New approach for Finite Difference Method for Thermal Analysis of Passive Solar Systems

Stanko Shtrakov and Anton Stoilov

Department of Computer’s systems, South - West University “Neofit Rilski”, Blagoevgrad, BULGARIA,
(Dated: February 17, 2005)

Mathematical treatment of massive wall systems is a useful tool for investigation of these solar
applications. The objectives of this work are to develop (and validate) a numerical solution model for
predication the thermal behaviour of passive solar systems with massive wall, to improve knowledge
of using indirect passive solar systems and assess its energy efficiency according to climatic conditions
in Bulgaria. The problem of passive solar systems with massive walls is modelled by thermal and mass
transfer equations. As a boundary conditions for the mathematical problem are used equations, which
describe influence of weather data and constructive parameters of building on the thermal performance
of the passive system. The mathematical model is solved by means of finite-differences method and
improved solution procedure. In article are presented results of theoretical and experimental study for
developing and validating a numerical solution model for predication the thermal behaviour of passive
solar systems with massive wall.

I. INTRODUCTION
The concept of passive solar systems is well-known method for use of solar energy as a source of heating in
buildings. There is a vast literature on this technology, but real objects (houses) with passive solar systems are
still rare. The major impediments to increase market penetration of passive solar systems is the lack of
available information and experience data for the efficiency and constructive parameters of passive solar
elements.
The main concept of indirect passive solar systems is Trombe-Michel wall. Most experimental and theoretical
data, published on Trombe wall performance are in form of overall building performance. Data of overall
performance is of limited use, as it only provides seasonal estimates of heat gains for specific building
designs, wall patterns and climates. Because of the large number of parameters and the wide range of weather
conditions, which influence the operation of massive walls, the assessment of the thermal behavior requires
the use of thermal simulation techniques.
Literature review shows that the problem of passive solar systems with massive walls is ordinarily modelled
by thermal and mass transfer equations [4,7]. As boundary conditions for the mathematical problem must be
used equations, which describe influence of weather data and constructive parameters of building on the
thermal performance of the passive system. The mathematical model, composed for the massive wall
performance, is usually very complicated and for solving the mathematical system of equations it is necessary
to apply a different set of assumptions.
The purpose of this article is to present the results of theoretical and experimental study for developing and
validating a numerical solution model for predication the thermal behaviour of passive solar systems with
massive wall.

2. MATHEMATICAL MODEL
The simulation scheme of typical passive solar system with massive wall is shown in Fig. 1. The massive wall
is usually mounted on the south facade of the house. It comprises three layers: a transparent cover (one or two
glasses or plastic plates), a massive wall (masonry, concrete) and air gap between transparent cover and
massive wall. At the bottom and the top of the massive wall, there are vents for allowing an air circulation
between air gap and room space. The external transparent cover transmits solar radiation in, but holds back
heat. Wall surface is painted black at its outer side and act as an absorber of solar radiation. It stores heat from
the day and releases it with time to the room space by radiative heat transfer. The air layer between
transparent cover and massive wall is heated in contact with the wall surface, rises and circulates towards the
room (when the vents are open).
 z In summer, inhabitants close the vents and air
circulation is stopped during the day. During the
night, vents in upper end of glazing can be
opened and natural ventilation in room space
can be organized.
Air gap The thermal analysis of such system is very
Massive Wall complicated. Mathematical model is based on
the transient performance of the system. It
Rooml comprises energy balance equations, written for
each element of the system. Since the wall is
taken to be large (in comparison with wall
Air circulation
thickness), the temperature variation in y -
d direction (wall width) will be neglected, and
Glass covers only two-dimensional problem can be
x
considered - the hight (z – direction) and
Vent
thickness (x – direction). The governing energy
Fig.1. Scheme of ‘Trombe Wall’ System conservation equation of heat transfer in
massive walls is:
∂T
= a( ∂ 2 + ∂ 2 )
2 2
T T
, (1)
∂τ ∂x ∂z
where T refers to the temperature in the wall, τ is time variable, x and z – space variables and a is a material
constant. The boundary conditions, needed for the solution of eqn (1), are derived from an energy balance for
elements of passive system. The cover glazing is assumed to absorb no solar radiation and only exchanges
heat by convection and radiation with the wall surface and the ambient. Heat transfer rate in elements of the
massive wall is given by:
at x = δ (an inner wall surface)
dT
λ = h cRoom ( T w 0 - T r ) + h rRoom ( T w 0 - T wall ) (2)
dx
at x = 0: (an outer wall surface)
dT (3)
-λ = q s ( τα ) e + h cgap ( T ag -T wn ) + h rgap (T g2 - T wn )
dx
at inner glass cover:

hcgap ( T ag - T g 2 )+ hrgap ( T w0 - T g 2 ) = h12 ( T g 2 - T g 1 ) (4)

at outer glass cover:

(5)
hc∞ ( T g 1 - T a )+hr ∞ ( T g 1 - T sky ) = h12 ( T g 2 - Tg1 ) ,

where Twn,Tw0 are temperatures of wall surfaces (x=δ and x=0)

Tag, Tg1, Tg2, Ta, Twall, Tr, Tsky - temperatures of air in the gap, glass covers, ambient air, averages of room’s
walls, air in room and sky, respectively
hc.. - Convective transfer coefficients, W/m2oK,
hr.. - Radiant transfer coefficients, W/m2oK,
h12 - heat transfer coefficient in space between glass covers, W/m2oK.
Next the air gap is considered. A differential formulation, which includes terms due to thermal capacity and
convective heat transfer from the wall surface and glazing cover to the air, leads to the equation:
dT ag
ρ G ag c p = ( hcgap ( T w0 - T ag ) + hcgap ( T g 2 - T ag ))Bdz (6)
dz
where ρ is the air density [kg/m3]; cp - heat capacity of air [J/kgoK]; Gag - air flow rate in a gap, [m3/s]; B -
wall width [m], (τα)e - an absorptance-transmittance product for the total insolation on a vertical surface
 The air flow rate for air circulation in the air cavity is determined by the average air velocity. According to
J.A.Duffie and W.A. Beckman [4], natural convection in cavity can be assessed by next expression:
2gH T m -Tr (7)
V= 2
C 1 ( Ag / Av ) + C 2 T m

where C1 and C2 are constants that depend on hydraulic characteristics of the gap,
Ag,Av - gap area and vent area [m2] ; H - wall height, Tm, Tr - average temperature of air in gap and
temperature of air in room.
System (1) – (7) is unsteady two-dimensional mathematical model of massive wall system. The model has a
combined system of algebraic and differential equations as a boundary condition. This model is rather
difficult to solve due to the complicate thermal and mass transfer processes in system. The main problems
which arise during solving processes are:
- Temperature variation in direction z (according to the wall height) requires solving all equations (2)
– (7) with respect to this variation. Hence, it is necessary to consider vertical temperature variations in the
glasses and gap. This can be done only by including new heat transfer equations (such as equation (1)) for
glass covers. This temperature variation is caused mainly by air circulation in the cavity (air convection).
- The circulation in the air cavity depends on the value of air velocity and therefore, on the
temperature difference in down and upper part of the air gap - eqn. (7). At the same time, the temperature
difference depends on the convective transfer coefficient hcgap, which is function of the air velocity. This
determines the mathematical model as nonlinearly and requires special iteration procedures for solving the
heat and mass transfer equations.
- The complicated form of equations for boundary conditions presumes many difficulties in trying to
solve the mathematical equations by regular numerical scheme. This goal is complicated additionally by
unregulated variation of ambient climatic parameters (ambient temperature and solar radiation).
From the point of view of engineering application, for the simulation model of passive solar system with
massive wall the following assumptions can be grounded:
- The literature review [1,2,5] and calculations, we carried out, showed that the heat transfer by
convection with air circulation is up to 10 - 15% of all heat transfer in massive wall. This heat transfer
determines very small vertical temperature variation in the wall (0.2 - 0.5 oC), because of the temperature
equalization by heat conductivity and the large heat capacity of the wall. On the base of these results, the
model can be simplified to one-dimensional one by assuming the different layers of wall construction to be at
uniform temperature at any given time. In this way, equation (1) can be simplified to one-dimensional
problem, referred only to variable x.
- Thermal characteristics of the massive wall and the air flow in cavity are considered as constants,
because of small temperature variation in thermal and mass transfer processes. Convection and radiation heat
transfer coefficients in energy balance equations are treated as depending on velocity and temperature
difference between corresponding elements.
- Lastly, an iterative calculation process can be organized, if equation (6) has solved separately, by
using uniform temperatures of wall surface and glass cover (a numerical method for solving differential
equation can be used with considering these temperatures known). The solution will determine the air
temperature variation in the air cavity. Separate finite difference method in z direction has been used for
solving the equation (6). On the base of received temperature rise in the duct, equation (6) can be rewritten as
algebraic scheme using mean air temperature in the air duct:
-
ρ G ag c p T m T r = hcgap ( T wn - T ma ) + hcgap ( T g 2 - T ma ) (8)
HB
where T m +T r
T ma =
2
and Tm and Tg are the air temperatures in bottom and top of the air gap. After solving the main task (1), (2),
(3), (4), (5) and (8) and receiving values for temperatures of wall surface and inner glass, the equation (6) can
be solved again and all procedure can be repeated. Process can be continued until sufficient accuracy is
arrived.
 3. FINITE DIFFERENCE APPROXIMATION
Outter
glass
Inner
glass Air gap The finite-difference form of differential
equation (1) is derived by integration over
control volume surrounded the typical node i, n
in solution grid (Fig.2). The indexes i and n
refer to the thickness (x) and the time (τ)
variable, respectively. An implicit time
approximation, which is stable for forward
integration in time, is developed for transient
n+1 i,n+1
differential equations. In this case a set of
i,n
n
i,n-1 i,n+1 simultaneous equations needs to be solved at
each time step. If the time interval is named ∆τ
i,n-1
n-1 = (τn, τn+∆τ) = (τn, τn+1), the time derivative
can be written using forward Euler formula for
discretization:

∂T n +1 - n (9)
i-1 i+1 I = T i T i
1 2 3 X =0
∂τ ∆τ
Fig.2. The mesh in time and space For the space derivative is applied symmetrical
Cranc-Nicolson’s scheme [3] for discretization:
∂T
2 n+1 n+1 n+1 n n n
= σ T i+1 - 2 T i + T i -1 + (1 - σ ) T i+1 - 2 T i + T i -1 (10)
∂ x2 ∆ x2 ∆ x2
where σ is a weight coefficient.
After substituting (9) and (10) in (1) and rearranging, the following general approximation is received:
∇ x2
n+1
T i -1 - (2 + ) T in+1 + T in++11 = F in , i = 1,2,..I (11)
σ∇ τ a in
n 1-σ ∆ x2 n
where : Fi = ( T in-1 - 2 T in + T in+1 ) +
Ti
σ σ∆τ ain
This is a system of (I-2) algebraic equations with I (i = 1,2....I) unknown node temperatures (with upper index
n+1). Temperatures with upper index n are considered as known, which are received from calculation, made
in former time step or from initial conditions in the first time step. Equation (11) yields a special diagonal
matrix of equations for the time layer n+1 at each space point. Boundary conditions (2), (3), (4), (5) and (8)
must be added to complete the system. These equations include a new unknown temperatures (Tag, Tg1, Tg2),
which requires preliminarily solving the boundary equations system.
Because of complicated nature of boundary conditions and numerous difficulties, which appear in solution
process, a new procedure for completing the algebraic system of equations is proposed. The mesh is
expanded by adding three new space layers, numbered 1, 2 and 3, as it is shown on Fig.2. These layers
correspond to the elements of Trombe wall - two glass covers and air in the gap with temperatures Tag, Tg1,
Tg2. This means that, to the algebraic system (11), it is necessary to add algebraic equations (3), (4) and (8).
As boundary conditions in this new system must be used only equations (2) and (5). In this way, mathematical
task becomes considerably easier to solve, because of simplified boundary condition system.

4. SOLUTION PROCEDURE
Equations (11) can be solved by standard algebraic methods. Because of simple form of algebraic system
(11), the well-known procedure with twofold calculation passage in the space direction of mesh is used [8].
This method is applicable for algebraic systems in form as follow:
(12)
Ai T i -1 + B i T i + C i T i+ 1 = F i
where i = 1,2........I-1, whit boundary conditions:
T0 = a0T1+ b0 and TI = aI TI-1+ bI . (13)
This algebraic system is similar to our problem (11), (2), (3), (5) and (8) with respect to the unknown
temperatures (superscript index n+1). In the next considerations the upper indexes of variables in finite
difference equations can be omitted for simplicity.
Solution for above system is wanted in form as follow:
Ti = αiTi+1 + βi , (14)
where ai and bi are unknown coefficients. This equation can be written for all indexes i, including i-1:

Ti-1 = αi-1 Ti + βi-1 (15)

Substituting (14) in (12) and rearranging, it can be received next formulas for coefficients:
Ci F i - Ai β i -1
αi= , βi= (16)
Ai α i -1 + B i A i α i -1 + B i
If it is take into account that for i = 0: α0 = a0 and β0 = b0 (from boundary condition eqn.(13)), coefficients αi
and βi can be calculated by using recurrent equation (16). This is the first calculation passage in the space
direction of the solution grid. Two equations are available for the last node I of the grid: second part of
equation (13) and equation (14) for I-1 node: TI-1 = α I-1 TI + βI-1 .
From these two equations, it is possible to determine TI:

a I β I -1 + b I (17)
T I =
1 - a I α I -1
Knowing TI, it can be made second calculation passage through the
solution grid by using recurrent equation (14) to calculate temperatures Ti (i = I-1, I-2......0) of all nodes of
the grid. This is a ordinarily procedure for solving the problem in time step. Receiving the temperatures for
time step n+1, it is possible to make next time step.
This simple procedure for solving the algebraic system of finite difference approximation can not be used
directly for the mathematical problem of massive wall system described above. System of equations (2), (3),
(4) and (8), which is added to main system (11), is not fully compatible with the system (12). Equation (4)
comprises four unknown temperatures - two temperatures of glass covers Tg1 and Tg2, temperature of air in
gap Tag and surface temperature Twn. This means, that equation (4) can not be expressed in form as the
equation (14).
To use a technique similar to mentioned above procedure for solving equations (12), a modified procedure
can be used. Instead of eqn (15), a new temperature function (with three consecutive temperatures) can be
considered:
Ti = αι Ti+1 + γiTi +2 + βi (18)
This means that, tree unknown coefficients αι, βi and γι, must be calculated in first calculating passage. After
doing similar steps as it has been made above (equation 16), equations for unknown coefficients αι, βi and γι
can be easier to receive. Here, these coefficients will not be described for the common case, but for the special
case of mathematical model for passive solar system with massive wall. After rearranging equations of
boundary conditions in standard form (equation (12)), with respect to temperature function (18) and indexes
in numerical grid (fig.2), the following equations for coefficient α, β and γ have been received:
- for outer glass cover – equation (5). It can be rewritten in form like eqn.13:
Tg1 = α1. Tg2 + β1, (19)
where the coefficients with index i=1 are:
h 12 , β1= h c∞
α1= Ta, γ1= 0 (20)
h 12 + h c ∞ h 12 + h c ∞
- for inner glass cover – equation (4). Substituting Tg1 from (19) and rearranging, the equation (4) can be
transformed in form as eqn.18 with coeficiens (i=2):
h cgap h12 β 1 hrgap
α2 = , β2= γ 2=
h12 (1 - α 1 ) + h gap h12 (1 - α 1 ) + h gap h12 (1 - α 1 ) + h gap
 (21)
,
where hgap = hcgap + hrgap
- for air gap – equation (8). Using analogous transformation, it can also be received equation in form as
eqn.18 with coefficients (i=3):
(22)
BHh cgap ( γ 2 + 1) 2 ρ Gc p BT r + BHh cgap β 2
α3 = , β3= , γ3=0
Dx Dx

where Dx = BH hcgap (g2+1) + 2 r G cp B

- for wall surface – equation (2). Here is necessary to approximate the x direction derivative of temperature
by finite difference. The appropriate coefficients for algebraic equation are: (i=4):
λσ /∆x F x - σ h rgap α 2 β 3 - σhrgap β 2 - σ h cgap β 3
α4 = , β4= , γ4=0 (23)
Ex Ex
where:
λσ λ∆x
E x = σ h rgap α 2 α 3 + σ h rgap γ 2 + σ h cgap α 3 + σ h cgap - -
∆x 2 a 4n ∆ τ
n n
T 5 -T 4 ∆ x T 4n
F x = λ (1 - σ ) +λ - q s (τα ) e - (1 - σ ) h cgap ( T 3n - T n4 ) + h rgap ( T n2 - T n4 )
∆x 2 a 4 ∆τ
n

- for ordinary wall layer - Here is valid the standard transformation (16) for equation (11) (i):
1
αi= , β i = ( β i - 1 - F in )α i , γ i = 0 (24)
∆ x2
2+ - α
σ∆τ a in-1
i -1

Temperature of the last wall's layer (inner wall surface) can be defined by following equation:
σλ
x β I −1 - FF
n+1 ∆ (25)
TI =
ρ c p ∆x σλ
σλ
σ h ROOM + + - α I −1
∆x ∆τ ∆x

where (1 - σ )λ n ∆x n
FF = σ hROOM T r + (1 - σ )( T nI +3 - T r ) hROOM + ( T I +3 - T nI +2 ) - ρ c p T I +3
∆x ∆τ
and hROOM = hcROOM + hrROOM
Knowing temperature TIn+1, other node
Ambient temperature
Temp. of inner surface temperatures can be calculated with the
Temp. of middle layer recurrent formula (18). It is possible, because
Temp of outside surf.
Temp. of inner glass coefficient γ for ordinary wall layers is zero.
40 This is the second calculation passage on the
35 grid.
30 To solve the mathematical problem by
Temperature [C]

25 proposed algorithmic scheme, the heat transfer

20 coefficients and the air velocity in gap, must be
15 known in advance. These can be calculated
10 with regarding temperatures and other needed
5
variables from previous time-step calculation or
from initial conditions in starting the
0
calculations. For improving the precision of
-5
numeric calculations, an iterative calculation
-10 have been organised until needed accuracy has
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hours arrived.
Fig.3. Temperature distribution in elements of
massive wall
5. NUMERICAL EXAMPLES
To verify the applicability of the above-proposed technique, a large number of numerical examples have been
carried. In Fig.3 and 4 is shown example of calculations for passive solar system with massive wall. Climatic
data are for Sofia, Bulgaria. Passive system is with south facing concrete wall with dimensions: height - 3 m;
width - 3.5 m and thickness - 0.3m. Climatic data (solar radiation and ambient temperature) are for February.
The set of month’s daily distribution of solar radiation and ambient temperature, estimated in hour-by-hour
period have been used. Five days period of simulation calculations was needed to exclude influence of initial
conditions. Following parameters are shown in Fig. 3 and 4: ambient temperature Ta, solar radiation qs, inner
glass temperature Tg2, outer surface
3500
Solar radiation temperature Twn, wall temperature in middle
layer Twm, inner surface temperature Tw0, heat
Heat load
3000
losses to the ambient air qt, heat transferred by
Heat by convection qf, heat transferred by radiation qr.
radiation
2500
Heat by 6. CONCLUSIONS
convection
In this paper, has been presented a
Energy [w]

2000 mathematical model for passive solar system

with massive wall. A finite difference solution
1500 scheme, based on implicit method was
developed for solving the combined system of
1000
algebraic and differential equations. A new
solution procedure, suitable for the presented
mathematical model was suggested.
500
Computer program for simulation calculations
was created, and with extensive numerical
0
experiments, the applicability of presented
1 3 5 7 9 11 13 15 17 19 21 23 model was verified.
Hours
The results from presented mathematical model
Fig.4. Energy balance in massive vall
would help researches in field of passive solar
systems to increase their knowledge and
experience for thermal and mass transfer processes. The designers of passive solar systems can use this model
to select optimal constructive parameters of the massive wall.

REFERENCES
[1] Akbarzadeh A., W.S.Charters and D.A. Lesslie, Thermocirculation characteristics of a Trombe wall passive test cell.
Solar Energy, 28(6), 461-468 (1982).
[2] Borgers T.R, H.Akbari, Free convective turbulent flow within the Trombe wall channel. Solar Energy, 35(1), 31-43
(1985).
[3] Cranc J., P.Nicolson. A practical method for numerical evaluation of solution of partial differential equationsof
heat-conduction type. Proc.Cambridge Philos.Sos.,1974, N 43
[4] Duffie J.A. and W.A. Beckman, Solar Energy Thermal Processes. Wiley, New York (1974)
[5] Duffin R.J., G.Knowles, A simple design method for the Trombe wall. Solar Energy, 34(1), 69-72 (1985).
[6] Duffin R.J., A passive wall design to minimize building temperature swings. Solar Energy, 33(3/4), 337- 432
(1984).
[7] Knowels T.R. , Proportioning composites for efficient thermal storage walls. Solar Energy, 31(3), 319- 326 (1983)
[8] Richtmayer R.D.,K.W. Morton, Difference method for initial-value problems. Interscience Publishers, 1967.