VISUAL HFLCAL – A SOFTWARE TOOL FOR LAYOUT AND

OPTIMISATION OF HELIOSTAT FIELDS

Peter Schwarzbözl1 , Mark Schmitz2 and Robert Pitz-Paal3

1
German Aerospace Centre, Solar Research, 51170 Köln, Germany, ++49 601 2967, [email protected]
2
FH Aachen, Solar-Institut Jülich, 52428 Jülich, Germany
3
German Aerospace Centre, Solar Research, 51170 Köln, Germany

Abstract

HFLCAL is a program for layout and optimization of heliostat fields of central receiver systems (CRS) based
on annual performance calculation. Computation time for performance estimation is saved by using a
simplified mathematical model for the concentrator optics: the reflected image of each heliostat is described
by a circular normal distribution. This approximation is justified when the standard deviation of the statistical
mirror error exceeds 1 mrad. The paper shows details about the mathematical model and its validation. The
HFLCAL code has been continuously used and enhanced in numerous R&D projects. The current features of
HFLCAL comprise (among others) automatic multi-aiming, secondary concentrator optics, tower reflector
systems, various receiver models and the ability of least-cost optimization with various optimization
algorithms. A graphical user interface has been added to the program that supports menu-driven interactive
commands and depicts calculation results in a display window.
Keywords: solar tower plant, central receiver system, heliostat field, modeling, simulation, optimization

1. Introduction

The computer program HFLCAL was developed by Michael Kiera at the German company Interatom during
the project GAST (GAS-cooled Solar Tower) in the early 1980’s ([1]). GAST was a bilateral German-
Spanish cooperation aiming at the development and the investigation of necessary solar specific components
and software for a gas-cooled tower power station of medium size.
The HFLCAL code was developed for two main tasks, the calculation of the annual plant output at a given
configuration and the layout and optimization of a total system with respect to maximum annual electric
energy yield per collector unit.
HFLCAL was acquired by the German Aerospace Centre (DLR) in 1994 and adapted it to run on current
personal computers instead of a mainframe machine. Since then the program was continuously used and
further developed in numerous R&D projects related to central receiver systems.

2. Mathematical Model of Solar Concentration

HFLCAL is a program for the layout of concentrator fields based on the calculation of intercepted power of a
larger number of heliostats. The HFLCAL model assumes all heliostats having well canted concentrating
facets of spherical curvature. The reflected image of each heliostat at a given point in time is described by
one single circular normal distribution of the energy flux
r²
1 −
F (r ) = e 2σ ² . (Eq. 1)
2πσ ²
In reality, the size and shape of the reflected image at a given point in time is influenced by the finite size of
the sun and the quality of the mirror curvature and the mirror surface. If the incident ray is not parallel to the
mirror’s normal, the reflected beam is further deformed by the astigmatic effect. In HFLCAL, the intercepted
energy from each heliostat over a certain period of time is of interest. Therefore, the aim-point uncertainty
due to the tracking mechanism has to be considered. All these influences are aggregated into the circular
normal distribution with one characteristic value σ, which can be assumed as a superposition of normal
distributions:

σ 2
= σ sun
2
+ σ beam
2
quality + σ astigm + ( 2 ⋅ σ track ) . (Eq. 2)
2 2

Sunshape
The solar intensity is not distributed uniformly across the sun disc but decreases towards the edge. It is often
described with the Kuiper sunshape model ([3], Figure 1 left). A normal distribution with σ=2.24mrad is
statistically comparable because it has the same root mean square deviation from the central ray. But the
graph in figure 1 left shows a systematic overestimation of the peak at the expense of the edge. Nevertheless,
the approximation of a heliostat image by a circular normal distribution is supported by the Central Limit
Theorem that states that the superposition of arbitrarily distributed quantities converges towards a normal
distribution. The convergence is accelerated when some of the quantities are already normally distributed.

0.035

0.030

"Kuiper"
0.025
"pillbox"
"Gaus s ian"
0.020

0.015

0.010

0.005

0.000
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
r [mrad]

Figure 1. Left: Normalized angular distributions of solar intensity with identical root-mean-sqare (RMS)
deviation from the central ray of 3.17mrad. “Kuiper” model sunshape ([3]), “Gaussian” normal distribution
with σ=2.24mrad and “pillbox” distribution with radius=4.48mrad. Right: Off-axis-reflection of light at a
spherical concentrator (Ψ: incident angle, f: focal distance of the concentrator, d: diameter of concentrator)

Beam Quality
The beam quality accounts for deviations of the mirror curvature from the ideal shape and imperfections of
the reflecting surface due to waviness and roughness. These mechanisms can be measured as the so-called
slope error and are described statistically by a circular normal distribution. The slope error is defined with
respect to the surface normal vector; therefore its effect is doubled in the reflected ray:

σ beam quality 2 = (2 ⋅ σ slope ) 2 . (Eq. 3)

Astigmatism
At off-axis reflection (Ψ≠0°) with a concentrator of spherical or paraboloidal shape parallel light rays are
concentrated into two focal lines rather than a single focal point (Figure 1 right). For f/d >> 1, rays in the
tangential plane (defined by the incoming and the reflected central ray) are concentrated in a line focus at
distance f·cosΨ from the mirror, while the rays in the perpendicular sagittal plane are concentrated in a line
focus f/cosΨ from the mirror. In between, at distance f, the reflected beam has minimal dimensions
D= d (1-cosΨ). At an arbitrary distance SLR from the mirror the image dimensions in the tangential and
sagittal plane are ([5]):
 SLR SLR
Ht = d ⋅ − cos Ψ ; Ws = d ⋅ ⋅ cos Ψ − 1 ; (Eq. 4)
f f

A heliostat usually consists of several facets and astigmatism occurs in the facet and in the total heliostat. In
HFLCAL, each heliostat is assumed to be canted correctly for Ψ=0°. The image dimensions due to off-axis
reflection are described as superposition of astigmatism in the heliostat and in the facet. The root-mean-
square of the image extensions in the tangential and sagittal plane is treated as diameter of a pillbox
distribution. It is incorporated into the HFLCAL image description as additional widening of the reflected
beam:

1
2 (H 2
t , hel + facet + Ws2,hel + facet )
σ astigm = ; (Eq. 5)
4 ⋅ SLR
Tracking
The heliostat reflecting surface is usually moved about two axes with motors and gears. Tracking errors, i.e.
deviations of the mirror normal from the intended direction, are caused by the finite motor step size,
tolerances of the gear boxes and wind loads on the structure. The tracking error is usually measured with
respect to the mirror normal vector; therefore its effect is doubled in the reflected ray (see Eq. 2). The
statistical deviations of the two tracking axes are combined into one circular symmetric distribution in
HFLCAL:

σ track = σ axis1 ⋅ σ axis 2 . (Eq. 6)

This is a conservative approach, valid for on-axis reflection. For off-axis reflection the influence of the
tracking uncertainty is dependent on the angle between the incident ray and the axes.
Performance Calculation
When the heliostat image is described with Eq. (1), the flux distribution has to be integrated along the
receiver aperture plane to get the intercepted power at a certain point in time. The intercept expressed as
efficiency then reads:
ζ 2 +γ 2
−
1 2σ 2
η incp =
2πσ 2 ∫∫ e
aperture
dζ d γ ; (Eq. 7)

The radiation power at time t from a single heliostat at location (x,y) in the field into a given aperture is then
calculated as:

P( x, y, t ) = DNI (t ) ⋅ FMir ⋅η refl ⋅η atmo ( x, y ) ⋅η cos ( x, y, t ) ⋅η b & s ( x, y, t ) ⋅η incp ( x, y, t ) , (Eq. 8)

where DNI is the direct normal solar radiation, FMir is the heliostat reflective surface, ηrefl is the mirror
reflectivity, ηatmo is the atmospheric attenuation depending on the distance from the receiver, ηcos is the cosine
factor depending on heliostat and sun position and ηb&s accounts for the losses due to blocking and
shadowing by the tower and neighbouring heliostats. To calculate the latter effect, a group of heliostats
around each heliostat is checked geometrically for shading and blocking interferences by projecting the
mirror outlines onto the planes of their neighbours.
The annual energy yield of a single heliostat is estimated as the weighted sum of power values for about 100
representative time points:

Ei ,annual ( x, y ) = ∑ w(t ) ⋅ Pi ( x, y, t ) ; (Eq. 9)

t

The weighting factor w(t) includes the corresponding interval length and the multiplicity of each
representative time point.
3. Validation

Pettit, Vittitoe and Biggs ([4]) early discussed the possibility of describing the angular distribution of
reflected light from a solar concentrator with a circular normal distribution. They calculated the convolution
of measured sunshapes of various widths with normally distributed mirror errors and compared the results
with circular normal distributions. They found very good agreement when the mirror error was more than
two times larger than the dispersion of the sun (σmirror ≥ 2σsun).
To assess the usability of the HFLCAL model here, it was compared with ray-tracing results performed with
MIRVAL ([2]). First, a single heliostat was considered consisting of 16 facets of 6m² each with 100%
reflectivity. The heliostat is canted for on-axis reflection and the focal length of the facets is equal to the slant
range of 141m. Figure 2 shows the results for on axis reflection of solar light with DNI of 1000 W/m². For
the ray-tracing calculations a “Kuiper”-sunshape as shown in Fig. 1 was used with RMS width of 3.17mrad.
Therefore, the beam dispersion for HFLCAL was set to σHFLCAL = ((2.24mrad)²+ σbeam quality²)½.

200 200

180 180
ray-tracing ray-tracing
160 160
HFLCAL HFLCAL

140 140

120 120

100 100

80 80

60 60

40 40

20 20

0 0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

a) mirror slope error: 0.0mrad (normal); σHFLCAL=2.24mrad b) mirror slope error: 0.0mrad (normal); σHFLCAL=2.63mrad
200 200

180 180
ray-tracing ray-tracing
160 160
HFLCAL HFLCAL

140 140

120 120

100 100

80 80

60 60

40 40

20 20

0 0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

c) mirror slope error: 1.0mrad (normal); σHFLCAL=3.00mrad d) mirror slope error: 2.0mrad (normal); σHFLCAL=4.58mrad

Figure 2: Flux profile of single heliostat with varying mirror beam error. Comparison of HFLCAL and
MIRVAL results.

Comparing Figures 2 a), c) and d) shows, that while for small mirror errors the HFLCAL image is too narrow
and has an overestimated peak flux, the deviations decrease as the mirror error increases and good agreement
is reached when σHFLCAL ≈ 2σsun. But when the root mean square deviation of the two-dimensional HFLCAL
image from the ray-tracing result is considered, a minimum deviation is reached for a σsun other than
2.24mrad (Figure 2 b)).
This was repeated for various mirror errors and the results are shown in Figure 3. As can be seen, the beam
dispersion for HFLCAL, σHFLCAL, (total sigma in Fig. 3 left) is converging towards 2 σslope because the
influence of the solar distribution vanishes with increasing mirror errors. The RMS deviation rapidly
decreases below 1% for realistic mirror slope errors >1mrad.

10 4.5%

9 4.0%

8 3.5%

RMS (HFLCAL-RayTracing)
7 3.0%
total sigma [mrad]

6
2.5%

5
2.0%
4
1.5%
3
1.0%
2
0.5%
1
0.0%
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
slope error (normal) [mrad]
slope error (normal) [mrad]

Figure 3: Best fit of HFLCAL image as a function of mirror slope error. Total sigma (left) and RMS
deviation of HFLCAL image from ray-tracing result at best fit (right).

[kW/m²]
500-550
kW/m² 450-500
500-550
400-450
450-500
350-400
400-450
350-400 300-350
300-350 250-300
250-300
200-250
200-250
150-200
150-200
100-150 100-150
50-100 50-100
0-50
0-50

-2.4 -1.9 -1.4 -1.0 -0.5 0.0 0.5 1.0 1.4 1.9 2.4
-2.4 -1.9 -1.4 -1.0 -0.5 0.0 0.5 1.0 1.4 1.9 2.4

Figure 4: Flux distribution of heliostat field for average incident angle of 45° (left: HFLCAL with
σHFLCAL=3.17mrad, right: MIRVAL with mirror slope error of 1mrad).

Finally, a rectangular field of 7x7 heliostats (96m² each) was considered with all heliostats aiming at the
centre of the target at an average slant range of 141m². The HFLCAL distribution agreed almost perfect with
the ray-tracing result. At off-axis reflection the image is distorted due to the astigmatic effect and the
HFLCAL model only gives an estimate of the mean diameter of the image but it does not reproduce the
correct shape due to the assumption of symmetry in Eq. 1 (see Figure 4). Nevertheless, the intercepted power
differs – depending on the aperture size - only by few percent. The different mechanisms in the field are
calculated as shown in Table 1.
 HFLCAL ray-tracing (Mirval)
cosine 0.7067 0.7067
blocking & shading 0.8015 0.8086
atmospheric att. 0.9764 0.9765
interception 0.9703 0.9755
total 0.5366 0.5443
Table 1: Comparison of loss mechanisms for 7x7 heliostat field between HFLCAL and MIRVAL.

4. System Layout and Optimization

A solar tower system is configured in HFLCAL by specifying the heliostat (dimensions, beam error as in Eq.
2, reflectivity), the tower (height, radius), the receiver (shape, size, inclination, thermal efficiency) and the
plant location (latitude, solar resource). The thermal receiver efficiency represents the link between the
thermal losses and the optical losses of the system. A load dependent model, which is usually non-linear, is
important for the correct weighting between time points.
The system layout is the determination of the required collector field size to reach the given plant design
power at the design point time. Thereby, HFLCAL starts with a gross field of hypothetical heliostat positions
described by a parameterized distribution, for example by linear functions of the heliostat angular and radial
distance (Figure 5 left). The annual yield of intercepted energy of each heliostat in the gross field is
calculated according to Eq. 9. Then the heliostats performing best on the annual basis are chosen and their
design point power is added up until the desired total power can be provided (Figure 5 right). The result is a
field layout with positions of single heliostats relative to the tower base.

Figure 5: Left: Parameterized heliostat field setup. Right: Screenshot of Visual HFLCAL showing gross field
and selected field. The colours describe the design point efficiency.

The simplified optical model of HFLCAL, as described in Chapter 2, allows performing a system layout
calculation in very short computation time, usually only few seconds. Therefore it is possible to include the
layout calculation in an optimization procedure. In principle, any optimization algorithm can be wrapped
around the HFLCAL core to perform subsequent system layout calculations while system parameters are
manipulated. Two options are currently implemented in Visual HFLCAL: Powell's Method as a direction set
method for multidimensional minimization ([6]) and a genetic algorithm that maximizes the system
performance by evolutionary mutation of system parameters ([7]). Best results were usually attained by
combining both optimization techniques.
Additionally, cost functions can be defined to calculate the design-dependent capital investment and O&M
costs for the specified solar tower system. This enables least-cost optimization of the solar part of a tower
plant by minimizing the cost per MWh thermal receiver energy.
5. The Visual HFLCAL Software

While some of the core routines of HFLCAL are still in the original FORTRAN77 standard, most new
features are programmed in FORTRAN95 and Visual-FORTRAN. It was decided to use Visual-FORTRAN
for the user interface rather than setting up the program completely new in some other code like Visual-C++
or linking the HFLCAL code as DLL to a separate user interface. Visual-FORTRAN offers sufficient
possibilities to create a windows-based graphical user interface with interactive dialogues, pop-up menus and
display of results. The advantage is that all routines are combined in a single software project and handled
with the same compiler. This facilitates the enhancement of the code with new features.
The Visual HFLCAL user interface allows navigating through the calculation process and provides
interactive menus for all input parameters. Very valuable is the graphical system display that directly depicts
definitions by the user and shows intuitive representations of the calculation results (Figure 5 right). The
calculation modes comprise field layout, annual performance calculation, flux density calculation, system
optimization and the creation of a field efficiency matrix to be used in other simulation tools (e.g. TRNSYS
[10]).
A menu driven user interface makes it easy to change parameters but does not allow the user to change or add
more complex information like model equations. Therefore, some routines were outsourced into a dynamic
link library and the source code can be manipulated by the advanced user. The outsourced routines comprise
the models for the thermal receiver efficiency and the models for cost calculations. Parameters for the user
defined routines can be provided through the user interface of the main program.

6. Applications

HFLCAL has been used for a wide range of solar tower technologies. These include direct steam systems
with cavity receiver, open volumetric air receiver with cylindrical aperture, pressurized air receiver with
secondary concentrator ([8]) and beam down systems with tower reflector and secondary concentrator ([11]).
The optics of a secondary concentrator for example, was included as incident-angle-dependent transmissivity
as an additional loss to the optical system (Figure 5 left). This resulted in long narrow field layouts for solar
gas turbine systems with pressurized receiver. Overall system optimization was performed including single
and multiple receiver clusters. (Figure 5 right, [9], [12], [13])

Figure 5: Left: HFLCAL-screenshot with field layout for receiver with secondary concentrator. The colours
represent the secondary transmissivity. Right: Field layout for large solar gas turbine plant with six receiver
clusters. Total intercept power is about 230 MW.
7. Conclusion

The HFLCAL model is suitable for the estimation of intercepted power of concentrating heliostats. It allows
fast performance calculation for large heliostat fields as a basis for system layout and optimization. The
HFLCAL model is not suitable for the simulation of non-concentrating or line-focusing mirrors. HFLCAL is
also not suitable for the detailed analysis of flux density distributions and the corresponding receiver design.
This should be done with detailed ray tracing codes.
HFLCAL shows two important features as compared to other codes. First, all calculations are based on
individual heliostats and the result of a HFLCAL field layout are single heliostat positions. And second, the
thermodynamic performance of the receiver can be included allowing overall system layout and optimization.
The graphical user interface of Visual HFLCAL adds the interactive usability of modern software.

References

[1] Kiera, M.: Heliostat Field: computer codes, requirements, comparison of methods. In: Becker, M.,
Böhmer, M. (Eds.): GAST – The Gas-Cooled Solar Tower Technology Program. Proceedings of the Final
Presentation. Springer Verlag, Berlin 1989
[2] Leary, P.L., Hankins, J.D.: A User’s Guide for MIRVAL - a computer code for comparing designs of
heliostat-receiver optics for central receiver solar power plants. Sandia Laboratories Report, SAND77-
8280, 1979
[3] Biggs, F., Vittitoe, C.: The Helios Model for the optical behaviour of reflecting solar concentrators.
SANDIA report SAND78-0347. Albuquerque, USA, 1979
[4] Pettit R. B., Vittitoe C. N., Biggs F.: Simplified Calculational Procedure for Determining the Amount of
Intercepted Sunlight in an Imaging Solar Concentrator. Journal of Solar Energy Engineering, Vol. 105,
1983
[5] Igel E. A., Hughes R.L.: Optical Analysis of Solar Facility Heliostats. Solar Energy, Vol. 22, 1979
[6] Press W.H., Flannery B.P., Teukolsky S.A. and Vetterling W.T.: Numerical Recipes in FORTRAN77:
The Art of Scientific Computing. Cambridge University Press, 1st Edition, 1986.
[7] Carroll, D. L.: Chemical Laser Modeling with Genetic Algorithms. AIAA J., Vol. 34, 2, 1996. (General
information about the genetic algorithm code used in HFLCAL can be found at www.cuaerospace.com).
[8] Buck R., Bräuning T., Denk T., Pfänder M., Schwarzbözl P. and Téllez F.: Solar-Hybrid Gas Turbine-
based Power Tower Systems (REFOS). ASME Journal of Solar Energy Engineering 2002, 124: 2-9
[9] Schmitz, M., Schwarzbözl, P., Buck, R., Pitz-Paal, R.: Assessment of the potential improvement due to
multiple apertures in central receiver systems with secondary concentrators. Solar Energy 80, 2006
[10] Schwarzboezl P.: A TRNSYS Model Library for Solar Thermal Electric Components (STEC).
Reference Manual. Release 3.0, November 2006 (available at
http://sel.me.wisc.edu/trnsys/trnlib/stec/stec.htm)
[11] Schmitz M.: Systematischer Vergleich von solarthermischen Turmreflektor- und Turmreceiversystemen.
Fortschritt-Berichte VDI Reihe 6 Nr. 556, VDI Verlag, Düsseldorf 2007.
[12] Pitz-Paal, R.; Schwarzbözl, P.; Buck, R.: Solarunterstützte Gasturbinensysteme - Von der
Mikrogasturbine zum GuD-Kraftkwerk. Fachverband VGB PowerTech e.V. (Ed.): VGB PowerTech, 10,
2007., S. 42 – 47
[13] Buck R., Schwarzbözl P.: Solarized Gas Turbine Power Systems. 4th International Conference "The
Future of Gas Turbine Technology". 15-16 October 2008, Brussels, Belgium.