Solar Energy Vol. 27, No. 6, pp. 555-560, 1981 0038-092X/81/060555-.06502.

0010
Printed in Great Britain. Pergamon P~ess Ltd.

SIMPLIFIED METHOD FOR PREDICTING

PHOTOVOLTAIC ARRAY OUTPUT

D. L. EVANS
Mechanical and Energy Systems Engineering,Arizona State University, Tempe, AZ 85281, U.S.A.

(Received 27 October 1980; accepted 6 July 1981)

Abstract--A simplified procedure for predicting the long term, monthly average electrical output of photovoltaic
arrays is presented. It is restricted to passively cooled, max-powertracked arrays, but is applicable to both south
facing, fixed fiat arrays and 2-D tracked concentrators. The procedure combines basic parameters characterizing
the array with the local monthly mean temperature and the monthly Kr (ratio of the total radiation on the
horizontal to the extraterrestrial radiation) to yield a monthlyaverage array efficiencywhich, when multiplied by
the monthlyarray insolation gives the electrical energy output.

I. INTRODUCTIONS design, cell temperature and array irradiation, and is

In photovoltaic (PV) system design it is necessary to typically characterized by [2, 3]:
predict the potential output of a given solar cell array
under various conditions. If the output is small enough rt, = rt,[l- fl(T~.,- T,)+ y logto Id. (2)
and coincides timewise with the load to be met, the solar
contribution to the load can often be easily estimated Most often this equation is seen with y = 0 , e.g.[3].
from the calculated array output. If the output is large Adding and subtracting Taj and Tu to the terms in
and/or does not coincide timewise with the load, then parentheses in eqn (2) and combining eqn (2) with eqn (1)
predicting the solar contribution to the load is not as easy yields:
since not all of the array output may be useful. However,
in this case, the calculation of array output is desirable Qa, = ( n,AI N~)['Z~ - fl'Z( L., - T,,.,)~
since it can play an important role in correlating system
- fl'Z(To.,- TM)I,- fl'Z(TM - L)I,
performance results [1].
This paper is devoted to a simplified procedure for + 3,Z/, 10810/,]. (3)
calculating array output starting with a minimum of input
From a simplified standpoint, it would be preferable to
information. The approach adopted here is to find a mean
calculate Q,, from:
monthly array efficiency that, when multiplied by the
mean monthly solar irradiation of the array, yields
Qae = ~AEIJNa (4)
essentially the same electrical energy production as the
summation of the hourly outputs over the month.
where 77 is a monthly average conversion efficiency and
2. T H E A S S U M P T I O N S
the summation is just the monthly average daily in-
solation on the array.
It is assumed here that the solar cell arrays are max-
For an appropriately defined (To-Ta), ( T a - TM),
power tracked; i.e. the voltage on the array is con-
(TM- TA, and log~o/, an expression similar to eqn (2)
tinuously adjusted such that the power produced is a
can be used to represent ~ (T, and TM have been added
maximum. However, other modes of operation can
and subtracted):
produce nearly this maximum amount of energy. Also,
the analysis and results that follow strictly apply only to
,7 = ,7,[I - fl(L - To)- #(To - T,,,,)- fl(TM - L)
passively cooled arrays in which the heat sink for the
+ y log~oI]. (5)
thermal energy deposited in the array is the ambient
environment.
When Qa, from eqn (3) is equated to Qa, resulting
3. THE ANALYSIS
from substituting eqn (5) into eqn (4), it is apparent that
(T~ - Ta), (Ta - TM) and logxoI could satisfy:
Based on insolation data integrated over one hour time
intervals, the monthly average daily array output is given
(T. - T,~)'ZI, = E(T~,, - TJI, (6)
by:
(T~ - T,,,)2~ = "~(To., - TM)I~ (7)
Qa, = A~,~,IJNa (1)
logto IE/~ = ~/~ logto/~. (8)
where the summation goes over all hourly intervals in the The term (To - Ta)
month and ~7~is the "hourly" efficiency. This efficiency Intuitively (T~- Ta) in eqn (6), represents the monthly
for max-power operation, is a function of array (and cell) average difference between the cell and ambient tem-

SE Vol.27, No. 6--H 555

556 D. L. EVANS
peratures during daylight hours. The (T~.~- Ta,~) term, to data is shown in Fig. 1, which relates long term
which it is related, should be driven by the insolation on UL(Tc-Ta)/(aO) to the long term average Kr, the
the array and the thermal losses from the array to the ratio of total radiation on the horizontal to extrater-
environment. Equating solar gains in the array to the restrial radiation[6]. Long term Kr's were also obtained
electrical output and the thermal losses yields, for a unit from the SOLMET data. If any radiation data exist for a
of aperture area: site they are usually total on the horizontal data from
apl~ = ~,pI~ + UL(T~.~ - T J (9) which Kr can be obtained; therefore, it is a convenient
correlation parameter. Its definition certainly hints that it
UL, the loss coefficient, is the thermal conductance (per may be a good relative measure of the average hourly
aperture area) for heat transfer from the cells to the insolation at a site.
surroundings[4]; it will be assumed that an adequate A tilt dependent correction factor, C¢, for converting
average value of UL can be defined. results of Fig. 1 to results for non-optimum tilts is shown
If rh is neglected in eqn (9) (it is usually small com- in Fig. 2. These data were calculated for summer and
pared to a), (T~,~- T J can be obtained from eqn (9) and winter months in Albuquerque, NM and Madison, Wl.
used in eqn (6) to yield: The correction factor for a particular tilt is multiplied by
the monthly UL(T~ - T~)/(ap) for an optimum tilt (from
U~(T. - T~)l(ap) : '~I~lXI~. (lO) Fig. 1) to obtain a monthly UL(Tc- T,)/(ap) for the
non-optimum tilt.
The quantity UL(T~- T~)/(ap) was computed monthly 2-D Tracking. The quantity UL(Tc - Ta)/(ap) was also
using eqn (10) for seven widely varying climatic locations calculated for 2-D tracked concentrating collectors. Here,
in the U.S. for which SOLMET[5] data were available. /,. was just the beam or direct normal radiation from the
The locations and number of years of data used are SOLMET data, Ib. Although only the data for monthly
shown in Table 1. optimum tilted flat arrays are shown in Fig. l, there is
Fixed flat arrays. The flat array calculations were made little detectable difference in the data for 2-D tracked
for south facing arrays tilted at an angle that gives concentrators as long as UI~ is the loss coefficient or
optimum incident energy each month. These optimum thermal conductance based on unit aperture area. This
angles are essentially only latitude dependent and can be similarity is a result of the term ~I2/~I~ most heavily
calculated using Table 2. The SOLMET data were con- weighting the high insolation periods near midday for
verted hourly for the local latitude to total radiation on fixed arrays.
the tilt using the standard year corrected total radiation Long term variations. A designer often needs to
and the direct normal radiation. Assumptions of uniform know what variation might be expected from the long
sky for adjusting diffuse radiation and a ground term averages shown in Fig. 1. Statistics of the scatter in
reflectivity of 0.2 were used. the difference between the monthly Kr at a given loca-
For each location, a long term average UL(Tc- tion and the long term average Kr for that same month
To)/(ap) was computed for each month from the multiple and same location show a nearly Gaussian distribution of
years of data with I~ = IF. A simple and effective cor- standard deviation, ~=0.042 (1). Likewise, individual
relation between this parameter for fiat arrays and solar data points (all months and locations) are scattered ver-
tically about the straight line fit in Fig. l, in a nearly
Gaussian fashion with a ~=0.15MJ.m-2.hr I or
Table 1. SOLMETsites and years of data used in this study 0.043 kW. m 2. This o- is shown on Fig. 1.
Number
Site o f Years Determining UL. Predicting UL is often difficult since
Albuquerque, NM 22
it involves knowing the thermal performance of the
Bismarck, ND 22
Madison, Wl 22 array, the wind speed and direction and secondary flow
Medford, OR 22
effects created by elements making up the array field.
P h o e n i x , AZ 22
Santa M a r i a , CA 16 For flat arrays, the Nominal Operating Cell Temperature
Washington D . C . / S t e r l i n g , VA 21
(NOCT)[7] is a useful concept in determining UL if the
application under study uses the array mounted similar
to that specified in the NOCT test procedure.
Table 2. Monthlyoptimumtilts--tilt angle (su) between the plane
of the flat a~ay and horizontal* Based on an energy balance, it can be shown (neglect-
Month s~(Delrees)**
ing the electrical output of the array as the NOCT test
January 0 + 29 does) that UL and a are related to NOCT via UL/a =
February .................. 0 + 18
March 0 + 3
GT,NOCT/(NOCT-To.NOCT).Traditionally, T~,Nocvis 20 C,
April ..................... O - 10 Gr.NOCT is either 3.6 or 2.88 MJ.hr -~.m 2 [1.0 or
May O - 22
flame . . . . . . . . . . . . . . . . . . . . . . 0 - 25
0.8 kW. m-2], and the wind speed is 1 m. s -~. At other
July O - 24 wind speeds this value has to be adjusted. This is dis-
August .................... O - 10
September O - 2
cussed later in this paper.
October ................... O ÷ 10
November O + 23
December . . . . . . . . . . . . . . . . . . O + 30
The term (T, - TM)
Manipulation of eqn (7) yields:
*Array is assumed to be south
facing
**0 is latitude in degrees (T~ - T~)= (X To.£/X/i)- T~. (11)
 Simplified method for predicting photovoltaic array output 557

3.5
± ~ ~ o.9
3.0
~ O.8
Z
c

% 2.5 ).7

2.0
I
~ o / D.5~
O O UL(T • ) ~ 0 . 7 8 9 + 2 . 9 9 6 KT M J / ( m 2 " h r : :
1.5 3.4
Ot)o ) O . 2 1 9 + C.83Z ~'T KW/m2
A -- Albuquerque, NM 3.3
Ii0
B -- B i s m a r c k ND
M- Madlson, Wl 3.2
0 -- M e d f o r d , OR
0.5 P -- P h o e n i x AZ
9.1
S -- 50ntcl M o r k L CA
W -- W e s h i n g t o n l S t e r l i n g , VA
I I I I I I I O
0.3 0.4 ' 0_.% O.6 0.7 0.8
KT

Fig. I. Long term monthly results for UL(T, - T,~)/(m) vs KT.

1.C

O.8 }
Cf

O.6 - - -

0.4
C : I. - 1.17xI0-~ ( S M - S ) Z ' J ~

0.2 "\
O0 10 20 30 40 50 60 70 80 90
i s M- s I DEVIATION FROM O P T I M U M TILT (DEG)

Fig. 2. Correction factor for converting to non-optimumtilts.

In essence, this term is a measure of the difference 4. T H E U S E O F T H E PROCEDURE

between the average temperature during high insolation The task of calculating the monthly electrical energy
hours of the day and the mean monthly temperature. output from max-power tracked PV arrays has now been
Typical Meteorological Year (TMY)[8] data for the reduced to evaluating the terms that appear in eqn (13)
seven sites noted previously (see Table 1) confirmed below, and eqn (4):
that[l] (To - TM)= 3 C +2C.
*/ = */r{]- fl[(Tc -- T a ) - ( T a - TM)- (TM -- T~)]
The term log,o 1 + 3' loglo I}. (13)
Equation (9) can be manipulated to give:
The problems of obtaining the monthly average daily
log;o I = X/~ log,o I~lXll (12) insolation on the array, EIJNd, are familiar to the solar
design community and will not be discussed here.
which defines Iog~o/. As was done for the term UL(Tc - Sources of data do exist, however[9, 10]. The parameters
To)/(ap), a value of loglol was calculated each month */,, T~, fl and Y are obtained from manufacturer's data. T,
using eqn (12) for the locations and number of years and */, are array and cell design dependent, while/3 and
shown in Table 1. The long term average results for y are essentially cell material dependent (e.g. for silicon
optimally tilted flat plates are shown in Fig. 3, along with fl-0.0045C-' and 7-0.12 and for cadmium sulfide
the best fit straight line and the standard deviation, or, of cells f l - 0.006 C-').
all the data (all months for all seven sites) about the The following example illustrates the use of this sim-
linear fit. There is little discernible difference for data for plified method. Consider a max-power tracked flat array
2-D tracked surfaces. No simple correction factor, such facing south, tilted up from horizontal at the local lati-
as that discussed above for correcting (To-Ta) for tude (35°N) in Albuquerque, NM (s = 0). Assume UL =
non-optimal tilts, has been found[l]. Fortunately, this 0.02kW.m-2.C -', a=0.88, p = l , ,/,=0.15 (15%),
term is usually small and does not significantly affect */. 7",=0C, fl=0.005C -1, and 3'=0. Note this cor-
558 D. L. EVANS

! i i !
o
o
-O.~

O
- Iog~01 : 0.640 - 0.732 K-I
- 0.4

M ~o

-o.~
w

n-
-O.~ o'

- 0.1 A -- AlbuquerqJe, NM
8 B]5 mar c_k, ND
M- Madison wr
0- Medford, OR
P Pnoer ix, AZ
5 -- Santa ~arba CA
V/ -- V~lshingtonlSterling, VA
~O.n'3 t t 0 .,7
0.4 ~5 0.6 0.8
gT

Fig. 3. Long term monthly results for log,o I vs Kr for seven cities using optimum tilts each month.

responds to a NOCT of 64 C at 1.0 kW. m -2, if the UL tioned above typically lowers the (To- T,), obtained by
given corresponds to 1. m. s-' wind speed. The com- neglecting the electrical output, by 2 to 4 C for con-
parison of the simplified technique with hourly simula- version efficiencies characteristic of present photovoltaic
tion will be made for the TMY month of January (KT =
0.614, T , =2C). Thus, from Fig. l, the optimum
Table 3. Comparison of simplified procedure for predicting out-
UL(Tc - T~)/(ap) is 0.73 kW" m -2. For the latitude and
put with TMY hourly computer simulation
month, the optimum tilt would be 64 °. This leads to a
The S y s t e m s
Cf = 0.90 from Fig. 2. Hence, for the non-optimum tilt: Location Albuquerque Madison Medford
Collector Flat Concen- Plat
Plate trator Plate
UL(T~ - T.)/(ap) = 0.90(0.73) = 0.66 k W " m 2. Concentration
Ratio 1 20 1
Tilt - Degrees 35 2-D I0
This leads to (T~- T,)= 0.66(1)(0.88)/0.02 = 29 C. UL - k W - m - ~ . C - I 0.02 0.01 0.02
Equation (13) then yields: a 0.88 0.88 .088
p 1.0 0.88 1.0
~ r (%) 15 15 15
~7 = 0.1511 - 0.005{(29) + (3) + (2)}] = 0.125 = 12.5 per cent. T r (C) 0 0 0
7 0 0 0
~(C -x) 0.005 0.005 0,005
The above result stems from an analysis that neglected
The R e s u l t s
the electrical output from the array in eqn (9). For many Albuquerque Madison Medford
purposes this efficiency is sufficiently accurate. However, MO. ~spm ~hcs ~spm ~hcs ~spm 1]hc$
if a more accurate W is required, an iterative procedure no. (~) (~) (%) (~) Or) (~)
based on eqn (9) can be invoked by substituting (a - rt) I 12.8 12.6 12.4 12.8 13.7 13.8
2 12.4 12.2 12.0 12.3 13.3 13.8
for a in UL(T~- T~)/(ap) to obtain a better value of 3 12.0 11.9 11.8 11.4 12.7 13.1
(T~-To) and new W. Such an iteration in this case 4 11.4 11.3 11.0 10.8 12.3 12.3
5 11.1 10.9 10.4 10.2 11.8 11.8
produces an adjusted (To - T,,) of 25 C, and r/= 0.128 or 6 10.8 10.7 9.9 10.2 11.2 11.1
12.8%. Hourly simulation for this array and location 7 10.7 10.4 9.6 9.8 10.7 10.3
8 10.6 10.5 9.7 10.1 11.0 11.4
yields an equivalent rl : 12.6%. 9 Ii.0 10.8 I0.I 10.4 11.5 11.7
10 11.3 12.2 10.9 10.8 12.3 12.3
The versatility of the procedure is shown by the com- 11 12.2 12.2 11.9 11.6 13.3 13.0
parisons with hourly simulations (using TMY data) in 12 12.6 12.6 12.5 12.2 14.1 13.9
Table 3. Yearly efficiencies were obtained by weighting yr 11.5 11.3 11.2 10.8 11.9 11.7
the monthly efficiencies with the monthly (or monthly
- calculated by S i m p l i f i e d P r o c e d u r e
average daily) array insolation. ~spm Method
This calculation process can be cast into a simpler ~bcs - calculated by H o u r l y Computer
Simulation
form when it is realized that the iteration process men-
 Simplified method for predicting photovoltaicarray output 559
NOCT !1 1.0 kWlm I & T i t 2 0 C
,o A 1 5o eo ,o 80 c

il Ill/'
NOT. "l°'llll[l['~'l'
Iltlllllq
NOCT tit 0.8 kW/m 2 & T " - 2 0 C UL/a
~01 11 [ 1 I I I I i ieO~ I l l kW/(m2"C)o.lo
' I1
i ~+ --i i i
? NOCT-T a at 0.8 kW/m 2 I L~.-I II
. . . .
.,, I I I o o6
~r ! t ! Fq~
WIND SPEED (m/s)

WIND ("~OI. ~--"

SPEED ~ _-~-.~.
(m/s) ARI
(OA SHA

,-,,,,<c> ~ # g ? o .~ e

!. : z t i . ,, ! ,o
,,,~_" /5 ] ! ]. 60
/ :t iiiiTi/JT/l , :" ]IAtF C
, . . . . . .

" ~ --'"LA~-L
~ 60

- 40 - 3 0 _20 - 10
,,,
10
IVV'I~,
20 30 40
t-i
50 60
i
70 80 90
~ i, l l~l -,~ o,
I s - S MI
D TcIC )

Fig. 4. Graph for estimating monthly average cell temperature (p = 1 assumed).

arrays (5-15 per cent). This iteration could be avoided plifications, can be used to construct the nomograph of
by, instead, neglecting the (To- TM) term in eqn (13) Fig. 4 for finding To. Entry into this graph is made either
which, fortuitously, is of the same magnitude. on axis A or axis B, depending on whether NOCT or
In essence, a monthly average cell temperature could UL/O~p is known. If entry is to be made on axis A, it
then be calculated from should be made at axis:
A1 if the NOCT is known for an insolation of i kW/m 2
T~ = (T~- T~)+ TM (14) and an ambient temperature of 20 C.
A2 if the NOCT is known for an insolation of I kW/m 2
where the term (T¢ - Ta), obtained from Fig. 1, neglects and the ambient temperature is different from 20 C. The
the electrical output in the energy balance of eqn (9). known ambient temperature should be subtracted from
Neglecting y, eqn (13) becomes the NOCT before entering this axis.
A3 if the NOCT is known for an insolation of
n = ,7,[1 - / 3 ( L - L)]. (15) 0.8 kW/m 2 and an ambient temperature of 20 C.
A4 if the NOCT is known for an insolation of
As mentioned previously, (To- To) for flat arrays is 0.8 kW/m 2 and the ambient temperature is different from
related to the NOCT and wind speed (there is also a 20 C. The known ambient temperature should be sub-
small dependence on ambient temperature which will be tracted from the NOCT before entering this axis.
neglected here). The array thermal studies of [7], along After entering axis A, one should proceed downward
with the present correlations and suggested sire- until intersecting the wind speed curve appropriate for
560 D.L. EVANS

the chosen location and month.t After the intersection a Qo, monthly average daily electrical output of photovoltaic
lateral move to the right should be made until reaching array
the K r curve for the location and month. If entry needs s array tilt up from horizontal
$M optimum s for maximum energy collection
to be made into Fig. 4 with a UtJap, one begins at axis B T~ ambient dry bulb temperature
and proceeds to the left until intersecting the appropriate Ta.Nocr ambient dry bulb temperature during NOCT test
K r curve. From the K r curve, one proceeds downward T~ monthly average cell temperature
until intersecting the appropriate sloping ] s - sM] curves TM mean monthly ambient temperature
T, reference temperature for cell efficiency
(labeled along axis C). A lateral move to the left then is UL loss coefficient or thermal conductance per aperture
made to reach the appropriate mean monthly tem- area between the cells and the ambient
perature, TM, curve for the month and site. A movement WS wind speed
downward to the horizontal axis D results in the monthly a solar absorptance of array
average cell temperature, Tc, that can be used in eqn temperature coefficient for cell efficiency
intensity coefficient for cell efficiency
(15). Y
7/ array monthly average conversion efficiency
For the Albuquerque example used above, Fig. 4 nl array hourly conversion efficiency
yields Tc = 29 C and rt = 0.15 [1-0.005(29-0)] = 12.8% for n, array eft. at Tc = T, and/i = 1. kW. m-2
a wind speed of 1 m. s -~. At 3 m. s -j wind speed, Tc = 0 latitude
solar reflectance (or transmittance) of optical com-
27C and r/= 13.0%. If this same array were used in P
ponents that may be between the array and the sun
January in Madison (Kr =0.45, T~ = - 9 C , s = 0 = standard deviation
43.1 °, W S = 4 m . s-l), Fig. 4 shows Tc = 11 C resulting
in rl = 14.2%. Electrical energy production could be
obtained from eqn (4) by using appropriate array in-
solation values for Albuquerque and Madison[9, 10]. REFERENCES
1. D. L. Evans, W. A. Facinelli and L. P. Koehler, Simulation
and simplified design studies of photovoltaic systems. Rep.
tNote there are two sets of wind speed curves in Fig. 4; the SAND 80-7013. Sandia Laboratories, Albuquerque (1980).
solid curves represent arrays that are open on both sides (e.g. 2. Photovoltaic systems concept study. Rep. AL0-2748-12.
free standing arrays with no thermal insulation) and the dashed Spectrolab, Inc., Sylmar, CA (1977).
curves represent arrays that are well insulated on their back sides 3. D. L. Evans and L. W. Florschuetz, Cost studies on ter-
(e.g. integral roof mounting). It is imperative that the mounting restrial photovoltaic power systems with sunlight concen-
and testing of the arrays during NOCT tests be equivalent to the tration. Solar Energy 19, 255 (1977).
way they will be mounted and used in the actual application. The 4. L. W. Florschuetz, On heat rejection from terrestrial solar
free standing NOCT's should not be used for arrays that will be cell arrays with sunlight concentration. Proc. l lth IEEE PV
insulated on the back and vice versa. Specialists' Con[erence, Scottsdale, AZ (1975).
5. SOLMET User's Manual. Hourly surface radiation-surface
meteorological observations. National Oceanic and Atmos-
Acknowledgement--This work was supported by the U.S. pheric Administration Environmental Data Service, Ashville,
Department of Energy through Sandia Laboratories contract NC (1979).
13-0313. 6. B. Y. H. Liu and R. C. Jordan, The long-term average
performance of flat-plate solar-energy collectors. Solar
NOMENCLATURE Energy 2, 53 (1%3).
A array area 7. J. W. Stultz and L. C. Wen, Low-cost silicon solar array
G factor for flat array tilt correction project. Rep. 5101-31. Jet Propulsion Laboratory, Pasadena
Gr,NOCT instantaneous total solar irradiation on the array during (1977).
NOCT test (1.0 or 0.8 kW. m-2) 8. I. J. Hall, R. R. Prairie, H. E. Anderson and E. C. Boes,
i an hourly subscript Generation of a typical meteorological year. Proc. 1978 Ann.
hourly beam (direct normal) solar radiation incident on Meeting Am. Sec. o[ ISES, Denver (1978).
array 9. S. A. Klein, W. A. Beckman and J. A. Duffle, Monthly
I, hourly solar energy incident on array average solar radiation on inclined surfaces for 261 North
IT hourly total (beam, diffuse and reflected) solar radia- American cities. Rep. 44-2. Solar Energy Laboratory, Uni-
tion incident on array versity of Wisconsin, Madison (1978).
K~ ratio of monthly total radiation on horizontal to mon- 10. D. L. Evans, W. A. Facinelli and L. P. Koehler, Simplified
thly extraterrestrial radiation (a clearness number) design guide for estimating photovoltaic flat array and sys-
Nd number of days per month tem performance. Rep. SAND 80-7185. Sandia Laboratories,
NOCT nominal operating cell temperature Albuquerque (1980).