 2008

[15] Kunert, et al..: Dimensioning heliostat drives considering dynamic wind loads. In: 15th International. SolarPACES Symposium. 2009

[16] Peterka et al.: Wind loads on heliostats and parabolic dish collectors, NASA STI/Recon Technical Report N 89 (1988): 22179

[17] Holmes.: Wind Loading Structures. 2nd edition published by Taylor & Francis, 2007

[18] Google: Heliostat Wind Tunnel Experiments, http://www.google.org/pdfs/google_heliostat_wind_tunnel.pdf (accessed July 18, 2014)