Energy Procedia 49 ( 2014 ) 2405 – 2413

Available online at www.sciencedirect.com

ScienceDirect

1876-6102 © 2013 Z. Tahboub. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).Selection and peer review by the scientifi c conference committee of SolarPACES 2013 under responsibility of PSE AG. Final manuscript published as received without editorial corrections. doi: 10.1016/j.egypro.2014.03.255

SolarPACES 2013

Modeling of irradiance attenuation from a heliostat to the receiver of a solar central tower

Z. Tahbouba, A. Oumbeb, Z. Hassarb, A. Obaidlic aSolar Eng., Masdar Clean Energy, P.O.Box 54115, Abu Dhabi, UAE

bTotal New Energies, R&D – Concentrated Solar Technologies, F-92069 Paris La Défense, France cProcess Performance Eng., Shams Power Co, P.O.Box 54115, Abu Dhabi, UAE

Abstract

A test bench located at Jabal Hafeet Mountain in Abu Dhabi has been prepared. Four pyrheliometers located at different altitudes from 340 m to 1035 m above mean sea level continuously recorded the Direct Normal Irradiance (DNI) with high accuracy. From the analysis of these measurements, a model simulating the attenuation of direct irradiance with distance of propagation, within 100 m above the surface (approximate height of a tower) was designed. A general shape of the relationship between the atmospheric optical depth from the top of the atmosphere to the ground reference and the extinction coefficient of the irradiance from the ground reference to the target was derived and validated with the use of ground measurements. This relation was used to estimate the DNI at two stations with mean path of sunlight from the reference station equals to 0.3 km and 0.6 km respectively. The comparison of the estimated and measured DNI at these two stations resulted in favorable bias and standard deviation values (bias of 0.3% and 0.4%, and standard deviation equals to 6% and 8% for the two stations respectively). Additionally, the applicability of this method in other locations was discussed.

© 2013 The Authors. Published by Elsevier Ltd. Selection and peer review by the scientific conference committee of SolarPACES 2013 under responsibility of PSE AG.

Keywords: Solar Tower; Irradiance Modeling; Attenuation; Aerosols; Heliostat.

1. Introduction

There has been a growing interest in CSP technologies in the Arabian Peninsula over the last five years. Among the different CSP technologies, Central Tower (CT) technology would be more penalized from the high atmospheric aerosol content in this region due to the extra path that the light has to travel between the heliostat and the receiver. A research program has been developed by Abu Dhabi Future Energy Company (Masdar) and Total for the evaluation of the DNI attenuation under common atmospheric conditions of the peninsula for the CT technologies.

© 2013 Z. Tahboub. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).Selection and peer review by the scientific conference committee of SolarPACES 2013 under responsibility of PSE AG. Final manuscript published as received without editorial corrections.

2406 Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413

The loss of solar radiation due to atmospheric attenuation is not a new study and has been questioned since the late seventies. Vittitoe and Biggs [1] have suggested approximate models to calculate the atmospheric transmittance under two different atmospheric conditions by applying LOWTRAN [2]. These proposed approximations are used in numerical codes such as DELSOL [3] and MIRVAL [4]. Later on, in the eighties, Pitaman and Vant-Hull [5] derived a wavelength independent physical transmission model for the propagation of the solar beam between the heliostat and the receiver. The transmission model was derived based on functional fits to the results data of Vittitoe and Biggs and contains five explicit physical variables: the site elevation above sea level, the atmospheric water vapor density in the air at the site, the attenuation coefficient corresponding to the wavelength band around 0.55 , the tower height and the slant range between the heliostat and the receiver. In a recent work, J. Ballestrin and A. Marzo [6] compared the results obtained by MODTRAN for atmospheric attenuation using a US 1976 standard atmosphere - rural atmosphere to those calculated from the physical model of Pitman and Vant-Hull under the same atmospheric conditions. The results found under different scenarios of slant ranges show a better fit comparing to those of DELSOL or MIRVAL codes. M. Sengupta and M. Wagner [7] from NREL have recently proposed an analytical model for the atmospheric attenuation between heliostats and the receiver as a function of the measured DNI. In addition, A. Alobaidli [8] has proposed using visibility measurements to calculate the extinction coefficient and then using Pitman and Vant-Hull model for estimating the atmospheric attenuation. In a recent work, N. Hanrieder et. al. [9] from DLR have proposed several sensors to measure the visibility and other atmospheric parameters that can be used to estimate the atmospheric conditions and also to update and modify the model if necessary. In this work, we use the experimental setup in Jabal Hafeet to derive and validate a relationship between the atmospheric optical depth and the extinction coefficient in the lower layer of the atmosphere.

Nomenclature

AOD Aerosol optical depth CSP Concentrated solar power CT Central tower DNI Direct normal irradiation SZA Solar zenith angle X Extinction coefficient z Difference in elevation of station 4 and station n τ Optical depth of atmosphere

2. Test setup and data selection

Jabal Hafeet is the highest mountain in UAE. In this experiment, four pyrheliometers are located at different altitudes from 340 m to more than 1000 m above mean sea level (depicted in Fig 1) to continuously record DNI with high precision (pyrheliometers are cleaned daily). Data were recorded for one year at a resolution of 10 minutes starting from February 2012 through end of January 2013.

(a) (b)

Fig. 1 (a) Schematic diagram of the test setup of Jabal Hafeet. (b) Aerial view of the four stations on Jabal Hafeet (Google Earth).

Pyrheliometer 3

Pyrheliometer 4

DNI

Pyrheliometer 1

H = 1035 m

5200 m

Pyrheliometer 2 Jabal Hafeet

H = 840 m

H = 595 mH =340 m

Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413 2407

All data were temperature corrected and calibrated. The calibration of the four pyrheliometers was carried out by placing all four at the top station for several days and then a calibration adjustment factor for each was applied to gain exactly coinciding readings (depicted in Fig. 2).

(a) (b)

Fig. 2 (a) Fine calibration of the four pyrheliometers against each other before the experiment started (b) Station at 1035 [m] during the calibration.

The calibration factor was calculated based on the relative difference of the temperature corrected DNI of each pyrheliometer to the mean value of the four.

(1)

Where, : DNI measured by pyrheliometer 1 to 4.

: Average DNI of the four pyrheliometers. Fig. 3 shows the relative difference to the mean before (Fig 3-a) and after (Fig 3-b) applying the calibration factor

to the temperature corrected measurements.

(a)

0

200

400

600

800

1000

1200

-2,0%

-1,5%

-1,0%

-0,5%

0,0%

0,5%

1,0%

1,5%

2,0%24. Jan. 25. Jan. 25. Jan. 26. Jan. 26. Jan. 27. Jan. 27. Jan.

DN

I [W

/m²]

Rel

ativ

e de

viat

ion

of D

NI

from

mea

n

110507 110512 110528 110529 mean DNI

2408 Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413

(b)

Fig. 3 Relative difference: (a) before and (b) after applying the calibration factor

The comparison of the DNI readings of the different stations in Jabal Hafeet should be carried out with precaution. The fact that the four stations differ in 1- the horizontal position and 2- the surrounding landscape may cause different blocking and shading schemes which will produce outlier data points that needs to be identified and deleted before concluding values of attenuation. Outlier data points may arise from different reasons such as:

Intermittent clouds that lead to shade some pyrheliometers and not the others. Blocking of the sun by nearby objects (due to the surrounding landscape) or passing by objects. For example

the station at 340 m is shaded by the mountain in the afternoon. Moreover, as four different instruments (pyrheliometers and trackers) are used, one should make sure that the

relative difference between any of the four instruments does not exceed a certain limit. The following explains the criteria used for data selection. The data were recorded every 10 minutes and the total

number of timestamps was equal to 52,704 (days and nights) or approximately 26,352 day data.

Relative difference of the pyrheliometers

As mentioned previously, the four pyrheliometers were temperature corrected and calibration adjusted. The relative difference (defined in Equation 1) of the four pyrheliometers was plotted against the mean irradiation (shown in Fig. 4).

Fig. 4 Relative difference between the four pyrheliometers

The relative difference of ±1 % was defined as the threshold of the data selection (shown in Fig. 4 as a two dashed red line). For different irradiation categories (in increments of 50 [W/m2]), the percentage of data points having an absolute relative difference greater than 1 % was calculated. Irradiance categories having a percentage of such data points higher than 5 % were excluded. This resulted that data with irradiances lower than 50 [W/m2] were

0

200

400

600

800

1000

1200

-2,0%

-1,5%

-1,0%

-0,5%

0,0%

0,5%

1,0%

1,5%

2,0%24. Jan. 25. Jan. 25. Jan. 26. Jan. 26. Jan. 27. Jan. 27. Jan.

DN

I [W

/m²]

Rel

ativ

e de

viat

ion

of D

NI

f rom

mea

n

110507 110512 110528 110529 mean DNI

0 200 400 600 800 1000-20

-10

0

10

20

Mean Irradiation [W/m2]

Rel

ativ

e D

iffer

ence

[%

]

1

2

3

4

Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413 2409

excluded from the analysis. For irradiance exceeding 50 [W/m2] only 3.1% of the data had an absolute relative error great than 1 %. 965 timestamps were eliminated when this constraint was applied and the percentage of data remaining from the original data set (excluding nights) was 96.3 %.

Cloud coverage

To give an example of the outliers that can result from cloud coverage, a scatter plot of the data gathered from 28th Jan to 16th Feb, 2012 is shown in Fig. 5. Here the DNI measured at altitude of 340 m is plotted against that measured at 1035 m. It is noticed that some points represent times when DNI340 is higher than DNI1035. These points are most probably created as a result of intermittent clouds covering the station at 1035 m causing lower DNI values at 1035 m than that at 340 m (defined as region 1 which are points above the DNI340=DNI1035 line indicated in blue). These values are not connected to atmospheric attenuation and thus are considered as outliers and should be deleted. Similarly, this can be generalized between any two stations. Therefore, for each timestamp, the following condition should apply otherwise the timestamp would be eliminated:

DNI4 >DNI3 & DNI3 >DNI2 & DNI2 > DNI1 (2)

Where, DNI4, DNI3, DNI2 and DNI1 are DNI measured in stations 4, 3, 2 and 1 consecutively.

9134 timestamps were eliminated when this constraint was applied and the percentage of data remaining from the original data set (excluding nights) was 61.7 %.

Fig. 5 Plot of data set gathered from 28th Jan to 16th Feb, 2012.

Shading by nearby landscape

To make sure that a station is not shaded by the surrounding mountain for certain SZA, all the data points of irradiation at each station were plotted against the SZA. Mornings and evenings were treated separately.

Note that in Fig. 6 and Fig. 7 below, the data were plotted prior to applying the constrains described in the two previous subsections to be able to detect patterns more clearly and consequently be able to define the limits of the SZAs more easily. In fact, the previous constraints overlap this constraint in a large number of instances. Fig. 6 shows the irradiation plotted against the SZA during the morning. Note that for stations 3 & 4 the sun is blocked by the mountain for SZA greater than 75°. This can be seen from Fig. 6-c & d where most of the data is equal to zero for these SZAs.

0 200 400 600 800 10000

100

200

300

400

500

600

700

800

900

1000

DNI @ 1035 m

DN

I @

340

m

Real dataset

DNI340=DNI1035

1

2410 Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413

(a) (b)

(c) (d)

Fig. 6 Morning - Irradiation vs. solar zenith angle for stations: (a) 4, (b) 3, (c) 2 and (d) 1.

616 timestamps were eliminated when this constrain was applied and the percentage of data remaining from the original data set (excluding nights) is 59.3 %.

Fig. 7 shows the irradiation plotted against the SZA during the afternoon. Note that for stations 4 the sun is blocked by the mountain for SZA greater than 60°. This can be seen from Fig. 7-d where most of the data is equal to zero for these SZA.

(a) (b)

(c) (d) Fig. 7 Afternoon - Irradiation vs. solar zenith angle for stations: (a) 4, (b) 3, (c) 2 and (d) 1.

0 20 40 60 80

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

0 20 40 60 80

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

0 20 40 60 80

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

0 20 40 60 80

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

20 40 60 800

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

20 40 60 800

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

20 40 60 800

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

20 40 60 800

200

400

600

800

1000

Sun Zenith Angle [degrees]

Irra

diat

ion

[W/m

2 ]

Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413 2411

920 timestamps were eliminated when this constrain was applied and the percentage of data remaining from the original data set (excluding nights) is 55.8 %.

3. Methodology

We propose a method to model the attenuation. It is based on the Beer-Lambert equation: DNI4 = Io * exp (-τ / cos (SZA)) (3) Where, DNI4: the direct radiation measured by station 4, Io: the extraterrestrial irradiance, τ: the optical depth of the atmosphere, SZA: the sun zenith angle.

Similarly, the irradiance propagating from station 4 to station n will be attenuated as follows:

DNIn = DNI4 * exp (-X * d) (4)

Where, DNIn: the direct radiation measured by station n (3, 2 or 1), X: the extinction coefficient of sunlight between locations of station 4 and station n, d: the distance of propagation of direct sunlight between station 4 and station n and is equal: d = z / cos (SZA) (5)

z being the difference in elevations of station 4 and station n. Fig 8 shows the parameters SZA, d and z.

Fig. 8 Difference in sun path length between stations 1 and 4

Practically, in equation (4), d would correspond to the distance between the heliostat and the receiver at the top of the central tower.

Considering the highest altitude as the reference for DNI (as previously mentioned), extinction coefficient (X) and atmospheric optical depth (τ) are computed for all the measured dataset. In order to eliminate cloudy situations, we remove all cases where the DNI change from a minute to another is greater than 10% and the cases where τ is greater than 1. The correlation coefficient between X and τ for all the dataset is 0.55 (depicted in Fig. 9). This correlation coefficient is good compared to the correlation coefficient between state-of-the-art estimated aerosol

2412 Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413

optical depth (AOD) and the measured AOD, which is usually less than 0.5 in the Middle East [10]. The color bar in Fig. 9 represents the density of points in one pixel of the graph.

Fig. 9 Variation of the extinction coefficient with the optical depth

From regression, the relation between X (calculated between station 1 and 4) and τ is: X = 0.23 * τ + 0.022 (6) This relation was validated by calculating the DNI of station 2 and 3 (using equations 4 and 6) and then by

comparing the calculated values to the ground measurements. The results are depicted in Fig. 10. Note that the correlation coefficients are 0.96 and 0.97 for station 2 and 3 respectively.

(a) (b)

Fig. 10 Calculated vs. Measured DNI in (a) station 2 and (b) station 3

A Similar model has been proposed by Sengupta and Wagner [7], but it has not been validated at a specific location. The model is analytical and is made for a constant tower height at 250 m; it computes the DNI received at the top of the central tower, assumed at 250 m height, from the DNI reflected by the heliostat. It is based on a linear

0.2 0.4 0.6 0.8 1 1.2 1.4

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Optical depth ( ) of the atmospheric column

Ext

inct

ion

co

effic

ien

t (X

) in

the

dir

ectio

n o

f pro

pag

atio

n

r = 0.550y = 0.227x +(0.022)

Number of points = 35,520

10

20

30

40

50

60

0 200 400 600 800 1000 12000

200

400

600

800

1000

1200

Measured DNI @ station 2 [W/m2]

Cal

cula

ted

DN

I [W

/m2 ]

rSTD = 8.01%bias = 0.25%

mean = 652.2

r = 0.961y = 0.979x +(13.932)Number of points = 14,070

5

10

15

20

25

30

0 200 400 600 800 1000

200

400

600

800

1000

Measured DNI @ station 3 [W/m2]

Cal

cula

ted

DN

I [W

/m2 ]

rSTD = 6.36%

bias = -0.42%

mean = 684.9

r = 0.972y = 0.989x +(6.753)Number of points = 14,070

5

10

15

20

25

30

35

40

45

Z. Tahboub et al. / Energy Procedia 49 ( 2014 ) 2405 – 2413 2413

relationship between 1) the optical depth difference between the current atmosphere (X) and a clean baseline atmosphere with no aerosol and 2) the optical depth of the 250 m lowest atmospheric layer (Y). If the distance d between the heliostat and receiver is known, and also DNIsfc is known, then the DNI at the receiver (DNIrec) can be calculated from the equation below:

DNIrec= DNIsfc. Exp(-Y*d/250) (7)

Where, DNIsfc is the DNI that is reflected from the heliostats and factors in cosine losses, blocking losses and reflectivity. The “clean baseline atmosphere” is computed using a radiative transfer model (RTM).

4. Proposed method for the generalization of the model

In preparation for any CSP project, a ground measurement of the solar resource is carried out. For CT technology projects, we propose the addition of a transmissometer to measure visibility in addition to the usual DNI measurement. The transmissometer will be used to calculate the extinction coefficient.

With the proposed model, the measured DNI will be used to compute the all-atmosphere optical depth, and the visibility will be used to compute the extinction in the lower layer of atmosphere. Then equation 4 can be applied to estimate the DNI that reaches the receiver. With the model presented in [7], the measured DNI will be used to compute the difference between the optical depth of the current atmosphere and the optical depth of a “no aerosol” atmosphere (the “no aerosol” atmosphere can be simulated with a RTM) and the visibility will be used to compute the optical depth of the bottom 250 m atmospheric layer. Then equation 7 can be applied.

5. Conclusion

A relationship between the atmospheric optical depth - from the top of the atmosphere to the ground reference - and the extinction coefficient of the irradiance - from the reference position to the target - was derived and validated using ground measurements. This was done using four DNI measurement stations distributed at different altitudes on Jabal Hafeet in UAE. In our setup the reference DNI (station 4) corresponds to the DNI received by a heliostat. The other three stations correspond to three receivers at various distances. We derived a relation between the optical depth and extinction coefficient using data from station 1 and 4. This relation was used to estimate the DNI at stations 2 and 3 which compared favorably with the ground measurements (bias of 0.3% and 0.4% and the standard deviations equal to 6 % and 8% for the two stations respectively). The applicability of this method in other locations is discussed.

References

[1] Vittitoe, C.N. and F. Biggs. Terrestrial Propagation Loss. Denver : presented at the American Section of the International Solar Energy Society, 1978. Sandia release SAND78-1137C, May 26, 1978.

[2] Selby, J. E. A. and R. A. MacClatchey. Atmospheric Transmittance from 0.25 to 28.5 μm: Computer Code LOWTRAN 3 . National Technical Information Service (ADA – 017723).

[3] Kistler, B. L. A User’s Manual for DELSOL3: A Computer Code for Calculation the Optical Performance and Optimal System Design for Solar Thermal Central Receiver Plants . 1986. Sandia Laboratories Report SAND86-8018.

[4] Leary, P. L., J. D. Hankins. A User’s Guide for MIRVAL— a computer code for comparing designs of heliostat-receiver optics for central receiver solar power plants. 1979. Sandia Laboratories Report, SAND77-8280.

[5] Pitman, C.L., L.L Vant-Hull. Atmospheric transmittance model for a solar beam propagating between a heliostat and a receiver. 1982. ASES Progress in Solar Energy, 1247– 1251.

[6] Ballestrin, J. and A. Marzo. Solar Radiation Attenuation in Solar Tower Plants. Granada : SolarPACES, 2011. [7] Sengupta, M., Wagner M.. Atmospheric attenuation in central reciever systems from DNI measurments. SolarPACES, 2012. [8] Al Hammadi, A.A, Master’s thesis, Aachen, April 2012. [9] Hanrieder, N et. al., Determination of beam attenuation in tower plants, SolarPaces, 2012. [10] M. Schroedter-Homscheidt and A. Oumbe (2013). Validation of an hourly resolved global aerosol model in answer to solar electricity

generation information needs. Atmospheric Chemistry and Physics, 13, 3777–3791, 2013