You are on page 1of 12

Assessing the Maximum

Stability of the Nonconvective


Zone in a Salinity-Gradient
Solar Pond
The quality of the stability of the nonconvective zone of a salinity-gradient solar pond
A. A. Abdullah (SGSP) is investigated for an operating protocol in which the flushing procedure exactly
Department of Mathematical Sciences, compensates for evaporation losses from the solar pond and its associated evaporation
Umm Al-Qura University, pond. The mathematical model of the pond uses simplified, but accurate, constitutive
Makkah 24382, Saudi Arabia expressions for the physical properties of aqueous sodium chloride. Also, realistic bound-
ary conditions are used for the behaviors of the upper and lower convective zones
K. A. Lindsay1 (LCZs). The performance of a salinity-gradient solar pond is investigated in the context
Department of Mathematical Sciences, of the weather conditions at Makkah, Saudi Arabia, for several thickness of upper con-
Umm Al-Qura University, vective zone (UCZ) and operating temperature of the storage zone. Spectral collocation
Makkah 24382, Saudi Arabia based on Chebyshev polynomials is used to assess the quality of the stability of the pond
throughout the year in terms of the time scale for the restoration of disturbances in tem-
perature, salinity, and fluid velocity underlying the critical eigenstate. The critical eigen-
value is found to be real and negative at all times of year indicating that the steady-state
configuration of the pond is always stable, and suggesting that stationary instability
would be the anticipated mechanism of instability. Annual profiles of surface tempera-
ture, salinity, and heat extraction are constructed for various combinations for the thick-
ness of the upper convective zone and storage zone temperature.
[DOI: 10.1115/1.4036773]

Keywords: stability, salinity-gradient solar pond, flushing, heat extraction, aqueous


NaCl

1 Introduction
The terminology salinity-gradient solar pond (SGSP) refers to
a salt lake that is managed to act as a large, low cost, collector of
solar energy via the low-cost production of hot saline. This cheap
heat can be used as process heat, e.g., for aquaculture, glasshouse
food production, or industrial/municipal purposes, to produce
electricity using an organic Rankine cycle, or for the purpose of
desalination, in particular, using the technology of membrane
distillation.
A typical SGSP comprises three distinct layers (or zones), as
illustrated in Fig. 1.
Figure 1 shows an UCZ of depth dUCZ meters overlying a non-
convective zone (NCZ), or barrier layer, of depth dNCZ meters
which in turn overlies a LCZ of depth dLCZ meters, often called
the storage zone because it is here that usable thermal energy is
normally stored. The barrier zone consists of a downward salinity
gradient from almost fresh water at the UCZ/NCS interface to sat-
urated saline (s.g.  1.2) at the NCZ/LCZ interface, and plays the
dual role of thermally insulating the LCZ from the atmosphere
and simultaneously inhibiting natural convection. In the absence
of a management plan, the accumulation of concentrated saline at
the surface of a SGSP (due to evaporation) would eventually
cause the pond to overturn and become mixed thermally and in
salinity as, for example, occurs during summer with Solar Lake
[1] close to the edge of the Red Sea at Taba, Egypt.

1
Corresponding author.
Contributed by the Solar Energy Division of ASME for publication in
the JOURNAL OF SOLAR ENERGY ENGINEERING: INCLUDING WIND ENERGY AND Fig. 1 The configuration of a salt-gradient solar pond is shown
BUILDING ENERGY CONSERVATION. Manuscript received October 24, 2016; final including the primary mechanisms governing its interactions
manuscript received April 13, 2017; published online June 8, 2017. Assoc. Editor: with its surroundings. In the case of a large pond, thermal
Carlos F. M. Coimbra. losses from the retaining walls of the pond may be ignored.

Journal of Solar Energy Engineering Copyright V


C 2017 by ASME AUGUST 2017, Vol. 139 / 041010-1

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


The practical management of an operating SGSP involves the that wall boundary effects cannot be ignored unlike their counter-
regulated extraction of heat from the pond, usually but not parts in a large pond.
exclusively from the LCZ (see Ref. [2]), and simultaneously the Finally, recall that SGSPs exist to generate large quantities of
implementation of a flushing protocol for the removal of saline low-grade heat over a significant period of time. Consequently, a
from the UCZ. The conservation of salt is managed through the realistic model of a SGSP should ideally include protocols for
use of an evaporation pond in conjunction with the SGSP. Spe- both the extraction of usable heat and for surface flushing, the
cifically, saline flushed from the solar pond is discharged under latter to prevent the pond from overturning. A common theme
gravity to the evaporation pond, where it is concentrated by running through the mathematical and numerical modeling in the
evaporation and recirculated to the base of the LCZ. The rate of aforementioned articles is the omission of both factors. Valder-
flushing of the SGSP in this symbiotic system is regulated to rama et al. [5] provide a clear account of how a salinity gradient is
exactly match the combined evaporation losses of the solar and established in a SGSP, and describe a management protocol under
evaporation ponds, for example, by using an ultrasonic sensor to which the pond will retain its integrity. They specifically comment
control the flow of flushing fluid (assumed here to be fresh that over significant periods of time the only noticeable changes
water) so as to maintain a fixed depth of concentrated saline in in the structure of the pond relate to minor variations in the loca-
the evaporation pond. tion of the UCZ/NCZ interface which one anticipates are due to
Historical observations of SGSPs over many years indicate that local weather. There is no discernable evidence of instability in
the stability of a well-managed SGSP is not in doubt. Such a pond their SGSP.
resides in a state of minimum gravitational energy with the salin- The work of this article is set out as follows. Section 2 introdu-
ity gradient overpowering any natural buoyancy arising from a ces the Makkah solar pond (MSP) and provides an overview of
temperature profile that increases with depth. Moreover, stability the modeling assumptions underlying the analysis of Secs. 3–5.
is further reinforced by the presence of saline viscosity which Section 3 develops the underlying mathematical equations and
always acts to damp convection. This article investigates the qual- associated boundary conditions. Section 4 derives the stationary
ity of this stability as measured by the time scale at which pertur- solution, and Sec. 5 develops the stability analysis of the station-
bations to the pond are restored under various operational ary solution for a large SGSP. Section 6 presents results and con-
conditions of heat extraction and UCZ thickness. cludes the article.
The literature on the behavior of SGSPs has largely divided
into the simulation of the temperature and salinity profiles of the
pond (Refs. [3–7] and the references therein) and linear stability 2 The Makkah Solar Pond
analyses (Refs. [8–12] and the references therein). With the
exception of density, one theme running through a number of The analyses to follow will be applied to the Makkah solar
these articles is the common assumption that the material proper- pond (MSP) sited on the main campus of Umm Al-Qura Univer-
ties of aqueous sodium chloride are constant [3,8,9,11,12]. This is sity, Al-Abdiah, Makkah, Saudi Arabia (see Ref. [13]). The oper-
a strong assumption, not made in this article. The introduction of ating fluid of the MSP is aqueous sodium chloride. However, the
nonconstant constitutive properties for saline significantly compli- analysis to follow is general, and for this reason the operating fluid
cates any theoretical or numerical analyses, but is nevertheless an of the MSP will henceforth be called “saline”. The MSP, illus-
essential step in the drive to model reality. For example, the varia- trated in Fig. 2, is of cylindrical construction with an internal
tions in temperature and salinity within the NCZ of a SGSP filled diameter of 12 m (surface area 113 m2) and has an operating depth
with aqueous sodium chloride cause dynamic viscosity to change of 3.1 m as determined by a fixed overflow system.
in value by a factor of five to six throughout the pond. Suarez The vertical retaining wall of the MSP comprises 40 cm of
et al. [6] and Giestas et al. [7] allow dynamic viscosity to depend rockwool insulation sandwiched between an inner reinforced con-
on temperature and salinity, and Suarez et al. [6] also allows crete wall of thickness 25 cm and an outer wall of thickness
thermal conductivity and the diffusivity of salt in water to depend 10 cm. The base of the MSP consists of 25 cm of reinforced con-
on temperature and salinity. crete resting on 40 cm of load-bearing polystyrene insulation
Appropriate boundary conditions on the surface of the UCZ are which itself rests on 10 cm of reinforced concrete. The immediate
also important for maintaining realism. Giestas et al. [3], Giestas environment of the pond is fine dry sand, and importantly, the
et al. [7], Hill and Carr [8], Akrour et al. [11], and Sodha and water table at the site of the pond is deep thereby minimizing ther-
Kumar [12] apply Robin boundary conditions on the surface of mal losses from the base of the pond. Calculations based on the
the UCZ. Only Boudhiaf et al. [9] and Suarez et al. [6] use a geometry of the retaining wall and base of the MSP and the
surface boundary condition that describes the heat transfer at the
surface of the pond in terms of the mechanisms of radiation, con-
vection, and evaporation, as used in this investigation. Hill and
Carr [8] conduct a theoretical investigation of the influence of a
porous layer in controlling the onset of convection in the NCZ,
while Shi et al. [4] conduct an experimental investigation of how
the presence of a porous layer influences the diffusion of heat and
salt in a novel configuration of a SGSP. The thermal and salinity
gradients in their experimental configuration would seem to act in
opposite directions unlike a conventional SGSP where both gra-
dients act in the direction of the acceleration due to gravity.
There have been a variety of careful experimental simulations
of the behavior of a SGSP, e.g., Refs. [4] and [10] and the referen-
ces therein. Shi et al. [4] use a pond of approximate dimension
0.29 m  0.19 m and depth 0.115 m, while Karim and Jomaa [10]
use a pond with approximate dimension 0.20 m  0.02 m and
depth 0.02 m. Despite the careful configuration of these miniature
solar ponds, how to extrapolating their experimental findings to
full-scale ponds and even pilot experimental ponds requires an
appreciation of the impact of wall boundary effects which essen-
tially decay in proportion to the ratio of the perimeter of a pond to Fig. 2 The experimental SGSP at Makkah, Saudi Arabia, show-
its surface area. In a miniature solar pond, one would conjecture ing floating rings to reduce wave action

041010-2 / Vol. 139, AUGUST 2017 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


Table 1 Monthly averages of insolation ESOLAR (kWh/m2/day), decreasing function of T and strictly increasing function of S. The
daily average Celsius temperature TAVE, daily average wind behavior of a SGSP is described by laws expressing conservation
speed VWIND (m/s), and daily average percentage relative humid- of mass, momentum, energy, and salt. Assuming that saline is
ity RH (%) are given. The data for ESOLAR, Clns, and VWIND are well modeled as an incompressible fluid, i.e., energy cannot be
taken from Ref. [14]. The insolation data and TAVE are taken
from Ref. [15].
stored by compressing fluid, then the fluid velocity behaves as a
solenoidal vector field. Conservation of linear momentum is
Mon ESOLAR TAVE VWIND RH% expressed by the equation
 
Jan 4.03 24.0 5.37 58 @vj @vj
qðT; SÞ þ vk
Feb 4.91 24.7 5.33 54 @t @xk
Mar 5.69 27.3 5.20 48   
Apr 6.39 31.0 4.67 43
@p @ @vj @vk
¼ qðT; SÞ g dj3  þ lðT; SÞ þ (4)
May 6.79 34.3 4.12 36 @xj @xk @xk @xj
Jun 7.07 35.8 4.41 33
Jul 6.95 35.9 5.40 34 where Einstein’s notation has been used, namely, the repetition of an
Aug 6.57 35.7 5.22 39 index implies summation of that index from 1 to 3. Also, g (assumed
Sep 6.06 35.0 4.38 45 constant) denotes the acceleration due to gravity, p (N/m2) denotes
Oct 5.43 32.2 3.99 50
Nov 4.24 28.4 4.30 58
hydrostatic pressure, l(T, S) (kg/m s) denotes the dynamic viscosity
Dec 3.78 25.6 4.98 59 of saline, and where v1, v2, and v3 (m/s) are the respective compo-
nents of fluid velocity with respect to the right-handed Cartesian
triad (e1, e2, e3) of unit vectors with e3 directed vertically downward.
Conservation of energy is expressed by the equation
materials used in their construction give the respective heat trans-
fer coefficients    
@U @U @ @T
qðT; SÞ þ vj ¼ jT ðT; SÞ þ r ðt; x3 Þ
@t @xj @xj @xj
hwall  0:03963 Wm1 K1 (1)  
@vj @vk @vj
þ lðT; SÞ þ (5)
hbase  0:02669 Wm1 K1 (2) @xk @xj @xk

The LCZ of the MSP has depth dLCZ ¼ 1.35 m as determined by where U(T, S) (J/kg) denotes internal energy, and jT(T, S) (W/m
the geometry of the salt charger. Thus, dUCZ þ dNCZ ¼ 1.75 m to K) denotes thermal conductivity. The function r(t, x3) appearing
give a fixed pond depth of 3.10 m. Finally, field measurements of in Eq. (5) is the volume rate of heating in the SGSP at depth x3
the temperature and salinity of the UCZ and LCZ confirm that and time t, and takes the form
both zones are well mixed thermally and in salinity. The analyses
to follow will therefore assume that the UCZ and LCZ behave as rðt; x3 Þ ¼ EðtÞ bðx3 Þ t 2 ½0; 24 (6)
lumped regions at respective temperatures TUCZ and TLCZ and
respective salt concentrations SUCZ and SLCZ. The commonly in which E(t) (W) is the irradiance received by the pond, and b(x3)
accepted view in a SGSP of any significant size is that the equilib- is the negative gradient of the fraction of irradiance reaching
rium distributions of temperature and salinity within the NCZ are depth x3 in the pond. An effective working formula for E(t) during
well described spatially by functions of depth alone. A realistic daylight hours is
analysis of a SGSP should recognize that the constitutive proper-  
ties of saline are dynamic functions of temperature and salinity. pE cos H þ cosðtp=12Þ
t 2 ½tdawn ; tdusk  (7)
The large changes in temperature and salinity throughout an 24 H cos H  sin H
operating SGSP thus drive significant variability in the
material properties of saline across the NCZ. Table 1 gives the where E is the insolation, and the critical hour angle, H, is calcu-
monthly average values at Makkah, Saudi Arabia, of insolation lated from the latitude k and the current declination angle d via
(kWh/m2/day) in the column headed ESOLAR, average Celsius the formula
temperature in the column headed TAVE, average wind speed
(m/s) in the column headed VWIND, and average percentage rela- H ¼ cos1 ðtan k tan dÞ (8)
tive humidity in the column headed RH%.
The value of a weather property with periodicity one year is Sunrise and sunset are the solutions of the equation E(t) ¼ 0, and
characterized by the Fourier expression therefore tdawn ¼ 12(1  H/p) and tdusk ¼ 12(1 þ H/p). The func-
tion b(x3) in the specification of r(t, x3) has expression
X
5
f ðtÞ ¼ a0 þ ak cosð2kptÞ þ bk sinð2kptÞ (3) X
4
k¼1 gk lk elk x3
k¼1
bðx3 Þ ¼ (9)
in which t 僆 [0, 1] and the nth day of the year corresponds to X
4
t ¼ n/365. The parameters a0 ; …; a5 ; b0 ; …; b5 in expression (3) gk
are estimated by an ordinary least squares fit of the monthly k¼1
weather averages given in Table 1. The details of the calculation
are given in the Appendix. A Fourier series with five trigonomet- where g1 ; …g4 and l1 ; …; l4 take the values quoted in Ref. [16],
ric terms has 11 parameters to be estimated from 12 data, and namely,
therefore provides the smallest minimum of the least squares tar-
g1 ¼ 0:190 l1 ¼ 20:0
get function within the framework of the ordinary least squares
g2 ¼ 0:230 l2 ¼ 1:75
methodology. (10)
g3 ¼ 0:301 l3 ¼ 0:0656
g4 ¼ 0:141 l4 ¼ 0:0102
3 Problem Formulation
Let q ¼ q(T, S) (kg/m3) denote the density of saline at Celsius Expression (9) is simply the derivative of the empirical formula
temperature T C and salinity S (kg/m3). Typically, q is a strictly proposed by Hull et al. [16] for the fraction of irradiance

Journal of Solar Energy Engineering AUGUST 2017, Vol. 139 / 041010-3

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


a SGSP is concerned with the long term behavior of these aver-
ages under the assumption that the daily pattern of insolation and
heat extraction at that time of year is repeated indefinitely.
A new system of approximate equations is constructed for these
averages by first integrating Eqs. (4)–(11) over (t, t þ Dt), where t
denotes time of year, and Dt is the duration of a day. The resulting
equations are divided by Dt and the limit taken as Dt ! 0 to obtain
the reformulated equations
 
@vj @vj @p
qðT; SÞ þ vk ¼ qðT; SÞ g dj3 
@t @xk @xj
  
@ @vj @vk
þ lðT; SÞ þ (16)
@xk @xk @xj
Fig. 3 The upper curve shows the average insolation (kWh/m2/
day) received by the Makkah solar pond throughout the year.    
The lower curves represent the average rate of heat extraction
@U @U @ @T
qðT; SÞ þ vj ¼ jT ðT; SÞ þ E bðx3 Þ
(kWh/m2/day) for a UCZ of thickness 0.25 m (top curve), 0.50 m @t @xj @xj @xj
(middle curve), and 0.75 m (bottom curve) with the LCZ main-  
@vj @vk @vj
tained at 95  C. þ lðT; SÞ þ (17)
@xk @xj @xk

penetrating to depth x3 in a solar pond. Finally, conservation of  


@S @S @ @S
salt is expressed by the equation þ vj ¼ jS ðT; SÞ (18)
@t @xj @xj @xj
 
@S @S @ @S
þ vj ¼ jS ðT; SÞ (11) in which E (constant) is the average daily insolation at time t of
@t @xj @xj @xj the year, that is
where jS ¼ jS(T, S) (m2/s) denotes the coefficient of diffusion of ð tþDt
1
salt in water. In the context of a SGSP, Eqs. (4)–(11) assume that E¼ EðsÞ ds (19)
heat flux, salt flux, and the components of the fluid stress tensor Dt t
have constitutive equations
The individual terms of Eqs. (16)–(18) are constructed using the
@T mean value theorem for integrals, and so all the functions appear-
qi ¼ jT ðT; SÞ (12) ing in Eqs. (16)–(18) are to be interpreted as daily average values.
@xi
Equations (16)–(18) now have a stationary solution, the stability
@S of which is to be investigated. If the daily average values of tem-
ci ¼ jS ðT; SÞ (13) perature and salinity are unstable, then the SGSP is unstable.
@xi Equations (16)–(18) are to be solved with boundary conditions
  on the UCZ/NCZ and NCZ/LCZ interfaces that reflect the fact
@vi @vj that the UCZ and LCZ of a managed salinity-gradient solar pond
rij ¼ pðx; tÞdij þ lðT; SÞ þ (14)
@xj @xi are assumed to behave as lumped convecting regions of fluid
mechanically isolated from the NCZ in the respect that the UCZ/
Simplified, but accurate, constitutive expressions for q(T, S), l(T, S), NCZ and NCZ/LCZ interfaces are flat and free from shear stress.
U(T, S), and jT(T, S) are given in Ref. [17] and for convenience Moreover, convection in the UCZ and LCZ enforces approxi-
are reproduced in the Appendix together with approximate maxi- mately uniform salinity and temperature distributions in these
mum relative errors over the range of Celsius temperatures and sal- zones.
inities relevant for a salinity-gradient solar pond. The constitutive
expression for jS(T, S) is 3.2 The UCZ/NCZ Interface. Let d1 ¼ dUCZ, then the
6
UCZ/NCZ interface is located at z ¼ d1. Salinity, mechanical, and
4:6741  10 thermal boundary condition on the UCZ/NCZ interface are now
jS ðT; SÞ ¼  3:8704  109 (15)
1000:0  3:7401 T expressed in terms of the daily averages of temperature and salin-
ity for the UCZ, namely, TUCZ and SUCZ, respectively.
for T > 0 based on a parameter fit to the data provided in Fig. 4 of
Bastug and Kuyucak [18]. The supremum relative error in this 3.2.1 Salinity Boundary Condition. Continuity of salinity is
specification is 1.41%, and the RMS relative error is 1.04%. enforced at the UCZ/NCZ interface so that

SNCZ ðx1 ; x2 ; d1 ; tÞ ¼ SUCZ ðtÞ (20)


3.1 Stationary Approximation. There are two immediate
difficulties to be overcome in the stability analysis of a SGSP. where the value of SUCZ(t) is determined by the management of
First is the strong dependence of the properties of saline on tem- the pond. Typically, the UCZ is flushed with fresh water at a fixed
perature and salinity. Temperatures in a SGSP typically range user-specified rate to remove salt deposited by virtue of evapora-
from 20  C to around 100  C, while salinities range from almost tion and to replace evaporation losses. Excess fluid is discharged
zero to saturation. Second, solar energy is received during day- under gravity to an evaporation pond by means of a fixed overflow
light hours, while heat is typically extracted during darkness so that the depth of the SGSP remains fixed. If flushing water is
hours. Consequently, Eqs. (4)–(11) have no stationary solution in supplied at the rate of F cubic meters per second per square meter
the strict sense of temperature and salinity profiles that are func- of pond area, then SUCZ(t) satisfies the equation
tions of position alone. However, diurnal and climatic time scales
are very different. On a diurnal time scale, it is reasonable to
dSUCZ @SNCZ ðd1 Þ
assume that the behavior of a solar pond is well described by the d1 ¼ jS ðTUCZ ; SUCZ Þ  ðF  FEVAP Þ SUCZ
behaviors of the daily averages of pressure, velocity, temperature, dt @z
and salinity at a given time of year. In this context, the stability of (21)

041010-4 / Vol. 139, AUGUST 2017 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


where FEVAP is the evaporation rate from the surface of the SGSP QEVA
per square meter. FEVAP ¼ (30)
LEVAP qðTAVE ; SUCZ Þ
3.2.2 Mechanical Boundary Condition. The mechanical
boundary conditions require the UCZ/NCZ interface to be flat and where LEVAP denotes the latent heat of evaporation of water.
free from shear stress. Thus The temperature boundary condition describing the behavior of
the UCZ enforces continuity of temperature at the UCZ/NCZ
@ 2 v3 ð x; y; d1 ; t Þ interface, namely,
v3 ð x; y; d1 ; t Þ ¼ ¼0 (22)
@z2
TNCZ ðx1 ; x2 ; d1 ; tÞ ¼ TUCZ ðtÞ (31)

3.2.3 Temperature Boundary Condition. The construction of


where TUCZ (t) is determined by conservation of energy for the
the thermal boundary condition for the UCZ begins by recogniz-
UCZ and is constructed by integrating Eq. (17) over the volume
ing that the heat flux normal to the surface of the UCZ (exposed
occupied by the UCZ to get
to the atmosphere), here taken to be z ¼ 0, is the sum of the aver-
age rate of heat losses due to radiation, convection, and evapora- ð d1 ð
dUðTUCZ ; SUCZ Þ
tion (see Fig. 1). The form of this boundary condition is qðTUCZ ; SUCZ Þ dAdz
0 A dt
q3 jz¼0 ¼ QRAD ðTUCZ Þ þ QCON ðTUCZ Þ þ QEVA ðTUCZ Þ (23) ð ð d1
 
¼ q3 jz¼0  q3 jz¼d1 dA þ AE bðzÞ dz
A 0
where TUCZ is the average daily temperature of the UCZ. The ð d1 ð
functions QRAD, QCON, and QEVA in Eq. (23) specify the surface  q : n dz dl (32)
rate of heat loss due to the mechanisms of radiation, convection, 0 @A
and evaporation, respectively, and are described by the constitu-
tive equations where A ¼ jAj denotes the surface area of the pond. The rate of
working of the stress tensor, namely, the term rijvi,j, is assumed to
QRAD ðTÞ ¼ rðT 4  TSKY
4
Þ (24) be negligible and is omitted from Eq. (32). Equation (32) is
divided by A and the expressions for b(z), the heat flux on the sur-
QCON ðTÞ ¼ CCON ðT  TAVE Þ (25) face of the pond and the heat flux at the UCZ/NCZ interface are
replaced by their definitions to get
QEVA ðTÞ ¼ CH ½CTB PWVP ðTÞð1  wÞ0:88 (26) ð d1 ð
1 dUðTUCZ ; SUCZ Þ
qðTUCZ ; SUCZ Þ dAdz
In expression (24), r denotes Stefan’s constant, T is the Kelvin A 0 A dt
temperature of the surface of the pond, and TSKY is the effective ¼ QRAD ðTUCZ Þ  QCON ðTUCZ Þ  QEVA ðTUCZ Þ
Kelvin temperature of the (clear sky) black-body radiation ð
received by the solar pond from the environment. The sky temper- 1 @TNCZ ð x; y; d1 ; tÞ
þ jT ðTUCZ ; SUCZ Þ dA
ature in this analysis will be specified using Swinbank’s [19] A A @x3
2 3
empirical (long-wave radiation) formula which asserts that X 4
lj d1
ð ð 6 gj e 7
3=2 1 d1 6 7
TSKY ¼ 0:0552 TAMB (27) 6 j¼1 7
 q : n dz dl þ E 61  7 (33)
A 0 @A 6 X 4 7
4 g 5
where TAMB is the ambient Kelvin temperature of the atmosphere j
j¼1
at 2 m above ground level. The analysis to follow will assume that
TAMB ¼ TAVE, the average daily temperature of the atmosphere at
2 m above ground level. Deacon [20] has shown that Swinbank’s The mean value theorem for integrals in combination with the
formula can be established from knowledge of atmospheric emis- assumed approximately spatial uniform temperature and salinity
sion by noting that at most localities there is a moderately strong throughout the UCZ allows the previous equation to be well
correlation between ambient temperature and the amount of water approximated by the equation
vapor overhead.
Expression (26) is Holman’s [21] model of the evaporation heat dUðTUCZ ; SUCZ Þ
flux from a water surface at Celsius temperature T. In this d1 qðTUCZ ; SUCZ Þ
dt
formula, w 僆 [0, 1] denotes relative humidity, and PWVP(T)
¼ QRAD ðTUCZ Þ  QCON ðTUCZ Þ  QEVA ðTUCZ Þ
denotes the saturation pressure of water vapor at Celsius tempera-
ture T. The constants CCON and CH appearing in Eqs. (25) and @TNCZ ðd1 ; tÞ
þjT ðTUCZ ; SUCZ Þ
(26) are defined by the respective formulae @x3
2 3
X4
CCON ¼ 5:7 þ 3:8 VWIND lj d1
ð ð 6 gj e 7
(28) 1 d1 6 7
CH ¼ 0:45ð0:37 þ 0:22 VWIND Þ 6 j¼1 7
 q : n dz dl þ E 61  7 (34)
A 0 @A 6 X 4 7
4 g 5
in which the expression for CCON is due to McAdams [22], and j
j¼1
VWIND is the wind speed in meters per second at 2 m above ground
level. Finally, the quantity
For a small or miniature solar pond, heat lost through the retaining
CTB ðSÞ ¼ 1:025  0:0246 e0:00879S (29) wall of the pond could be important. This case is treated by
replacing q : n in the surface integral of Eq. (34) by the expression
is the Turk and Bonython [23–25] correction for the vapor pres- hWALL ðzÞðTUCZ  TAVE Þ, where hWALL(z) characterizes the ther-
sure of water taking account of salinity. The average evaporation mal conductivity of the retaining wall. The stability analysis to
rate FEVAP in Eq. (21) is computed from expression (26) using the follow will assume a large solar pond which corresponds to the
formula limit of condition (34) as A ! 1. Thus

Journal of Solar Energy Engineering AUGUST 2017, Vol. 139 / 041010-5

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


dUðTUCZ ; SUCZ Þ is manipulated similarly to Eq. (34). Heat lost through the retain-
d1 qðTUCZ ; SUCZ Þ ing wall of a large pond is ignorable to get the final condition
dt
¼ QRAD ðTUCZ Þ  QCON ðTUCZ Þ  QEVA ðTUCZ Þ
2 3 dUðTLCZ ; SLCZ Þ
X4 dLCZ qðTLCZ ; SLCZ Þ
lj d1 2 3 dt
6 gj e 7
@TNCZ ðd1 ; tÞ 6 7 X4
6 j¼1 7 6 gj elj d2 7
þjT ðTUCZ ; SUCZ Þ þ E 61  7 (35) 6 j¼1 7
@x3 6 X 4 7 6 7 @TNCZ
4 gj 5 ¼ E6 7  W  jT ðTLCZ ; SLCZ Þ
6 X 4 7 @x3
j¼1 4 gj 5
j¼1
3.3 The NCZ/LCZ Interface. Let d2 ¼ dUCZ þ dNCZ, then hBASE ðTLCZ  TAVE Þ (41)
the NCZ/LCZ interface is located at z ¼ d2. Salinity, mechanical,
and thermal boundary condition on the NCZ/LCZ interface are
now expressed in terms of the daily averages TLCZ of temperature 4 Stationary Solution
and SLCZ of salinity for the LCZ. Equations (16)–(18) together with the salinity, mechanical, and
3.3.1 Salinity Boundary Condition. Continuity of salinity at temperature boundary conditions have a stationary solution in
the NCZ/LCZ interface gives which the fluid is motionless, i.e., v ¼ 0, and the daily averages of
pressure, temperature, and salinity within the NCZ are functions
SNCZ ðx1 ; x2 ; d2 ; tÞ ¼ SLCZ ðtÞ (36) of depth alone, that is,

T ¼ TðzÞ ;  ;
S ¼ SðzÞ 
p ¼ PðzÞ (42)
where the value of SLCZ (t) is determined by the presence of the
salt charger, the role of which is to maintain undissolved salt at 
the base of the LCZ. Thus, salt diffused upward from the LCZ is The functions TðzÞ; SðzÞ,  satisfy the ordinary differential
and PðzÞ
replaced immediately, which means that the LCZ behaves as satu- equations
rated saline of approximately uniform concentration determined dP  
by the constitutive formula ¼ q T; S g (43)
dz
 
SLCZ ¼ SSAT ðTLCZ Þ (37) d   dT
jT T; S ¼ E bðzÞ (44)
dz dz
where SSAT(T) denotes the saturation concentration of saline at
temperature T.  
d   dS
jS T; S ¼0 (45)
3.3.2 Mechanical Boundary Condition. The mechanical dz dz
boundary conditions require the NCZ/LCZ interface to be flat and
free from shear stress. Thus where z 僆 (d1, d2) and jT, jS, and q are nonconstant functions of

T and S.
v3 ð x; y; d2 ; t Þ ¼ 0 The temperature and salinity boundary conditions on z ¼ d1 are
@ 2 v3 ð x; y; d2 ; t Þ (38) constructed from Eqs. (21), (22), and (35), and the equivalent con-
¼0 ditions on the NCZ/LCZ interface are constructed from Eqs. (37),
@z2 (38), and (41).
The conditions to be satisfied on the UCZ/NCZ boundary at
3.3.3 Temperature Boundary Condition. The temperature z ¼ d1 are
boundary condition on the NCZ/LCZ interface requires continuity
of temperature, namely, jS ðTUCZ ; SUCZ Þ @SNCZ ðd1 Þ
SUCZ  ¼0 (46)
ðF  Fevap Þ @z
TNCZ ðx1 ; x2 ; d2 ; tÞ ¼ TLCZ ðtÞ (39)
dTðd1 Þ
where the behavior of TLCZ(t) is determined by conservation of jT ðTUCZ ; SUCZ Þ  QRAD ðTUCZ Þ  QCON ðTUCZ Þ
dz 2 3
energy for the LCZ. Integration of the energy equation (17) over
the volume occupied by the LCZ gives X4
lj d1
6 gj e 7
ð ð d2 þdLCZ 6 7
6 j¼1 7
dUðTLCZ ; SLCZ Þ  QEVA ðTUCZ Þ þ E 61  7¼0 (47)
qðTLCZ ; SLCZ Þ dz dA 6 X 4 7
A d2 dt 4 gj 5
ð
j¼1
¼ QHEAT FLUX NCZ=LCZ BDRY jz¼d2 dA
A
2 3  1 Þ. Similarly, the conditions
where TUCZ ¼ Tðd1 Þ and SUCZ ¼ Sðd
X 4
6 gj e lj d2
7 to be satisfied on the NCZ/LCZ boundary at z ¼ d2 are
6 j¼1 7
6 7
þA E 6 7  A W  A ðTLCZ  TAVE Þ hBASE SLCZ  SSAT ðTLCZ Þ ¼ 0 (48)
6 X 4 7
4 gj 5
j¼1
2 3
X 4
ð ð d2 þdLCZ 6 gj e lj d2
7
ðTLCZ  TAVE Þ hWALL ðzÞ dz dl (40) dTðd1 þ d2 Þ 6 j¼1 7
6 7
@A d2 jT ðTLCZ ; SLCZ Þ  E6 7
dz 6 X 4 7
4 g 5 j
where W is the daily rate of extraction of heat from the LCZ per j¼1
square meter of pond surface area, and where all insolation pene-
trating to the LCZ is assumed to be absorbed there. Equation (40) þW þ hBASE ðTLCZ  TAVE Þ ¼ 0 (49)

041010-6 / Vol. 139, AUGUST 2017 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


 2 Þ. The management protocol
where TLCZ ¼ Tðd2 Þ and SLCZ ¼ Sðd consistent with intuition and is explicable in terms of a reduction
for the SGSP considered here is one in which the pond is operated in the heat flux reaching the UCZ from the LCZ of the solar pond.
so as to maintain the average temperature of the storage layer at a
given user-specified value. The choice of this temperature then
determines how much heat can be extracted from the LCZ of the 4.3 Salinity of UCZ. Figure 5 illustrates the annual profile of
pond at each time of year. the model average of daily salinity at the Makkah solar pond for
The numerical solution of Eqs. (44) and (45) treats Eqs. (48) various thicknesses of upper convective zone.
and (49) as initial boundary conditions and Eqs. (46) and (47) as Recall that at any time of year the management strategy for the
terminal boundary conditions. Specifically, an interior shooting Makkah solar pond supplies flushing fresh water at a rate designed
procedure manipulates the salinity gradient on the NCZ/LCZ to balance evaporation losses from the solar and evaporation
interface. This procedure is embedded within an exterior shooting ponds. Based on this management protocol, the findings illustrated
procedure which manipulates the initial temperature gradient on in Fig. 5 are again consistent with intuition for the following
the NCZ/LCZ interface. reasons.
(a) The rate at which flushing water is supplied is independent
4.1 Heat Extraction From the LCZ. Figure 3 illustrates the of the thickness of the UCZ. Relatively speaking, therefore,
average amount of heat that can be drawn from the Makkah solar a thin UCZ is flushed more vigorously with fresh water
pond throughout the year in order to sustain a user-specified aver- than a thick UCZ, and consequently the salinity of a thin
age temperature of 95  C neglecting heat losses through the retain- UCZ should be less than that of a thicker UCZ at any given
ing wall of the pond but taking account of heat losses via its base. time of year.
Figure 3 demonstrates that less heat can be extracted from a (b) The reduction in the salinity of the UCZ during summer
thicker UCZ, hence an important reason for trying to restrict the months is entirely predictable and is due to the increased
thickness of the UCZ, e.g., by using floating rings on the surface flushing rate during the heat of the summer months for any
of the UCZ as illustrated in Fig. 2. given thickness of UCZ.
A qualitatively similar behavior to that illustrated in Fig. 3 is (c) The diffusion coefficient of salt in water is a strictly
found when the LCZ is maintained at 80  C. In this case, the increasing function of temperature in the range of tempera-
amount of heat that can be extracted is approximately 5% more tures relevant to the operation of a SGSP. Thus, the rate of
than that illustrated in Fig. 3, but of course, this heat is now diffusion of salt from an LCZ at 95  C is greater than that
delivered at approximately 80  C rather than 95  C. This finding from an LCZ at 80  C. This increased flux of salt is mani-
is consistent with intuition in the respect that the extra heat that fest as an increased salinity of the UCZ.
can be extracted at 80  C is essentially the additional heat that
would have been lost through the base of the pond if the LCZ
was operated at an average temperature of 95  C as opposed to
5 Stability of Infinite Pond
80  C.
The stability of the stationary distributions of pressure, temper-
ature, and salinity for a SGSP pond is investigated by considering
4.2 Temperature of UCZ. Figure 4 illustrates the annual
the behavior of the perturbed state
profile of the model average of daily temperature at the Makkah
solar pond for various thicknesses of upper convective zone.
vj ej ¼ e ðuðx; y; z; tÞ e1 þ vðx; y; z; tÞ e2
Figure 4 illustrates that the average temperature of the UCZ
is always several Celsius degrees below that of the ambient tem- þ wðx; y; z; tÞ e3 Þ þ Oðe2 Þ
perature. This finding is in strong agreement with observation.  þ e pðx; y; z; tÞ þ Oðe2 Þ
p ¼ PðzÞ (50)
Moreover, the average temperature of the UCZ is marginally
higher for thicker UCZs, which is again consistent with intu- T ¼ TðzÞ þ e hðx; y; z; tÞ þ Oðe2 Þ
ition. A thicker UCZ has a larger heat capacity for a given sur-  þ e /ðx; y; z; tÞ þ Oðe2 Þ
S ¼ SðzÞ
face area of pond, and thus has a marginally larger ability to
retain heat.
where e is a small parameter. Expressions (50) are now substituted
A qualitatively similar behavior to that illustrated in Fig. 4 is
into Eqs. (16) and (18), the resulting equations are divided by e,
found when the LCZ is maintained at 80  C. In this case, the aver-
and the limit is then taken as e ! 0 to get the linearized momen-
age temperature of the UCZ is marginally less than that when the
tum equations
pond is operated with the LCZ at 95  C. This finding is again

Fig. 4 The upper curve shows the average daily Celsius tem- Fig. 5 The annual profile of the average daily salinity of the
perature at the Makkah solar pond throughout the year. The UCZ of the Makkah solar pond is shown for a UCZ of thickness
lower curves represent the average daily Celsius temperature 0.75 m (top curves), for a UCZ of thickness 0.50 m (middle
of the UCZ for a thickness of 0.25 m (bottom curve), 0.50 m (mid- curves), and for a UCZ of thickness of 0.25 m (bottom curves).
dle curve), and 0.75 m (top curve) with the LCZ maintained Solid and dashed lines are associated with LCZ temperatures
at 95  C. of 95  C and 80  C, respectively.

Journal of Solar Energy Engineering AUGUST 2017, Vol. 139 / 041010-7

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


 
@u @p df4 ðzÞ@u @w 5.2 Linearized Boundary Conditions at the NCZ/LCZ
f1 ðzÞ ¼  þ f4 ðzÞ 䉭3 u þ þ Interface. The linearized boundary conditions (37), (38), and (41)
@t @x dz @z @x
  at the NCZ/LCZ interface are, respectively,
@v @p df4 ðzÞ @v @w
f1 ðzÞ ¼  þ f4 ðzÞ 䉭3 v þ þ dSSAT ðTLCZ Þ
@t @y dz @z @y /ðd2 ; tÞ  hðd2 ; tÞ ¼ 0 (58)
dT
@w @p df4 ðzÞ @w
f1 ðzÞ ¼ gðf2 ðzÞ h þ f3 ðzÞ /Þ  þ f4 ðzÞ 䉭3 w þ 2
@t @z dz @z
(51) @ 2 wðd2 ; tÞ
wðd2 ; tÞ ¼ ¼0 (59)
@z2
together with the linearized conservation equations of energy and
salt, namely,  
@hðd2 ; tÞ @/ðd2 ; tÞ
dLCZ f5 ðd2 Þ þ f6 ðd2 Þ
@h @/ @t @t
f5 ðzÞ þ f6 ðzÞ @hðd2 ; tÞ
@t @t ¼ f8 ðd2 Þ  f9 ðd2 Þ hðd2 ; tÞ  f11 ðd2 Þ /ðd2 ; tÞ
@h @z
¼ w f7 ðzÞ þ f8 ðzÞ 䉭3 h þ ð2f9 ðzÞ þ f10 ðzÞÞ
@z  hBASE hðd2 ; tÞ (60)
@/ df11 ðzÞ df9 ðzÞ
þ f11 ðzÞ þ /þ h (52)
@z dz dz
5.3 The Eigenvalue Problem. A solution of Eqs. (52)–(54) is
@/ dS @h sought in the form
¼ w þ f12 ðzÞ 䉭3 / þ f13 ðzÞ
@t dz @z
@/ df13 ðzÞ df14 ðzÞ ðh; /; wÞ ¼ ertþimxþiny ðhðzÞ ; /ðzÞ ; wðzÞÞ (61)
þ ð2f14 ðzÞ þ f15 ðzÞÞ þ hþ / (53)
@z dz dz
Let X ¼ 䉭3 w. When ansatz (61) is substituted into Eqs.
The auxiliary functions f1 ðzÞ; …; f15 ðzÞ are defined in the Appen- (52)–(54), the outcome is
dix. The nonconstant nature of the coefficients of Eqs. (51)–(53)
 
means that the analysis of stability for the NCZ of a SGSP differs df1 ðzÞ dw
significantly from that of a thin layer. Despite the complexity of r f1 ðzÞ X þ
dz dz
Eqs. (51)–(53), the analysis can nevertheless proceed by taking  
curl curl of Eq. (51) to obtain d2 X
¼ a2 gf2 ðzÞ h  a2 gf3 ðzÞ / þ f4 ðzÞ  a2
X
  dz2
@ df1 ðzÞ @ @w 2
df4 ðzÞ dX d f4 ðzÞ
f1 ðzÞ ð䉭3 wÞ þ þ2 þ ðX þ 2a2 wÞ (62)
@t dz @t @z dz dz dz2
¼ gf2 ðzÞ 䉭2 h þ gf3 ðzÞ 䉭2 / þ f4 ðzÞ 䉭23 w
    d2 w
df4 ðzÞ @w d2 f4 ðzÞ @ 2 w  a2 w  X ¼ 0 (63)
þ2 䉭3 þ  䉭 w (54) dz2
2
dz @z dz2 @z2
 
where w is the axial component of fluid velocity. Equations d2 h 2
rðf5 ðzÞ h þ f6 ðzÞ /Þ ¼ w f7 ðzÞ þ f8 ðzÞ  a h
(52)–(54) are solved with the linearized forms for the boundary dz2
conditions on the UCZ/NCZ and NCZ/LCZ interfaces. dh df9 ðzÞ
þ ð2f9 ðzÞ þ f10 ðzÞÞ þ h
dz dz
5.1 Linearized Boundary Conditions at the UCZ/NCZ d/ df11 ðzÞ
þ f11 ðzÞ þ / (64)
Interface. The linearized boundary conditions (21), (22), and (35) dz dz
at the UCZ/NCZ interface are, respectively,
 
@/ðd1 ; tÞ @/ðd1 ; tÞ   d S d2 / 2 d/
d1 ¼ f12 ðd1 Þ þ f14 ðd1 Þ  F þ Fevap /ðd1 ; tÞ r / ¼ w þ f12 ðzÞ  a / þ ð2f14 ðzÞ þ f15 ðzÞÞ
@t @z dz dz2 dz
@Fevap ðd1 Þ dh df13 ðzÞ df14 ðzÞ
þ f13 ðd1 Þ hðd1 ; tÞ þ SNCZ hðd1 ; tÞ þ f13 ðzÞ þ hþ / (65)
@T dz dz dz
@Fevap ðd1 Þ
þ SNCZ /ðd1 ; tÞ (55)
@S
5.3.1 The UCZ/NCZ Interface. The boundary conditions to be
2 satisfied by h, /, X, and w at the UCZ/NCZ interface are con-
@ wðd1 ; tÞ structed from conditions (55)–(57) and take the form
wðd1 ; tÞ ¼ ¼0 (56)
@z2

  r d1 /ðd1 Þ
@hðd1 ; tÞ @/ðd1 ; tÞ d/ðd1 Þ  
d1 f5 ðd1 Þ þ f6 ðd1 Þ ¼ f12 ðd1 Þ þ f13 ðd1 Þ hðd1 Þ þ f14 ðd1 Þ  F þ Fevap /ðd1 Þ
@t @t dz
@hðd1 ; tÞ @Fevap ðd1 Þ  @Fevap ðd1 Þ 
¼ f8 ðd1 Þ þ f9 ðd1 Þ hðd1 ; tÞ þ S ðd1 Þ hðd1 Þ þ S ðd1 Þ /ðd1 Þ ¼ 0 (66)
@z @T @S
dðQRAD þ QCON þ QEVA ÞðTUCZ Þ
 hðd1 ; tÞ þ f11 ðd1 Þ /ðd1 ; tÞ
dT
(57) wðd1 Þ ¼ Xðd1 Þ ¼ 0 (67)

041010-8 / Vol. 139, AUGUST 2017 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


r d1 ðf5 ðd1 Þ hðd1 Þ þ f6 ðd1 Þ /ðd1 ÞÞ y0 ðd1 Þ ¼ y1 ðd1 Þ ¼ 0 (78)
dhðd1 Þ
¼ f8 ðd1 Þ þ f9 ðd1 Þ hðd1 Þ r d1 ðf5 ðd1 Þ y2 ðd1 Þ þ f6 ðd1 Þ y3 ðd1 ÞÞ
dz
dðQRAD þ QCON þ QEVA ÞðTUCZ Þ dy2 ðd1 Þ
 hðd1 Þ þ f11 ðd1 Þ /ðd1 Þ ¼ f8 ðd1 Þ þ f9 ðd1 Þ y2 ðd1 Þ
dT dz
(68) dðQRAD þ QCON þ QEVA ÞðTUCZ Þ
 y2 ðd1 Þ þ f11 ðd1 Þ y3 ðd1 Þ
dT
5.3.2 The NCZ/LCZ Interface. The boundary conditions to be (79)
satisfied by h, /, X, and w at the NCZ/LCZ interface are con-
structed from conditions (58)–(60) and take the form NCZ/LCZ interface:
dSSAT ðTLCZ Þ
/ðd2 Þ  hðd2 Þ ¼ 0 (69) y0 ðd2 Þ ¼ y1 ðd2 Þ ¼ 0 (80)
dT

wðd2 Þ ¼ Xðd2 Þ ¼ 0 (70) dSSAT ðTLCZ Þ


y3 ðd2 Þ  y2 ðd2 Þ ¼ 0 (81)
dT
dLCZ r ðf5 ðd2 Þ hðd2 Þ þ f6 ðd2 Þ /ðd2 ÞÞ
dLCZ r ðf5 ðd2 Þ y2 ðd2 Þ þ f6 ðd2 Þ y3 ðd2 ÞÞ
dhðd2 Þ
¼ f8 ðd2 Þ  f9 ðd2 Þ hðd2 Þ  f11 ðd2 Þ /ðd2 Þ  hBASE hðd2 Þ dy2 ðd2 Þ
dz ¼ ðf9 ðd2 Þ þ hBASE Þ y2 ðd2 Þ  f8 ðd2 Þ  f11 ðd2 Þ y3 ðd2 Þ
(71) dz
(82)
5.4 Results. When expressed in terms of the variables
y0 ðzÞ; …; y3 ðzÞ defined by
5.5 Numerical Procedure. Chebyshev collocation is used to
treat the eigenvalue problem posed by Eqs. (62)–(65). In this
y0 ðxÞ ¼ wðzÞ; y1 ðzÞ ¼ XðzÞ
(72) approach, the interval [d1, d2] is dissected by the Chebyshev
y2 ðxÞ ¼ hðzÞ; y3 ðzÞ ¼ /ðzÞ nodes

the eigenvalue problem to be investigated requires the calculation zj ¼ d2 cos2 ðjp=2NÞ þ d1 sin2 ðjp=2NÞ
of the value of r with largest real part, where y0 ðzÞ; …; y3 ðzÞ sat-
isfy the differential equations where j ¼ 0; …; N, and N denotes the maximum order of the
Chebyshev polynomial expansion used to represent the unknowns
d 2 y0 of the problem. In the calculations to follow, N ¼ 41. The first
 y1  a2 y0 ¼ 0 (73)
dz2 derivative of yk(z) (k ¼ 0; …; k ¼ 3) at each internal node
  z1 ; …; zN1 is expressed in terms of the values yk ðz0 Þ; …; yk ðzN Þ
df1 ðzÞ dy0 using the standard Chebyshev collocation differentiation matrix.
r f1 ðzÞ y1 þ Second order derivatives, when needed, are likewise computed at
dz dz
  the internal nodes z1 ; …; zN1 by two applications of the Cheby-
2 2 d 2 y1
¼ a gf2 ðzÞ y2  a gf3 ðzÞ y3 þ f4 ðzÞ  a2 y1 shev collocation differentiation matrix.
dz2 Using this procedure, Eqs. (62)–(65) are collocated at the inter-
df4 ðzÞ dy1 d2 f4 ðzÞ   nal nodes z1 ; …; zN1 , and in so doing each equation contributes
þ2 þ y1 þ 2a2 y0 (74) exactly (N  1) linear equations connecting the elements of the
dz dz dz2
4(N þ 1) column vector Y defined by
rðf5 ðzÞ y2 þ f6 ðzÞ y3 Þ
  Y ¼ ½y0 ðz0 Þ; …y0 ðzN Þ; y1 ðz0 Þ; …; y1 ðzN Þ;
d 2 y2 2 dy2
¼ f7 ðzÞ y0 þ f8 ðzÞ  a y2 þ ð2f9 ðzÞ þ f10 ðzÞÞ y2 ðz0 Þ; …; y2 ðzN Þ; y3 ðz0 Þ; …; y3 ðzN ÞT (83)
dz2 dz
df9 ðzÞ df11 ðzÞ dy3
þ y2 þ y3 þ f11 ðzÞ (75) The result of this procedure is 4(N  1) linear equations connect-
dz dz dz
  ing the unknown values of the components of Y. The remaining
dS d 2 y3 2
eight equations are constructed by enforcing the four boundary
r y3 ¼  y0 þ f12 ðzÞ  a y3 conditions (77)–(79) on z ¼ d1 and the four boundary conditions
dz dz2
(80)–(82) on z ¼ d2. The final outcome is a generalized eigenvalue
dy3 dy2
þ ð2f14 ðzÞ þ f15 ðzÞÞ þ f13 ðzÞ problem of type AY ¼ r(a) BY, where A and B are
dz dz 4(N þ 1)  4(N þ 1) real matrices in which B is singular, Y is a
df13 ðzÞ df14 ðzÞ column vector of length 4(N þ 1), and r(a) is an eigenvalue. The
þ y2 þ y3 (76)
dz dz stability of the SGSP is investigated by maximizing the real part
of r over all possible values of a.
in the interval z 僆 (d1, d2) together with four boundary conditions This is a challenging problem from a numerical perspective due
on the UCZ/NCZ interface z ¼ d1 and four boundary conditions to wide ranging numerical values for the elements of the matrices
on the NCZ/LCZ interface z ¼ d2, namely, A and B. The calculations were therefore done in quadruple preci-
UCZ/NCZ interface: sion (machine error 1035) using ABSOFT FORTRAN 95.
dy3 ðd1 Þ
r d1 y3 ðd1 Þ ¼ f12 ðd1 Þ þ ðf14 ðd1 Þ  F þ FEVAP Þy3 ðd1 Þ 6 Results and Conclusions
 dz 
@FEVAP ðd1 Þ  The stability analysis, when applied to the Makkah solar pond,
þ f13 ðd1 Þ þ S ðd1 Þ y2 ðd1 Þ recovers eigenvalues with negative real parts under all circum-
@T
stances. Consequently, the average steady-state configurations
@FEVAP ðd1 Þ  of temperature and salinity for the MSP are stable to small pertur-
þ S ðd1 Þ y3 ðd1 Þ (77)
@S bations of temperature and salinity. Of course, this is not a

Journal of Solar Energy Engineering AUGUST 2017, Vol. 139 / 041010-9

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


through the base of the pond in addition to any usable heat
that is extracted, then the lower section of the NCZ is often
at least as hot if not hotter than the LCZ. Consequently,
heat extracted from the LCZ is partially replenished from
the NCZ, and the larger the heat capacity of the NCZ, the
more able it is to perform this task.
(2) At first sight, a thicker UCZ might appear inconsistent with
increased stability. However, the temperature difference
across the NCZ is relatively independent of its thickness at
any given time of year, and the flushing protocol likewise
ensures that the same is true for the salinity difference across
the NCZ. Consequently, a thicker UCZ and thinner NCZ
translates into a larger downward density gradient, which in
combination with viscous damping, improves stability.

Fig. 6 The value of s 5 jrj21 is shown for the Makkah solar 6.2 Second Conclusion. Figure 6 reveals that under the pro-
pond for UCZs of thickness 0.25 m, 0.50 m, and 0.75 m when the posed operating protocol, a salt-gradient solar pond is least stable
LCZ is operated at 95  C (solid lines) and at 80  C (dashed lines) during winter months, a finding which at first sight might seem
counter-intuitive. Ordinarily, one usually associates more intensive
surprising finding in the respect that the pond is subjected to per- heating, such as that which occurs during summer months, with
turbations in temperature and salinity on a daily basis due to reduced rather than increased stability. The explanation resides in
weather but yet retains its integrity as an operating SGSP over a the fact that the proposed operating protocol requires the amount of
long period of time. However, what is unknown are: heat extracted from the LCZ of the solar pond to be such that it
maintains the temperature of the LCZ at a user-specified value
(a) the nature of the development of instability independent of the time of year. However, the average temperature
(b) the dependence of the strength of the restoring forces of the UCZ is significantly lower during winter months, and so a
on the time of year direct consequence of this operating protocol is that the average
(c) the dependence of the strength of the restoring forces temperature difference across the NCZ is significantly larger dur-
on the thickness of the UCZ and the operating temperature ing winter months for a given thickness of UCZ. By contrast, the
of the LCZ salinity gradient is largely unchanged throughout the year. Because
As mentioned previously, the first finding is that the critical a downward temperature gradient acts to destabilize, whereas a
eigenvalue is always real which suggests that instability would downward salinity gradient acts to stabilize, then the proposed
occur via blow-up, that is, by the mechanism of stationary stabil- operating protocol degrades pond stability during winter months
ity. Observations of naturally occurring solar ponds indicate that due to the presence of the increased temperature gradient.
this is exactly what happens in practice, e.g., Solar Lake at Taba, For UCZ thicknesses of 0.50 m or less, recovery times are longer
Sinai, Egypt. during winter months and shorter during summer months when the
The strength of the restoring forces acting within the pond is LCZ operates at 95  C as opposed to 80  C. Beyond a UCZ of
measured in this investigation in terms of the time scale s ¼ jrj1 . 0.50 m thickness, the recovery time for an LCZ operating at 80  C
Large values of s are associated with weak restoring forces, is always less than that when the LCZ operates at 95  C. We offer
whereas small values of s are associated with strong restoration. no simple explanation for this behavior other than to note that it is
Figure 6 plots the value of s throughout the year for the Makkah probably a feature of the structure of the temperature and salinity
solar pond for various thicknesses of upper convective zone. gradients in the NCZ. In a thin NCZ, these gradients are largely
Recall that calculations are based on a management protocol in constant. By contrast, the lower section of a thick NCZ is more
which the rate of flow of fresh flushing water exactly matches the effectively insulated from surface effects so that the temperature
rate of evaporation from the solar pond and an evaporation pond gradient tends to be low in the vicinity of the NCZ/LCZ interface
(area 200 m2). The following conclusions can be drawn from Fig. 6. and high in the vicinity of the UCZ/NCZ interface. This effective
insulation partly accounts for why a thin UCZ is conducive to
increased thermal efficiency for a SGSP.
6.1 First Conclusion. Solar ponds with thicker UCZs have
shorter recovery times (and are therefore more stable) for a given Acknowledgment
operating temperature of the LCZ. While at first sight this might
seem contrary to the practical management of a solar pond in the The authors are grateful for the financial support for this project
respect that one strives to minimize the thickness of the UCZ, there from King Abdulaziz City of Science and Technology, Saudi Ara-
is no contradiction. As Fig. 4 illustrates, the reason for controlling bia through its National Science, Technology and Innovation Plan
the thickness of the UCZ is to increase the useful heat output from (Research Project No. 12-MAT2969-10).
the pond rather than improve its stability. In reality, the extent to
which the thickness of the UCZ can be controlled depends on the
flushing process and the prevailing weather at the site of the pond.
The Makkah solar pond uses a floating array of rings (see Fig. 2) to Appendix: Constitutive Formulae
reduce the influence of wind-driven wave action as suggested by
Leblanc et al. [2]. However, the site experiences strong winds, and qðT; SÞ ¼ 1004:3512  0:2381 T þ 0:6819 S
even with these rings in place, the thickness of the UCZ at the Mak-
kah pond drifts between 0.4 m and 0.5 m.  ð2:0560 T 2 þ 0:1134 TS þ 0:1370 S2 Þ  103 (A1)
Because the sum of the thicknesses of the UCZ and NCZ is
fixed in a SGSP, then a thinner UCZ translates to a thicker NCZ. jT ðT; SÞ ¼ 0:7282  0:1040  103 S  0:1609 expð0:0121 TÞ
Two further comments relevant to this feature are now discussed. (A2)
(1) The increase in usable heat output in the presence of a thin
UCZ originates from the increased heat capacity of the 4:6741  106
jS ðT; SÞ ¼  3:8704  109 (A3)
NCZ. Specifically, because the LCZ continually sheds heat 1000:0  3:7401 T

041010-10 / Vol. 139, AUGUST 2017 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


SSAT ðTÞ ¼ 316:0155 þ 0:3422  101 T þ 0:8403  103 T 2 (A4) Approximation of Weather Properties
 Let tk denote the time at the end of the kth month and start of the
425:73 (k þ 1)th month of the year with t0 ¼ 0 denoting the start of the
lðT; SÞ ¼ exp 10:6333 þ þ 0:3096  102 S year and t12 ¼ 1 denoting the end of the year. Suppose further
100 þ 0:7255 T
# that the annual property to be interpolated is modeled by the
0:1888 S 0:2982  101 S2 expression
 ; þ (A5)
100 þ 0:7255 T ð100 þ 0:7255 T Þ2
X
M
 f ðtÞ ¼ a0 þ aj cosð2jptÞ þ bj sinð2jptÞ (A8)
4017:4528
PðT; SÞ ¼ exp 23:5402   0:1465  102 S j¼1
T þ 234:4956
#
0:2616 S 0:8802  101 S2 where t 僆 [0, 1]. Specifically, the value of f on the nth day of the
þ  (A6) year is obtained by substituting t ¼ n/365 in expression (A8).
T þ 234:4956 ðT þ 234:4956Þ2 The model average of f(t) over the kth month is
UðT; SÞ ¼ 6723:8865 þ 4114:7270 T  3:2519 TS ð tk
1
6 1 fk ¼ f ðtÞ dt (A9)
þ 1:15783  10  expð0:1084  10 SÞ (A7) tk  tk1 tk1

When expressed in terms of sincðxÞ ¼ sin x=x


Maximum Relative Error
X
M
Property Units Maximum relative error fk ¼ a0 þ sinc /jk ðaj cos wjk þ bj sin wjk Þ (A10)
j¼1
3
Density q kg/m 0.4019%
Thermal conductivity jT W/m K 1.1436% where
Diffusivity of salt in water jS m2/s 1.4184%
Solubility kg/m3 0.2040% #
Dynamic viscosity kg/m s 0.9584%
/jk ¼ pj ðtk  tk1 Þ
(A11)
Vapor pressure N/m2 1.4592% wjk ¼ pj ðtk1 þ tk Þ
Internal energy J/kg Not applicable
Suppose now that Fk is the data value of fk , then the ordinary least
squares estimate of the coefficient a0 and (aj, bj) for j ¼ 1; …; M
Auxiliary Functions is obtained by choosing these coefficients to minimize the value
The auxiliary functions f1 ðzÞ; …; f15 ðzÞ in Eqs. (51)–(53) are of the target function
defined as follows:
X
12
Uða0 ; …; aM ; b0 ; …; bM Þ ¼ ðFk  fk Þ2 (A12)

f1 ðzÞ ¼ qðT; SÞ k¼1
 
@q T; S Specifically, a0 ; …; aM ; b0 …; bM are required to satisfy the
f2 ðzÞ ¼
@T conditions
 
@q T; S
f3 ðzÞ ¼ @U @U @U @U
@S ¼  ¼ ¼ ¼  ¼ ¼0 (A13)
@a0 @aM @b0 @bM

f4 ðzÞ ¼ lðT; SÞ
 
  @U T; S Conditions (A13) give rise to a system of (2 M þ 1) linear equa-
f5 ðzÞ ¼ q T; S tions for the coefficients a0 ; …; aM ; b0 …; bM , which in turn deter-
@T
  mine the functional form of f(t). Note that if (2 M þ 1) exceeds the
  @U T; S number of data (12 in this case), then the minimum value of U ¼ 0
f6 ðzÞ ¼ q T; S
@S  is achievable in an arbitrary number of ways. Therefore, the num-
  dU T; S ber of variable coefficients in expression (A8) must be less than
f7 ðzÞ ¼ q T; S
dz

f8 ðzÞ ¼ jT ðT; SÞ
 
@jT T; S d T
f9 ðzÞ ¼
@T dz
 
@jT T; S d S
f10 ðzÞ ¼
@S
  dz
@jT T; S d T
f11 ðzÞ ¼
@S dz

f12 ðzÞ ¼ jS ðT; SÞ
 
@jS T; S d S
f13 ðzÞ ¼
@T
  dz
@jS T; S d S
f14 ðzÞ ¼
@S dz
 

@jS T ; S d T

f15 ðzÞ ¼ Fig. 7 The annual profile of the insolation is shown based on
@T dz expression (A8) with M 5 5 (dashed line) and M 5 2 (dotted line)

Journal of Solar Energy Engineering AUGUST 2017, Vol. 139 / 041010-11

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm


the number of data, which in this work imposes the constraint [9] Boudhiaf, R., Moussa, A. B., and Baccar, M., 2012, “A Two-Dimensional
M  5. In particular, the choice M ¼ 5 will yield the smallest mini- Numerical Study of Hydrodynamic, Heat and Mass Transfer and Stability in a
Salt Gradient Solar Pond,” Energies, 5(12), pp. 3986–4007.
mum for U, and consequently the best ordinary least squares fit to [10] Karim, C., and Jomaa, S. M., 2012, “Experimental Simulation of the Salt Gradi-
the data. ent Solar Pond Stability,” J. Renewable Sustainable Energy, 4(6), p. 063143.
In the context of the annual profile of insolation, Fig. 7 illus- [11] Akrour, D., Tribeche, M., and Kalache, D., 2011, “A Theoretical and Numerical
trates how expression (A8) varies throughout the year for M ¼ 2 Study of Thermosolutal Convection: Stability of a Salinity Gradient Solar
Pond,” Therm. Sci., 15(1), pp. 67–80.
and M ¼ 5. When M ¼ 5, expression (A8) is seen to capture the [12] Sodha, M. S., and Kumar, A., 1983, “Linear Stability Analysis of Double-
shoulder in insolation around late September and the fact there is Diffusive Solar Pond,” Energy Conserv., 23(2), pp. 67–71.
a clear minimum in average monthly insolation in December (in [13] Abdullah, A. A., Lindsay, K. A., and AbdelGawad, A. F., 2016, “Construction
agreement with the data of Table 1). On the other hand, expres- of Sustainable Heat Extraction System and a New Scheme of Temperature
Measurement in an Experimental Solar Pond for Performance Enhancement,”
sion (A8) fails to capture these features when M ¼ 2. Sol. Energy, 130, pp. 10–24.
[14] Matti Tukiainen, 2005, “Mecca, Saudi Arabia - Sunrise, Sunset, Dawn and
Dusk Times,” Yl€ oj€arvi, Finland, accessed May 22, 2017, http://www.gaisma.
com/en/location/mecca.html
References [15] Weather Atlas, 2002, “Monthly Weather Forecast and Climate Mecca, Saudi
[1] Cohen, Y., Krumbein, W. E., Goldberg, M., and Shilo, M., 2003, “Solar Lake Arabia,” Yu Media Group d.o.o. Beograd, Belgrade, Serbia, accessed May 22,
(Sinai)—1: Physical and Chemical Limnology,” Limnol. Oceanogr., 22(4), pp. 2017, http://www.weather-atlas.com/en/saudi-arabia/mecca-climate
597–608. [16] Hull, J. R., Nielsen, C. E., and Golding, P., 1988, Salinity-Gradient Solar
[2] Leblanc, J., Akbarzadeh, A., Andrews, J., Lu, H., and Golding, P., 2011, “Heat Ponds, CRC Press, Boca Raton, FL.
Extraction Methods From Salinity-Gradient Solar Ponds and Introduction of a [17] Abdullah, A. A., Lindsay, K. A., and AbdelGawad, A. F., 2015, “Parsimonious Con-
Novel System of Heat Extraction for Improved Efficiency,” Sol. Energy, stitutive Expressions With Good Accuracy and Suitable for Modelling the Properties
85(12), pp. 3103–3142. of Aqueous Sodium Chloride in Solar Ponds,” Sol. Energy, 122, pp. 617–629.
[3] Giestas, M. C., Milhazes, J. P., and Pina, H. L., 2014, “Numerical Modeling of [18] Bastug, T., and Kuyucak, S., 2005, “Temperature Dependence of the Transport
Solar Ponds,” Energy Procedia, 57, pp. 2416–2425. Coefficients of Ions From Molecular Dynamics Simulations,” Chem. Phys.
[4] Shi, Y., Yin, F., Shi, L., Sun, W., Li, N., and Liu, H., 2011, “Effects of Porous Lett., 408(1–3), pp. 84–88.
Media on Thermal and Salt Diffusion of Solar Pond,” Appl. Energy, 88(7), pp. [19] Swinbank, W. C., 1963, “Long-Wave Radiation From Clear Skies,” Q. J. R.
2445–2453. Meteorol. Soc., 89(381), pp. 339–348.
[5] Valderrama, C., Gibert, O., Arcal, J., Solano, P., Akbarzadeh, A., Larrotcha, E., [20] Deacon, E. L., 1970, “The Derivation of Swinbank’s Long-Wave Radiation
and Cortina, J. L., 2011, “Solar Energy Storage by Salinity Gradient Solar Formula,” Q. J. R. Meteorol. Soc., 96(408), pp. 313–319.
Pond: Pilot Plant Construction and Gradient Control,” Desalination, 279(13), [21] Holman, J. P., 1992, Heat Transfer, 7th ed., McGraw-Hill, New York.
pp. 445–450. [22] McAdams, W. H., 1954, Heat Transmission, 3rd ed., McGraw-Hill Kogakusha,
[6] Suarez, F., Tyler, S. W., and Childress, A. E., 2010, “A Fully Coupled, Tran- Tokyo, Japan, p. 249.
sient Double-Diffusive Convective Model for Salt-Gradient Solar Ponds,” Int. [23] Bonython, C. W., 1956, “The Influence of Salinity Upon the Rate of Natural Evapo-
J. Heat Mass Transfer, 53(9), pp. 1718–1730. ration,” UNESCO, Canberra, New South Wales, Australia, Report No. 65-71.
[7] Giestas, M. C., Pina, H. L., Milhazes, J. P., and Tavares, C., 2009, “Solar Pond [24] Bonython, C. W., 1965, “Factors Determining the Rate of Solar Evaporation in
Modeling With Density and Viscosity Dependent on Temperature and the Production of Salt,” Second Northern Ohio Geological Society Symposium on
Salinity,” Int. J. Heat Mass Transfer, 52(12), pp. 2849–2857. Salt, Northern Ohio Geological Society, Cleveland, OH, May 3–5, pp. 152–167.
[8] Hill, A. A., and Carr, M., 2013, “The Influence of a Fluid-Porous Interface on [25] Turk, L. J., 1970, “Evaporation of Brine: A Field Study on the Bonneville Salt
Solar Pond Stability,” Adv. Water Resour., 52, pp. 1–6. Flats, Utah,” Water Resour. Res., 6(4), pp. 1209–1215.

041010-12 / Vol. 139, AUGUST 2017 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/936206/ on 06/12/2017 Terms of Use: http://www.asm

You might also like