You are on page 1of 8

Available online at www.sciencedirect.

com

Applied Thermal Engineering 28 (2008) 801–808


www.elsevier.com/locate/apthermeng

Review

A survey of wind convection coefficient correlations


for building envelope energy systems’ modeling
J.A. Palyvos *
Solar Engineering Unit, School of Chemical Engineering, National Technical University of Athens, Greece GR-15780

Received 30 September 2007; accepted 6 December 2007


Available online 15 December 2007

Abstract

The thermal losses to the ambient from a building surface or a roof mounted solar collector represent an important portion of the
overall energy balance and depend heavily on the wind induced convection. In an effort to help designers make better use of the available
correlations in the literature for the external convection coefficients due to the wind, a critical discussion and a suitable tabulation is
presented, on the basis of algebraic form of the coefficients and their dependence upon characteristic length and wind direction, in addi-
tion to wind speed. Finally, simple average correlations are produced from the existing ones, useful for quick, gross estimates.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: External convective losses; Wind loss coefficient; Forced convection heat loss; Review of heat convective coefficients

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 801
2. The traditional correlation for the wind loss coefficient and its variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 802
3. Boundary layer type correlations for the wind loss coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804
4. Correlations for the wind loss coefficient explicit in V, L, and/or the wind direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806
5. External convection algorithms used in building simulation programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806
6. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 807
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 807

1. Introduction ing/cover heat losses were in the range 26.5–40.0% of the


incident solar radiation, depending on duct geometry,
It is well documented that the thermal losses from ambient temperature, solar energy flux, and wind velocity.
external building surfaces and such solar components as A more pronounced ‘‘cover” heat loss to the ambient
collectors, chimneys, and ventilated photovoltaic arrays, was recorded in the case of a ventilated photovoltaic array,
constitute a large portion of the respective energy balance. situated close to the chimney and having the same dimen-
In support of this statement, recent calculations and mid- sions and tilt angle.
winter data regarding a tilted solar chimney mounted on Major parameter affecting the usual modeling of losses
an NTUA campus building’s rooftop showed that glaz- to the ambient from such building envelope related compo-
nents is the wind convection coefficient, hw, a quantity not
*
Tel.: +30 210 7723297; fax: +30 210 7723298. very well documented, improper use of which can easily
E-mail address: jpalyvos@chemeng.ntua.gr cause 20–40% errors in energy demand calculations [1].

1359-4311/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2007.12.005
802 J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808

Nomenclature

A surface area of plate (m2) Vaz wind speed at height z (m/s)


C circumference of plate (m) Vo wind speed at standard conditions (m/s)
cp heat capacity of air (J/kg K) Vw wind speed at monitored surface (m/s)
hw forced convection coefficient due to wind (W/ Vf free stream wind speed (10 m above roof) (m/s)
m2 K) w wind velocity component in z-direction (m/s)
j Colburn j-factor, dim/less (=St Pr2/3) Wf wind modifier, dim/less (1 for windward, 0.5 for
k thermal conductivity of air (W/m K) leeward)
L length (m) x distance from leading edge of plate (m)
L1, L2 dimensions of plate (m) z height of monitored wall above ground (m)
LR characteristic length for Re (=4A/C) (m) z0 height at which Vf measurements are taken (m)
l length along wind direction (m)
Nu Nusselt Number, hL/k (=St Re Pr) Greek letters
Nux local Nusselt Number a angle of attack (°)
NuL average (over L) Nusselt Number a, b terrain-dependent coefficients, dim/less
Pr Prandtl Number, cpl/k h wind direction (°), + east of North
Re Reynolds Number, LVq/l (=LV/m) l shear viscosity of air (N s/m2)
Rec critical Reynolds Number m kinematic viscosity of air (m2/s)
Rf surface roughness factor-multiplier q density of air (kg/m3)
St Stanton Number, h/qcpVf (=Nu/RePr) u yaw angle (°)
Ta ambient temperature (K, °C in Eq. (1))
u wind velocity component in x-direction (m/s) Subscripts
v wind velocity component in y-direction (m/s) L average over the length L
V wind speed (m/s) x local value

Compared to radiation losses, on the other hand, external  wind tunnel measurements and model studies on rela-
convection losses are 3–4 times as big [2–4]. And since in tively small plates and bluff bodies-obstacles (cf. [19–
many situations there has been enough skepticism toward 21]);
the ‘‘standard computational correlation” for the wind loss  full-scale/field data recordings, i.e. measurements on
coefficient, namely that of Nusselt–Jürges [5], a plethora of actual building facades and roofs (cf. [22–26]).
analogous correlations have appeared in the literature in
recent years. (One of the early reviews on the subject is that Thus, the prospective designer/modeler must be aware of
of Cole and Sturrock [6].) the diversity of the available correlations and must make
A discussion of the ‘‘pros” and ‘‘cons” of using a simple sure that he has examined the specific conditions under
linear correlation such as the Nusselt–Jürges one, should be which they have been produced, before he can safely use
preceded by a rough categorization of the various expres- them. After all, there is an obvious lack of physical equiva-
sions that have been used so far. On the basis of parametric lence between easily controlled indoor experimental studies
dependence, the wind convection loss coefficient for an and the hard reality of the field. Out there, monitoring the
external flat surface or bluff body has appeared in the liter- wind or establishing uniform conditions for the relevant
ature as: measurements is a formidable job. It has been reported,
for example, that average wind speed measurements in the
 an experimentally determined constant value, usually proximity of an external surface such as a solar collector
given on tables (cf. [7–10]); cannot have less than a ±0.5 m/s variation [27].
 a very simple expression, usually linear or power law
function of the wind speed (cf. [9,11–13]); 2. The traditional correlation for the wind loss coefficient and
 an expression involving, in addition, a characteristic its variants
length of the surface in question, either explicitly or
implicitly, i.e. via the Reynolds Number (cf. [14–17]); In its most general form, the Nusselt–Jürges correlation
 an analogous relation which also takes into account the between the wind convection coefficient, hw, and the paral-
wind direction with respect to the surface (cf. [1,3,18]). lel to the surface component of the wind velocity, Vw,
which drives the convection can be written as
On the basis of the various test rigs, whose data were     n 
used to produce the correlations for the wind loss coeffi- 294:26
hw ¼ 5:678 a þ b V w 0:3048 ð1Þ
cient, there are expressions stemming from: 273:16 þ T a
J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808 803

Table 1 Table 3
Values for the constants in Eq. (1) [28,29] Relation of the parallel component of the wind velocity to the free stream
value [22]
Vw < 4.88 m/s 4.88 6 Vw < 30.48 m/s
Windward Leeward
Surface texture a b n a b n
Vf > 2 m/s Vw = 0.25Vf Vw = 0.3 + 0.05Vf
Smooth 0.99 0.21 1 0 0.50 0.78
Rough 1.09 0.23 1 0 0.53 0.78 Vf < 2 m/s Vw = 0.5

where a, b, n are empirical constants, and Ta the ambient generated the above equations were taken on a vertical
temperature in °C. The correction in the innermost paren- square copper plate with 0.5 m sides subjected to parallel
thesis is dictated by the fact that the original correlation flow of air, a situation that is hard to meet in real life.
was developed for a reference temperature of 294.26 K In order to circumvent the lack of proper values for Vw,
(21.1 °C) [28,29]. (This temperature correction is equivalent Ito et al. [22] tried to correlate this parallel to the surface
to a density correction, which is necessary since the mass component with the free stream velocity, Vf, introducing,
velocity rather than the linear velocity is more appropriate at the same time, gross wind direction, i.e. differentiat-
for the description of forced convection [30]). The three ing the results for windward1 and leeward conditions (cf.
constants in Eq. (1), which take values that depend on Table 3). This procedure, which at the time was also
the external surface texture and the wind speed, are listed adopted by ASHRAE [37], allows the use of a single corre-
in Table 1, giving hw in SI units. lation for the convection component of the outside heat
Actually, the original 1922 global correlation of Nus- transfer coefficient for either direction, namely
selt–Jürges [5] based on their copper plate data, if written
hw ¼ 18:63V 0:605
w ð5Þ
for SI units would have the form
hw ¼ 7:13V 0:78 0:6V w This power law equation (which, aside from the S.I. units,
w þ 5:35e ð2Þ
was actually derived by Kimura on the basis of earlier data
that is, it includes a decay term. This equation was special- of his Japanese group [37, p. 78]), as well as the linear Eqs.
ized two years later by Jürges [31] for three types of sur- (3) and (4) have been adopted as algebraic forms by most
faces (cf. Table 2). However, it was the original authors researchers for their own data fitting procedures. An exten-
who first suggested that, for practical calculations, it is suf- sive but not exhaustive tabulation of such equations are gi-
ficient to use a linear interpolation formula for wind speeds ven in Tables 4 and 5.
up to 5 m/s and to ignore the second term for higher speeds It turned out that in many cases the linear regression
[5]. In SI units, the proposed original linear equation would equations were equally effective in fitting the experimental
be data, even though fundamental heat transfer theory pre-
hw ¼ 5:8 þ 3:95V w ð3Þ dicts a power relation between convective coefficient and
wind speed. On the basis of data generated by thirty such
having constants which are very close to the values used in linear correlations listed in Table 4 for windward surfaces,
recent years on the basis of the McAdams [28] recast of Eq. a purely empirical ‘‘average” correlation of this simple but
(1), which for smooth surfaces is convenient form can be derived, namely
hw ¼ 5:7 þ 3:8V w ð4Þ
hw ¼ 7:4 þ 4:0V f ðwindwardÞ ð6Þ
This last and much quoted correlation has been widely
used in modeling, simulations, and relevant calculations In the wind velocity range 0–4.5 m/s, the maximum devia-
(cf. [32–34]), in spite of its shortcomings. It has been ar- tions of the individual windward equations’ predictions
gued, for example, that this dimensional equation includes from those of Eq. (6) average to 18%. A similar ‘‘average”
radiation loss in addition to convection [35], and that the correlation can be derived for leeward surfaces, using the
average-across the surface-wind speed as well as its direc- remaining six correlations of Table 4
tion must be considered [36]. Moreover, the data which
hw ¼ 4:2 þ 3:5V f ðleewardÞ ð7Þ

Table 2 In the same wind velocity range, the maximum deviations


Values for the constants in Eq. (1) based on the original correlations of of the individual leeward equations’ predictions from those
Jürges [31], ignoring the decay terma
of Eq. (7) average to 22%.
Vw < 4.88 m/s 4.88 6 Vw < 30.48 m/s It should be noted at this point that many of the individ-
Surface texture a b n a b n ual correlations listed in Tables 4 and 5 use constants with
Hydraulically smooth 0.973 0.214 1 0 0.499 0.775
Rolled 1.005 0.214 1 0 0.497 0.780
Very rough 1.087 0.226 1 0 0.522 0.784 1
A surface or data are classified as windward if the angle of incidence
a 0.6 V
The decay term is ae , where a takes the values 5.12, 5.35, and 5.84 between the normal to the monitored surface and the wind direction is less
for the three surface types, respectively. than ±90° and leeward for all other directions [23].
804 J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808

Table 4
Linear equation form for the wind convection coefficient (W/m2 K): hw = a + bV
a b Comments References
5.8 3.95 Wind tunnel measurement (WTM), plate, parallel flow, Vw < 5 m/s [5]
5.7 3.8 WTM, plate, parallel flow, Vw < 5 m/s, smooth surfaceb [28]
6.2 4.3 WTM, plate, parallel flow, Vw < 5 m/s, rough surface [4]
6.05 4.08 WTM, vertical plate, rough surface, V 6 5 m/s [30]
5.82 4.02 WTM, vertical plate, rolled surface, V 6 5 m/s [30]
7.82 3.50 Very smooth surface, no speed limit [39]
8.9 3.71 Smooth wood, plaster, no speed limit [39]
10.7 4.96 Cast concrete and smooth brick, no speed limit [39]
23 5.7 WTM, exposed face of a 23 cm cube for the range 3–10 m/s [6]
11.4 5.7 Nocturnal field measurements (NFM) on heated strips, exposed surface [6]
5.7 6.0 NFM on heated strips, normal surface [40]
0 5.7 NFM on heated strips, leeward facing surfaces [6]
5.8 2.9 NFM, wall, Vf > 3 m/s, windward (if leeward and Vf > 4 m/s, hw = 13 W/m2 K) [22]
8.7 9.4 NFM, Vw > 4 m/s, leeward, h independent of wind direction [22]
2.8 3.0 Vw, revised Ref. [5] data to exclude radiation and free convection contribution [35]
6.22 1.824 Kimura’s ‘‘4th floor model”, field measurement (FM) on window, windward surface [41]
6.22 0.4864 Kimura’s ‘‘4th floor model”, FM on window, leeward surface [41]
7.55 4.35 NFM in a Canadian Arctic location, window, Vf rooftop speed [42]
5.8 4.1 Based on Ref. [5] data, Vw (for roofs design V’s are 1, 3, and 9 m/s) [43]
4.5 2.9 Smooth surfaces (glass, paint) at ordinary temperatures (rough: 50% higher) [11]
8.55 2.56 Rectangular plate exposed to varying wind directions, Vbar sqrt(u2 + v2 + w2) [44]
0.036 2.2 Laboratory measurements, inclined and yawed real collector, Vw, leeward [3]
5.1 1.7 FM on facßade of tall building, Vw = f1(Vf) = 1.8Vf + 0.2 windward [23]
5.1 1.7 FM on facßade of tall building, Vw = f2(Vf) = 0.2Vf + 1.7 leeward [23]
7.0 2.1 Flat plate PV module, experimental, 1.0 < Vw < 1.5 m/s, 0° < Ta < 35 ° [65]
6.47 6.806 ASHRAE/DOE-2 model, rough surfacesa – excluding radiation [45]
4.955 1.444 Daytime FM on central region of vertical wall, Vf, plate shielded from sun [13]
8.91 2.00 FM on plane, smooth facßade test surface, Vf (Vw = 0.68Vf  0.5), windward [24]
4.93 1.77 FM on plane, smooth facßade test surface, Vf (Vw = 0.68Vf  0.5), leeward [24]
10.03 4.687 Indoor laboratory measurements on box-type solar cooker [12]
12.2 6.548 Indoor laboratory measurements on basin-type solar still [12]
8.3 2.2 FM, collector mimic on 35° pitched roof, Vf, incidence angle i = 0° [46]
6.5 3.3 FM, collector mimic on 35° pitched roof, Vf, incidence angle i = 90° [46]
6.42 3.96 Multipoint FM, V = sqrt(avg(u2 + v2 + w2)), developed turbulent boundary layer on horizontal roof [26]
4.47 10.21 Multipoint FM, V = sqrt(avg(u2 + v2 + w2)), on vertical wall [26]
4.214 3.575 Collector glass cover [9]
5.82 4.07 Based on photovoltaic systems’ analysis [67]
5.5 2.2 Outdoor measurements on PV modules without considering wind direction [68]
a
For smooth surfaces hw = 3.12 + 3.83V  0.047 V2 – excluding the constant 5.11 W/m2 K radiation contribution. A recent similar equation is
hw = 12.667 + 1.5946V + 0.0041V2 [63].
b
ASHRAE proposes a = 5.62, b = 3.9 ([60], p. 3.14). Another variant is a = 5.67, b = 3.86 [64].

more significant figures than one would expect to obtain j-factor, expressing the analogy between (sublimation)
during a wind speed measurement in the field. Also, small mass transfer and (convection) heat transfer. Using the rel-
differences in the values of the constants in the Nusselt–Jür- evant definitions, j = St Pr1/3, and St ¼ qcphV f ¼ RePr
Nu
, where
ges equations, as reported by various authors, could be the St is the Stanton Number, Pr the Prandtl Number, q the
result of units-conversion errors which probably have been density, and cp the heat capacity of air, respectively, the
propagated over the years by indirect referencing. average Nusselt number, NuL , on an inclined rectangular
plate subject to an oncoming airflow can be written as
3. Boundary layer type correlations for the wind loss NuL ¼ 0:86Re1=2 Pr1=3 ð20:000 < Re < 90:000Þ ð8Þ
coefficient
In this correlation, and in similar ones involving square
Thermal boundary layer theory has led to correlations plates or plates with stabilizing extensions (cf. Table 6),
for hw or, equivalently, for the Nusselt Number, Nu, which the characteristic length for Re is LR = 4A/C, where A is
implicitly involve a characteristic length via the Reynolds the plate area and C its circumference. If the plate is rect-
Number, Re, which also appears in the correlations. In a angular, with sides L1 and L2, then
series of wind tunnel experiments on naphthalene plates,
Sparrow and his group [18,19,15,38] have produced corre- 2L1 L2
LR ¼ ð9Þ
lations of the form j = aReb, where j is the familiar Colburn L1 þ L2
J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808 805

Table 5
Power law form for the wind convection coefficient (W/m2 K): hw = a + bVn
a b n Comments References
a
0 7.13 0.78 Wind tunnel measurement (WTM), plate, parallel flow, 5 < Vw < 24 m/s [5]
0 7.11 0.775 WTM, plate, parallel flow, hydraulically smooth surface, 5 < Vw < 24 m/s [31]
0 7.52 0.784 WTM, plate, parallel flow, very rough surface, 5 < Vw < 24 m/s [31]
0 7.2 0.78 Recast of data in [5,31], V > 5 m/s, smooth surface [28]
0 7.6 0.78 Recast of data in [5,31], V > 5 m/s, rough surface [28]
0 6.97 0.666 WTM, flat plate, measurement on the rear [47]
0 6.60 0.6 Vertical surface behind wedge-separated subsonic flow [48]
0 18.65 0.605 Field measurements (FM), Vw–Kimura’s ‘‘6th floor model”b [37]
0.685 11.8 0.5 WTM, small (0.16 m) collector mimic [49]
0 2.38 0.89 FM, window, low-rise building, forced convection only, windward (MoWiTT)c [45]
0 2.86 0.617 FM, window, low-rise building, forced convection only, leeward (MoWiTT)c [45]
0 16.15 0.397 FM, flat vertical panel, windward, Vw = f(Vf) = 0.68Vf  0.5 and 0.2Vf  0.1 [24]
0 16.25 0.503 FM, flat vertical panel, Leeward, Vw = f(Vf) = 0.157Vf  0.027 [24]
0 16.21 0.452 FM avg correlation for combined windward and leeward conditions [24]
0 14.82 0.42 FM on a 6th floor vertical surface in 200 mm recess, windwardd [50]
0 15.06 0.53 FM on a 6th floor vertical surface in 200 mm recess, leewardd [50]
0 9.4 0.5 FM, collector mimic on 35° pitched roof, Vf [46]
18.192 0.0378Tav 0.8 External coefficient in large commercial ducts, Tav = (Tduct,surf + Texterior)/2 [66]
Note. Ref. [69] examines convective cooling of photovoltaics.
a
ASHRAE proposes b = 7.2 for 5 6 Vw 6 30 ([60], p. 3.14).
b
Vw = 0.25Vf for Vf > 2 m/s, Vw = 0.5 for Vf 6 2 m/s (windward) and Vw = 0.3 + 0.05Vf (leeward) [45].
c
Same for DOE-2 with an additional multiplier for rough surfaces, Rf.
d
Also: correlations for 4 more depths and alternate correlations involving Vf instead of Vw.

Table 6
Boundary layer equation form for the wind convection coefficient (W/m2K): Nu = a RebPrc + d
a b c d Comments References
0.10 0.666 0 0 Wind tunnel measurement (WTM), vertical plate, windward [47]
0.20 0.666 0 0 WTM, vertical plate, leeward [47]
0.42 0.6 0 0 WTM, considers house as equivalent sphere [14]
0.931 0.5 0.333 0 WTM, global correlation for inclined (attack) and yawed square plate [18]
0.86 0.5 0.333 0 WTM, global correlation for inclined rectangular plate of finite width [19]
0.930 0.5 0.333 0 WTM, global correlation for pitch and yaw-square plate [38]
0.86 0.5 0.333 0 WTM, global correlation for collector with extension surfacesa [15]
0.0253 0.8 0 3 WTM, local Nu equation for mixed (lam. and turbulent) flow over a flat plate [20]
0.036 0.8 0.333 f1(Pr)b Laboratory, avg Nu on inclined and yawed real collector, with L = L(L1, L2, u)e [3]
0.032 0.8 0 84.5 Laboratory, avg Nu on inclined and yawed real collector (Pr = 0.706) [3]
1.23 0.5 0.333 0 WTM, plate flush on wooden roof model, angle of attack a < 40° [51]
0.90 0.5 0.333 0 WTM, plate flush on wooden roof model, angle of attack a P 40° [51]
0.568 0.524 0 0 WTM, flat-plate model collector flush on 30° roof of model house [52]
1.067 0.466 0 0 WTM, flat-plate model collector flush on 45° roof of model house [52]
f2(Re, Pr)c 0.8 1 0 Nu for parallel to plate flow, turbulent boundary layer [9]
0.023 0.891 0 0 Multipoint field measurement (FM), Re = (l/m)sqrt(avg(u2 + v2 + w2)), developed turbulent [26]
boundary layer on horizontal roof
0.0296Rf 0.8 0.333 0 FM, local Nu for horizontal roof, turbulent flow – Rf surface roughness multiplier [25]
0.037Rf 0.8 0.333 0 FM, avg Nu, horizontal strip of length L, turbulent flow, DT = 0, L > xc  0 [25]
0.664Rf 0.5 0.333 0 FM, avg Nu, horizontal strip of length L, turbulent flow, DT < 0, L > xc = 5  105l/(qVf) [25]
0.037Rf 0.8 0.333 f3(Pr)d FM, avg Nu, horizontal strip of length L, turbulent flow, DT < 0, L P xc = 5  105l/(qVf) [25]
a
The hydrodynamic dimensions are used for the characteristic length in Re.
b
f1(Pr) = 95.0Pr1/3.
c
f2(Re, Pr) = 0.037/(1 + 2.443Re0.1(Pr2/3  1).
d
f3(Pr) = 871RfPr1/3.
e
The characteristic length is L = L1L2/(L1 cos u + L2 sin u), with L1, L2 the collector side lengths and u the yaw angle.

The choice of such a length was dictated by the intuitive Another experimental series involving a real (albeit half-
involvement of the surface area of the plate, and the very size) solar collector under controlled environmental condi-
high transfer rates observed at the edges, thus bringing also tions in the laboratory [3], produced a slightly different cor-
the circumference into the picture [19]. Eq. (9) is nothing relation for the average Nusselt Number, namely,
more than the simplest combination which yields a length 4=5
dimension. NuL ¼ 0:036ReL Pr1=3  95Pr1=3 ð10Þ
806 J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808

The critical Reynolds Number was estimated at 70,000, evaluation of the thermo-physical properties which partic-
implying an early development of turbulence which can ipate in the Reynolds and Prandtl Numbers, for specific
be explained by the obvious lack of a sharp leading edge temperatures, although the latter are not always reported
on the collector. As the latter was both inclined and yawed (cf. [17]). They have the general form hw = aVb/Lc, that
with respect to the on-coming air, the appropriate charac- is, they include a decay of the wind convection coefficient
teristic length for Re was along the surface in the direction of the wind. For example,
L1 L2 the fully turbulent flow convective coefficient can be writ-
L¼ ðu < 90 Þ ð11Þ ten as
L1 cos / þ L2 sin /
hw ¼ 5:74V 4=5 L1=5 ð13Þ
with a similar equation for yaw angles, u, greater than 90°
[3]. It turned out, however, that, as in the case of Sparrow’s for a flat surface such as a solar collector [17]. If the latter is
yawed plate [18], the influence of the yaw angle was small, flush-mounted on the (inclined) roof of a house, then
over the entire range of u’s tried (0, 90, 135, and 180°), but  
8:6V 0:6
not insignificant. The relevant reduction of hw with u was hw ¼ max 5; 0:4 ð14Þ
in the range 5–15% [3]. On the other hand, the wind direc- L
tion has also been found not to have a significant effect in where L is the cube root of the house volume in meters [32,
the case of large walls [13]. p. 166]. The constant 5 in this last relation represents the
Table 6 includes additional boundary layer type correla- minimal hw value which is observed in solar collectors un-
tions. Among them, is the following equation for the local der zero wind.
Nusselt Number, Nux, The rest of the correlations in Table 7 are explicit only in
Nux ¼ 0:037Rf Re4=5 Pr1=3  871Rf Pr1=3 ð12Þ V, having the form hw = aVb, i.e. without the decay factor.
Each one is given for either wall surfaces or roofs of low-
in which x denotes distance from the leading edge and Rf is rise isolated buildings, and for a particular wind direction.
a surface roughness dependent convection multiplier, As an example, for a ±45° surface-to-wind angle the rele-
assuming values in the range 1.00 < Rf < 2.10. (The similar- vant equation for walls is
ity between Eqs. (10) and (12) is worth noting.) The same
hw ¼ 3:34V 0:84
f ð15Þ
authors also give relations involving explicitly the wind
incidence angle, h, for the forced convection heat transfer according to recent CFD calculations [1].
coefficient averaged over rectangular roof strips [25].
5. External convection algorithms used in building simulation
4. Correlations for the wind loss coefficient explicit in V, L, programs
and/or the wind direction
In view of the lack of a universally acceptable wind con-
A number of correlations for hw, explicit in V and L are vection coefficient or correlation (as clearly demonstrated
also listed in Table 7. Some of them have been the result of by the large, yet non-exhaustive, compilations of Tables

Table 7
Explicit in V, L form for the wind convection coefficient (W/m2 K): hw = aVbLc
a b c Comments References
8.6 0.6 0.4 Convection over buildings, L = cube root of building volume [14]
5.79 0.8 0.2 Single flat-plate PV panel, Vw P 0.3 m/s, L = ‘geometric scale’ [65]
2.537WfRfa 0.5 0.5 BLAST model: V = Vaz, wind speed modified for height zb [16]
2.537WfRfa 0.5 0.5 TARP model:V = Vaz, wind speed modified for height zc [16]
5.74 0.8 0.2 Fully turbulent flow over horizontal, constant temperature surface [17]
5.1 0.5 0.5 Field measurements (FM), plates, global, 313 K mean plate-air temperature [46]
11.42 0.891 0.109 FM, local h, V = sqrt(avg(u2 + v2 + w2)), both for horizontal and vertical enveloped [26]
5.15 0.81 0 FM, walls of isolated, low-rise building, 0° angle of attacke [1]
3.34 0.84 0 FM, walls of isolated, low-rise building, ±45° angle of attacke [1]
4.78 0.71 0 FM, walls of isolated, low-rise building, ±90° angle of attacke [1]
4.05 0.77 0 FM, walls of isolated, low-rise building, ±135° angle of attacke [1]
3.54 0.76 0 FM, walls of isolated, low-rise building, ±180° angle of attacke [1]
5.11 0.78 0 FM, roof of isolated, low-rise building, 0° angle of attacke [1]
4.60 0.79 0 FM, roof of isolated, low-rise building, ±45° angle of attacke [1]
3.67 0.85 0 FM, roof of isolated, low-rise building, ±90° angle of attacke [1]
a
Wf = wind modifier (1 for windward surfaces, 0.5 for leeward ones), Rf = surface roughness multiplier, L = (surface area/perimeter).
b
Vaz = V0(z/z0)1/a (z0 = 9.14, a is a terrain-dependent coefficient).
c
Vaz = V0b(z/z0)a (a, b are terrain-dependent coefficients).
d
Pr has been evaluated at 293 K.
e
DT = surface-to-air temperature difference = 10 K, wind speed 1–5 m/s.
J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808 807

4–7), users of large building energy simulation programs a quick gross estimate, one can always use the average
such as EnergyPlus [53], TRNSYS [54], and ESP-r [55], simple correlations proposed in this work, i.e. Eqs. (6)
should be able to choose the most suitable form offered and (7).
by the respective program. In fact, they should be also In view of the above, the expressions from the literature
given the option to supply ‘‘their own” model, as the cor- which are listed in Tables 4–7 should prove useful, pro-
relations for the external convection coefficient adopted vided the reader has access to the original works cited
by such programs are not standardized [56]. therein in order to assess the respective applicability to
EnergyPlus, for example, offers six ‘‘outside convection his/her own problem at hand. However, the obvious lack
algorithms: a simple second degree polynomial in Vw pro- of generality of the existing wind convection coefficient cor-
posed by ASHRAE [8] (which, however, includes a con- relations presents a challenge for future research. More
stant radiation component of about 5 W/m2 K), the more realism is needed, i.e. field rather than laboratory measure-
detailed algorithm of Sparrow et al. [19] as well as similar ments, as well as some sort of standardization in the choice
expressions in the BLAST and TARP programs [16] and, of such things as the wind velocity sensors or the measure-
finally, the MoWiTT and the DOE-2 models [45]. The ment topology, e.g. height above ground and/or distance
MoWiTT algorithm offers a reasonable balance between from the facßade wall or roof, etc. In this way, a much smal-
accuracy and ease of use [57] while the DOE-2 model is a ler set of well proven and generally accepted correlations
combination of the MoWiTT and BLAST algorithms may emerge, which will greatly help the designer/modeler.
(see Tables 4, 5 and 7 for the respective expressions).
Most of the relevant components in TRNSYS, on the References
other hand, ask the user to supply the heat transfer coeffi-
cient (although the linear form of Eq. (4) is built in, for [1] M.G. Emmel, M.O. Abadie, N. Mendes, New external convective
example, in the Type 19 subroutine). The most popular heat transfer coefficient correlations for isolated low-rise buildings,
algorithm used for the external convective coefficient is Energ. Build. 39 (2007) 335–342.
the quadratic polynomial in Vw [58,59], whose constant [2] P.I. Cooper, The effect of inclination on the heat loss from flat-plate
solar collectors, Solar Energy 27 (5) (1981) 413–420.
coefficients are tabulated for six types of surface texture [3] G. Thomaidis, J.A. Palyvos, N.G. Koumoutsos, The influence of the
[60]. wind on the efficiency of the flat-plate solar collector, in: Proceedings
Finally, external convection in ESP-r is modeled via of the 1st National Conference on ‘‘Soft Energies”, Institute of Solar
time varying wind driven convection coefficients, calculated Engineering, Salonica, Greece, 20–22 October, 1982, HEX, pp.1–15
(in Greek).
with a choice of two different correlations: a simple linear
[4] M.G. Davies, Build. Heat Transfer, John Wiley & Sons Ltd.,
one (cf. [35]), with suitable amendment of the velocity value Chichester, England, 2004, § 5.7.
on the basis of surface orientation and wind direction rela- [5] W. Nusselt, W. Jürges, Die Kühlung einer ebenen Wand durch einen
tive to the surface [61], and the MoWiTT model [45]. (The Luftstrom (The cooling of a plane wall by an air flow), Gesundheits
ESP-r repertoire, however, is much more flexible when it Ingenieur 52. Heft, 45. Jahrgang, (30 Dezember 1922) pp. 641–642.
comes to internal building surfaces, for which it offers an [6] R.J. Cole, N.S. Sturrock, The convective heat exchange at the
external surface of buildings, Build. Environ. 12 (4) (1977) 207–214.
option of five empirical correlations and fixed values for [7] ASHRAE, Cooling and heating load calculation manual, GRP 158,
the internal convection coefficient [62]). ASHRAE, New York, second printing 1979, p. 3.12.
[8] ASHRAE, Handbook of Fundamentals, Atlanta, GA, USA, 1993.
6. Conclusions [9] U. Eicker, Solar Technologies for Buildings, John Wiley & Sons Ltd.,
Chichester, England, 2003, § 3.1.10.4.
[10] P. Konttinen, T. Carlsson, P. Lund, T. Lehtinen, Estimating thermal
Since the importance of the external convection coeffi- stress in BIPV modules, Int. J. Energ. Res. 30 (2006) 1264–1277.
cient due to the wind in, practically, all the energy calcula- [11] P.J. Lunde, Solar Thermal Engineering, John Wiley & Sons, New
tions involving the building envelope has been well York, USA, 1980, p. 17.
documented, the prospective designer/modeler must be [12] S. Kumar, V.B. Sharma, T.C. Kandpal, S.C. Mullick, Wind induced
heat losses from outer cover of solar collectors, Renew. Energ. 10 (4)
extra careful when seeking a correlation for hw to suit
(1997) 613–616.
his/her needs. This is also true when one uses the major [13] S.E.G. Jayamaha, N.E. Wijeysundera, S.K. Chou, Measurement of
building energy analysis programs, which accept user sup- the heat transfer coefficient for walls, Build. Environ. 31 (5) (1996)
plied correlations and constant values for hw or offer a 399–407.
menu of such correlations. [14] J.W. Mitchell, Heat transfer from spheres and other animal forms,
Being aware of the diversity of the available expressions, Biophys. J. 16 (1976) 561.
[15] E.M. Sparrow, S.C. Lau, Effect of adiabatic co-planar extension
some of which are the result of laboratory tests while others surfaces on wind-related solar-collector heat transfer coefficients,
stem from actual field measurements, one must make sure Trans. ASME J. Heat Transfer 103 (1981) 268–271.
that he/she has fully examined the specific conditions under [16] T.M. McClellan, C.O. Pedersen, ASHRAE Trans. 103 (2) (1997) 469–
which they have been produced, before he/she can safely 484.
[17] E. Sartori, Convection coefficient equations for forced air flow over
use any one of them. Moreover, the understandable lack
flat surfaces, Solar Energy 80 (2006) 1063–1071.
of physical equivalence between certain experimental stud- [18] E.M. Sparrow, K.K. Tien, Forced convection heat transfer at an
ies and the reality of the field, should help in narrowing inclined and yawed square plate – application to solar collectors,
down the options. If, on the other hand, one merely seeks Trans. ASME J. Heat Transfer 99 (1977) 507–512.
808 J.A. Palyvos / Applied Thermal Engineering 28 (2008) 801–808

[19] E.M. Sparrow, J.W. Ramsey, E.A. Mass, Effect of finite width on [45] M. Yazdanian, J.H. Klems, Measurement of the exterior convective
heat transfer and fluid flow about an inclined rectangular plate, film coefficient for windows in low-rise buildings, ASHRAE Trans.
Trans. ASME J. Heat Transfer 101 (1979) 199–204. 100 (1) (1994) 1087–1096.
[20] X.A. Wang, An experimental study of mixed, forced, and free [46] S. Sharples, P.S. Charlesworth, Full-scale measurements of wind-
convection heat transfer from a horizontal flat plate to air, Trans. induced convective heat transfer from a roof-mounted flat plate solar
ASME J. Heat Transfer 104 (1982) 139–144. collector, Solar Energy 62 (1998) 69–78.
[21] J.L. Francey, J. Papaioannou, Wind-related heat losses of a flat-plate [47] H.H. Sogin, A summary of experiments on local heat transfer from
collector, Solar Energy 35 (1985) 15–19. the rear of bluff obstacles to a low speed airstream, Trans. ASME J.
[22] N. Ito, K. Kimura, J. Oka, A field experiment study on the convective Heat Transfer (1964) 200–202.
heat transfer coefficient on exterior surface of a building, ASHRAE [48] J.W. Mitchell, Base heat transfer in two-dimensional subsonic fully
Trans. 78 (1972) 184–191. separated flows, Trans. ASME J. Heat Transfer (1971)
[23] S. Sharples, Full-scale measurements of convective energy losses from 342–348.
exterior building surfaces, Build. Environ. 19 (1984) 31–39. [49] N. Onur, J.C. Hewitt Jr., A study of wind effects on collector
[24] D.L. Loveday, A.H. Taki, Convective heat transfer coefficients at a performance, in: Proceedings of the ASME Solar Energy Division,
plane surface on a full-scale building facade, Int. J. Heat Mass Century 2 Solar Energy Conference, San Francisco, CA, August 19–
Transfer 39 (8) (1996) 1729–1742. 21, 1980.
[25] R.D. Clear, L. Gartland, F.C. Winkelmann, An empirical correlation [50] A.H. Taki, D.L. Loveday, External convection coefficients for framed
for the outside convective air-film coefficient for horizontal roofs, rectangular elements on building facades, Energ. Build. 24 (1996)
Energ. Build. 35 (2003) 797–811. 147–154.
[26] A. Hagishima, J. Tanimoto, Field measurements for estimating the [51] S. Shakerin, Wind-related heat transfer coefficient for flat-plate
convective heat transfer coefficient at building surfaces, Build. solar collectors, Trans. ASME J. Solar Energy Eng. 109 (1987)
Environ. 38 (7) (2003) 873–881. 108–110.
[27] G.L. Morrison, D. Gilliaert, Unglazed solar collector performance [52] N. Onur, Forced convection heat transfer from a flat-plate model
characteristics, Trans. ASME J. Solar Energy Eng. 114 (1992) 194–200. collector on roof of a model house, Wärme und Stoffübertragung 28
[28] W.H. McAdams, Heat Transmission, third ed., McGraw-Hill (1993) 141–145.
Kogakusha, Tokyo, Japan, 1954, p. 249. [53] http://www.eere.energy.gov/buildings/energyplus/.
[29] J.A. Clarke, Energy Simulation in Building Design, second ed., [54] http://sel.me.wisc.edu/trnsys/.
Butterworth-Heinemann, Oxford, England, 2001, § 7.6.2. [55] http://www.esru.strath.ac.uk/Programs/ESP-r.htm.
[30] A. Schaak, Industrial Heat Transfer, Chapman & Hall, London, [56] S.J. Rees, D. Xiao, J.D. Spitler, An analytical verification test suite
1965. for building fabric models in whole building energy simulation
[31] W. Jürges, Der Wärmeübergang an einer ebenen Wand (Heat transfer programs, ASHRAE Trans. 108 (1) (2002) 30–42.
at a plane wall), Beiheft Nr. 19 zum ‘‘Gesundh.- Ing.” (1924), [57] F.C. McQuiton, J.D. Parker, J.D. Spitler, Heating, Ventilating, and
appearing in Gesundheits-Ingenieur 9. Heft, 48. Jahrg., 1925, p. 105. Air Conditioning, Analysis and Design, fifth ed., John Wiley & Sons
[32] J.A. Duffie, W.A. Beckman, Solar Energy Thermal Processes, third Inc., New York, 2000, p. 226.
ed., John Wiley & Sons Ltd., Hoboken, NJ, USA, 2006, § 3.15. [58] D. Bradley, TRNSYS engineer, private communication.
[33] D.Y. Goswami, F. Kreith, J.F. Kreider, Principles of Solar Engi- [59] J.S. Coventry, in: Proceedings of the 40th ANZSES Solar Energy
neering, second ed., Taylor and Francis, Philadelphia, PA, USA, Conference, 2002.
2000, p. 98. [60] ASHRAE, Handbook of Fundamentals, SI Edition, Atlanta, GA,
[34] G.N. Tiwari, Solar Energy-Fundamentals, Design, Modeling and USA, 1997.
Applications, Alpha Science International Ltd., Pangbourne, UK, [61] http://www.esru.strath.ac.uk/Programs/ESP-r_CodeDoc/esrubld/
2002, p. 74. convect2.F.html.
[35] J.H. Watmuff, W.W.S. Charters, D. Proctor, Solar and wind induced [62] ESP-r Ver. 9, Data Model Summary, § 5.6, ESRU, University of
external coefficients for solar collectors, Comples. Int. Rev. d’Hellio Strathclyde, 2001.
Tech. 2 (1977) 56. [63] L. Zhang et al., Research on system identification of wall surface heat
[36] M.V. Oliphant, Measurement of wind speed distributions across a transfer processes, Exp. Heat Transfer 15 (2002) 31–47.
solar collector, Solar Energy 24 (1980) 403–405. [64] C. Cristofari et al., Thermal modeling of a photovoltaic module, in:
[37] ASHRAE Task Group, Procedure for determining heating and cooling Proceedings of the 6th IASTED International Conference Modeling,
loads for computerizing energy calculations. Algorithms for building Simulation, and Optimization, September 11–13, 2006, Gaborone
heat transfer subroutines, ASHRAE, New York, 1975, pp. 76–78. Botswana, pp. 273–278.
[38] K.K. Tien, E.M. Sparrow, Local heat transfer and fluid flow [65] T. Schott, Operation temperatures of PV modules, in: W. Palz, F.C.
characteristics for airflow oblique or normal to a square plate, Int. Treble (Eds.), Proceedings of the 6th E.C. PV Solar Energy
J. Heat Mass Transfer 22 (1979) 349–360. Conference, London, 15–19 April 1985, D. Reidel Publ. Co., 1985,
[39] B. Jennings, Environmental Engineering, International Textbook pp. 392–396.
Company, 1970. [66] C. Wray, Duct thermal performance models for large commercial
[40] N.S. Sturrock, Localized boundary layer heat transfer from external buildings, Lawrence Berkeley National Laboratory Report, LBNL-
building surfaces, Ph.D. Thesis, University of Liverpool, 1971. 53410, 2003.
[41] K. Kimura, Scientific Basis of Air Conditioning, Applied Science [67] P. Nolay, Developpement d’une methode generale d’analyse des
Publishers Ltd., London, England, 1977. systemes photovoltaıques. MS Thesis, Ecole des Mines, Sophia-
[42] K. Nicol, The energy balance of an exterior window surface, Inuvik, Antipolis, France, 1987.
N.W.T., Canada, Build. Environ. 12 (1977) 215–219. [68] K. Furushima, Y. Nawata, M. Sadatomi, Prediction of photovoltaic
[43] CIBS, Chartered Institute of Building Services Guide Book A, Section (PV) power output considering weather effects, in: Proceedings of the
A3, CIBS, London, 1979. SOLAR 2006 – Renewable Energy, Key to Climate Recovery, July 7–
[44] F.L. Test, R.C. Lessmann, A. Johary, Heat transfer during wind flow 13 2006, Denver, Colorado, USA.
over rectangular bodies in the natural environment, Trans. ASME J. [69] L. Wen, Investigation of the effect of wind cooling on photovoltaic
Heat Transfer 103 (1981) 262–267. arrays, Report DOE/JPL-1012-69, 1982.

You might also like