Você está na página 1de 11

G Model

ENB-2826; No. of Pages 11 ARTICLE IN PRESS


Energy and Buildings xxx (2010) xxx–xxx

Contents lists available at ScienceDirect

Energy and Buildings


journal homepage: www.elsevier.com/locate/enbuild

A criterion for the assessment of the reliability of ASHRAE conduction transfer


function coefficients
Giuseppina Ciulla ∗ , Valerio Lo Brano, Aldo Orioli
Dipartimento di Ricerche Energetiche ed Ambientali, University of Palermo, Viale delle Scienze edificio 9, 90128 Palermo, Italy1

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a new mathematical approach, which, when applied to conduction transfer functions
Received 14 January 2010 (CTFs) of a multi-layered wall, is able to predict the reliability of building simulations. This new procedure
Received in revised form 4 March 2010 can be used to identify the best set of CTF coefficients, which are a critical point at the core of the
Accepted 14 March 2010
ASHRAE calculation methodology founded on the transfer function method. To evaluate the performance
of different CTF coefficient sets, the authors performed a large number of thermal simulations on the
Keywords:
multi-layered walls included in the ASHRAE Handbook, volume fundamentals, and on other walls typical
CTF coefficients
of Mediterranean building heritage. Those data were employed to test an algorithm able to assess the
ASHRAE heat balance method
Thermal building simulation
reliability of the simulations. The numerical results show that it is possible to select the optimal number
of coefficients on the basis of the size of the poles of a CTF. The proposed criterion, which employs a
pole threshold value, is highly accurate, fast and easy to integrate in the most diffuse building simulation
tools.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction objects, can modify the time the peak of the space cooling load
occurs and its magnitude.
Dynamic thermal simulation is a very important tool for sizing As well known by the designers of the ancient buildings placed
heating or air-conditioning systems and for exploring the passive all over the Mediterranean area, the thermal inertia of the room
thermal behaviour of buildings. Heating and air-conditioning sys- envelope is an effective resource that can give benefits to eco-
tem design requires reliable heating and cooling load calculations nomics, energy saving and comfort, especially during the summer
because they can significantly affect the cost of building construc- season.
tion, the operating cost and the energy consumption. Moreover, These important topics must be accurately evaluated in order
especially for buildings concerned with intensive human activity, to avoid employing safety factors in the load calculation results,
badly sized heating and air-conditioning systems can lower the which could lead to improperly sized heating and air-conditioning
comfort and the productivity of the occupants. For the above rea- systems. The artistic and cultural worth of historic and monumental
sons, ASHRAE focuses on non-residential cooling and heating load buildings, which are usually very thick and massive, should not be
calculations [1]. disfigured by the presence of heating and air-conditioning systems,
The heat balance (HB) method is the procedure recommended which can be uselessly oversized because of unreliable heating and
for reliable results because it takes account of surface-by-surface cooling load calculations.
conductive, convective and radiative heat exchanges for each room The accurate assessment of the dynamic thermal behaviour of
surface and of the heat balance for the room air. buildings requires the analysis of many thermal phenomena:
The HB method also considers that the energy absorbed by walls,
floors, furniture, etc., affects the room cooling load only after a time • The heat conduction through exterior walls, roofs, ceilings, floors
lag because there is always a delay between the time the solar radi- and interior partitions;
ation enters the room, or an energy source is turned up, and the • the solar radiation through transparent surfaces;
consequence of its presence. The thermal inertia of all the com- • the latent and sensible heat generated in the space by occupants,
ponents of the room envelope, as well as those of the contained lights and appliances;
• the ventilation and infiltration of outdoor air and other miscella-
neous heat gains.
∗ Corresponding author. Tel.: +39 091236118; fax: +39 091484425.
E-mail address: ina@dream.unipa.it (G. Ciulla). Heat conduction, which is the slowest phenomenon, mainly
1
http://www.dream.unipa.it. affects the time response of the building. Even though it can be

0378-7788/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.enbuild.2010.03.012

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
2 G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx

Nomenclature
 discrete time, s
am , bm , cm , dm elements of the transmission matrix of the m layer thermal conductivity, W/(m K)
mth layer m layer density, kg/m3
A indoor surface area, m2 Фj CTF coefficient
A(s), B(s), D(s) overall wall conduction transfer matrix ele-
ments Subscripts
A0 , An , Bn Fourier coefficients j current CTF number
bn , cn , dn transfer function coefficients nx, ny, nq number of CTF coefficients
cm layer thermal specific heat, J/(kg K)
C layer thermal capacity, J/(m2 K)
C0 , C1 residuals related to the origin double pole due to the studied through many different approaches, two computational
input signal, W/m2 , W/(s m2 ) methods have been widely used: the finite difference and transfer
d, n coefficients of Den(z), Num(z) function (TF) methods. Because of the computational time advan-
Den(z), Num(z) denominator and numerator of G(z) tage, the conduction transfer method was selected for the HB
f(t) generic periodic signal method described by ASHRAE.
Fo Fourier number The wall conduction process is formulated using conduction
G(s), G(z) generic system Laplace/z-transfer function transfer functions (CTFs), which relate the conductive heat fluxes
hi , ho indoor and outdoor surface convection coefficients, to current and past surface temperature and past heat fluxes. For a
W/(m2 K) given wall, the conductive inside heat flux qki, and the conductive
I(t), O(t) time-dependent generic input and output outside heat flux qko, are [1]
I(s), O(s) input and output Laplace transform
I(z), O(z) input and output z-transform 
nz 
ny

Lm layer thickness, m qki, = −Z0 Tsi, − Zj Tsi,(−j) + Y0 Tso, + Yj Tso,(−j)


N sample number of P j=1 j=1
Np number of poles of truncated transfer function
Ns number of layers

nq

− j qki,(−j) (1)
ok , o(t) output time-dependent signal, W/m2
pn generic nth pole, h−1 j=1

pn∗ generic pole giving stabilised PME, h−1


p∗n−1 pole preceding pole p∗n , h−1 
ny

nx
qko, = −Y0 Tsi, − Yj Tsi,(−j) + X0 Tso, + Xj Tso,(−j)
pth threshold pole value, h−1
pth, threshold pole value for time step , h−1 j=1 j=1

P period of f(t), s

nq
PME percentage mean error − j qko,(−j) (2)
qki, qko, conductive inside and outside discrete heat fluxes,
j=1
W/m2
qCTF
i,
, q F
i,
CTF and Fourier sampled conductive inside heat where Xj = outside CTF coefficients, j = 0, 1, . . ., nx; Yj = cross CTF
fluxes, W/m2 coefficients, j = 0, 1, . . ., ny; Zj = inside CTF coefficients, j = 0, 1, . . ., nz;
qso (s), qsi (s) Laplace transforms of conductive outside and ˚j = flux CTF coefficients, j = 0, 1, . . ., nq;  = discrete time; ı = time
inside heat fluxes step; Tsi = inside-face temperature; Tso = outside-face temperature;
rn residual of nth pole, W/m2 The FTs are
 nx  nz 
R layer thermal resistance, (m2 K)/W qko  X z −j q  Zz −j
qki 
j=0 j j=0 j
Ri , Ro indoor and outdoor surface resistances, (m2 K)/W  =  nq − ki  = nq 
Tso Tsi =0  z −j Tsi Tso =0 j z −j Tso Tsi =0
Sie thermal structure factor j=0 j j=0
s Laplace variable  ny −j
q  Yz
j=0 j
t continuous time variable, s = − ko  = nq (3)
te, sol–air temperature at time , K Tsi Tso =0 j z −j
j=0
trc indoor air temperature, K
TH, TH% threshold value The effectiveness of the z-transfer function in thermal dynamic
Ti (s), To (s) inside and outside air temperature Laplace trans- simulations was first claimed by Stephenson and Mitalas [2,3] who
forms studied one-dimensional transient heat transfer through multi-
Tsi , Tso inside-face and outside-face temperature, K layer walls, roofs, and floors. Mitalas and Arsenault [4] wrote the
Tsi, , Tso, inside-face and outside-face discrete temperature, first program for computing transfer function coefficients. Hittle
K and Pedersen [5] presented a detailed derivation of transfer func-
Tsi (s), Tso (s) inside-face and outside-face temperature tions for multi-layer slabs based on the method of Peavy [6].
Laplace transforms Other procedures for performing CTF calculations have been
U overall heat transfer coefficient, W/(m2 K) developed with the aim of improving the method accuracy and of
z z-transform variable reducing the time consumption. Ouyang and Haghighat [7] pre-
Zj , Xj , Yj CTF coefficients sented a procedure based on the state-space principle, and Wang
and Chen [8] proposed a frequency-domain regression method,
Greek symbols which has been recently improved [9].
˛m layer thermal diffusivity, m2 /s It is well known that for a wall surrounded by fluids whose
ı time step, s temperature changes over time, the CTF coefficients should be
mathematically infinite but, as the absolute value of them decreases

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx 3

very quickly with increasing order of the addendum, it is easy to default setting, or users do not have enough experience in making
perform a truncation procedure to select the finite numbers nx, ny the wisest decision. Sometimes the coefficients used are not enough
and nq. for the required accuracy, and sometimes their excessive number
Anyway, in order to accomplish a correct truncation procedure causes the increase in the computational time.
it is necessary to evaluate the effect on the numerical response This paper aims to define a tool useful to assess whether a given
produced by the insertion or elimination of a single coefficient, and set of CTF coefficients, matched with a defined time step, will lead
it can be consciously executed only if a great number of coefficients to a reliable simulation or not. Obviously, such a tool should be able
of each order are available. Some of the most widely used software, to predict the reliability of a simulation without performing it and
like HVACSIM+ [10], TRNSYS [11] and EnergyPlus [12,13], employ using the least amount of computation.
truncated CTFs to perform accurate simulations of complex thermal As it will be shown in the following, the goal can be reached by
systems and design cooling load calculations. observing the size of the poles of CTFs in order to verify whether a
Although the above methods and software are widely used to benchmark was surpassed. The poles are the roots of the equation
predict the hourly energy performance of building envelopes, the B(s) = 0, and B(s) is the denominator that appears in Eq. (A4). This
reliability of this approach cannot be always claimed. Actually, procedure performs its task very accurately and can be easily nested
some researchers focused on the assessment of the numerical sta- into building simulation programs in order to automatically fit the
bility of transfer functions and the influence of the time step and number of the CTF coefficients with the selected time step.
thermal inertia.
Spitler and Fisher [14] already pointed out that computational 2. The assessment of CTF coefficients
inaccuracy sometimes occurs in calculating the dynamic behaviour
of thermal systems. Kosny and Kossecka [15] asserted that in most In order to describe how with a mathematical approach it is
of the whole building thermal modelling computer programs, for possible to predict the reliability of a set of CTF coefficients, a brief
several structural and material configurations of building envelope introduction to the problem theory is necessary.
components containing high thermal mass, the one-dimensional As shown in Appendix A, the generic CTF can be written in the
analysis could generate serious errors in building loads estima- form:
tion. The inaccuracy of such computer models can cause incorrect
energy estimations and improper sizing of the heating and cooling Num(z) n0 + n1 z −1 + n2 z −2 + n3 z −3 + · · · + nNp z −Np
G(z) = = (4)
equipments. For a wall composed of 20% framing materials and 80% Den(z) 1 + d1 z −1 + d2 z −2 + d3 z −3 + · · · + dNp z −Np
insulating materials, DOE-2, TRNSYS or EnergyPlus have an error in
the calculation of thermal performances that may exceed 44%; for It is known that the number Np of the employed coefficients
concrete framing the errors may exceed 27%. should assure that the following stationary condition is respected
Dos Santos and Mendes [16] employed many numerical meth- by each CTF:
ods for integrating the governing differential equations in the air Np
n
domain and assessed the results in terms of accuracy and time com- kNp k = U (5)
puting. They also showed the influence of the simulation time step k
dk
on room air temperature and humidity profiles within the building
were U is the overall heat transfer coefficient of the wall or roof. This
envelope.
relation can be exactly verified only by using an infinite number of
Li et al. [17] presented an analysis of the applicability of calcu-
coefficients. Actually, it easy to find that a set of coefficients, whose
lation methods of CTFs and indicated many sources of error in CTF
last values are near to zero, does not verify the relation (5) though
computation. Using a 24-h outside sinusoidal temperature, they
the CTF allows a very accurate evaluation of the thermal behaviour
evaluated the heat flux on the inside surface of some walls, whose
of a wall or roof. This is the reason why relation (5), which is never
thickness and thermal properties varied between a large range of
affected by the time step, is almost useless for defining the number
values, and they claimed that the frequency-domain regression
of CTF coefficients adequate to perform a good simulation.
method was the most accurate.
On the contrary, it is important to observe that the time step
The authors have already pointed out the reliability of CTFs
ı plays an important role because it directly affects both the coef-
in the case of walls characterized by a high thermal inertia, typ-
ficients of the numerator and those of the denominator of G(z).
ical of traditional South-European buildings [18–21], and have
Actually, if ı is changed, a different set of CTF coefficients will be
assessed the numerical stability and sensitivity analyses of the
obtained, and G(z) will change.
time step and thermal inertia on the prediction of thermal per-
Observing the expression of the denominator, as obtained with
formance of walls. They observed that, because the conduction
Eq. (A9):
transfer functions become progressively more unstable if the time
step is decreased, the simulation of buildings that have a time

Np
response much longer than the selected time step, can yield phys- Den(z) = (1 − e−pn ı z −1 ) (6)
ically meaningless results.
n=1
With the aim of investigating some critical points at the core of
the ASHRAE calculation methodology, the authors have calculated it is easy to state that, as the time step increases, each term e−pn ı
the CTF coefficients of multi-layered walls and compared the results decreases, and the corresponding quantity 1 − e−pn ı tends towards
with those obtained by the Fourier periodic steady state algorithm. unity.
Though a heuristic approach to the problem may suggest that a For a pole pn there is a value of ı that makes the pole contri-
wall with a high thermal inertia would need a CTF containing many bution irrelevant to evaluating Den(z), and this is more effective
coefficients or a longer time step, the weak point is the early and the greater the value of the pole is. Obviously, for such a value of
easy determination of the number of coefficients best fitting the ı the contributions of the poles greater than pn will vanish, which
selected time step or vice versa. Actually, before starting with the is equivalent to neglecting the effect of the poles farthest from the
calculation of the HB, it is sensible to know if the selected number origin, which allows the system to follow the fastest changes in the
of CTF coefficients will be able to adequately represent the thermal input signal.
behaviour of a given wall, roof or floor. Unfortunately, the number More complicated is the way the time step affects the numerator
of coefficients cannot be always selected, because it is a software coefficients because it is also necessary to take into account the

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
4 G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx

traditional expression [1], chapter 28, as the output signal:


 
  dn (qi,−nı ) 
qi, = A bn (te,−nı ) − − trc cn (9)
A
n=0 n=1 n=0

where qi, = heat gain through wall or roof, at calculation


hour , A = indoor surface area,  = discrete time, ı = time step,
te,−nı = sol–air temperature at time  − nı, trc = constant indoor
room temperature, and bn , cn , dn = transfer function coefficients.
The heat gain was evaluated using the sol–air temperature listed
in [22] (chapter 28, Table 1) and the room indoor temperature
trc = 24 ◦ C.
The authors carried out a comparison between simulation data
obtained from truncated CTFs and those obtained from the Fourier
periodic steady state algorithm. This algorithm was chosen because
Fig. 1. Shapes of e−pn ı for different poles. it is based on a mathematical basis quite different from the z-
transform, and this is very important to avoid using a self-referring
validation criterion.
terms of Eq. (A11): In a periodic steady state any time-discrete function f(t), known
in correspondence to the N instants in which the period P was

Np
subdivided, can be approximated by the finite Fourier series:
ok = o(t)t=kı = C1 + C0 kı + rn e−pn kı (7)
n=1
A0 
N/2  2n
2n

f (t) = + An cos t + Bn sin t (10)
Again for the generic term rn e−pn ı
it is possible to find a value of 2 P P
n=1
the time step ı that makes the effect of the pole pn negligible. This
value is affected by the size of the related residual rn , and rn is not where
always greater then rn+1 . Because the residuals of the poles farthest
from the origin are on average smaller than the nearest ones, they 2
N kP
2
N kP
2kn

A0 = f An = f · cos
do not have an effect contrary to what was observed before about N N N N N
the influence of the time step. k=1 k=1

Considering, for sake of simplicity, Den(z), Fig. 1 shows the


2
N kP
2kn

shapes of the quantities e−pn ı versus ı for some ordered poles: Bn = f · sin (11)
In order to evaluate, for a given time step, the threshold pth N N N
k=1
above which the poles pn > pth can be neglected, the value TH, or its
percentage value TH%, that makes e−pn ı negligible has to be defined: The exact mathematical solution of the heat flux caused by
TH% ln[TH%/100] a sinusoidal outside temperature variation and a constant inside
e−pth ı ≤ → pth ≥ − (8) temperature can be easily calculated, but to perform an accurate
100 ı
evaluation, a great number of harmonic components has to be used,
The next step is the definition of TH% that still assures reliable and consequently the period P has to be divided by a number N
calculations. Table 1 lists some values of pth evaluated with the bigger than 24, which is the value related to the 1-h sampled data.
previous relation for different time steps. In order to know this large amount of outside temperature val-
With the aim of verifying the effectiveness of the pth approach, ues it is necessary to use an interpolation rule able to reproduce
the authors applied it to the walls and roofs contained in [1] (chap- the physical signal that originated the discrete input. This goal was
ter 18, Tables 16 and 17), for various numbers of poles and using suitably reached by means of a cubic spline, which is a special func-
time steps of 1–3 h. tion defined piecewise by third order polynomials. The interpolated
The CTFs, or better the TF coefficients for the presence of the sol–air temperature is shown in Fig. 2.
convection exchange coefficients on outside and inside faces, were After calculating the time responses to each harmonic input
calculated by performing each possible combination between the component, for the linearity of the system, the response to the
number of poles and coefficients, including those in the range from generic f(t) can be obtained by their summation, and this response,
1 to 20. which is mathematically exact in periodic steady state conditions,
To better take into account the simultaneous presence of more can be used to benchmark the accuracy of the heat gain calculated
than one TF, the authors used the heat gain evaluated with the by Eq. (9).
The evaluations were performed using N = 240 (120 harmonic
components) and, in order to get the periodic steady state response
Table 1
Values of pth evaluated with the relation (8).
of the system, the calculation with CTFs was repeated 300 times
until the output signal was stabilised [23].
Time step [h] TH% The comparison was based on a percentage mean error param-
0.01% 0.05% 0.10% 0.50% 1.00% eter PME:
 
0.1 92.103 76.009 69.078 52.983 46.052  CTF F 
1  qi, − qi,
24/ı
0.2 46.052 38.005 34.539 26.492 23.026
PME = F
× 100 (12)
0.5 18.421 15.202 13.816 10.597 9.210 24/ı q
1.0 9.210 7.601 6.908 5.298 4.605 =1 i,
2.0 4.605 3.800 3.454 2.649 2.303
3.0 3.070 2.534 2.303 1.766 1.535 where qCTF
i,
and qFi, are the sampled heat fluxes at the generic dis-
6.0 1.535 1.267 1.151 0.883 0.768 crete time  calculated using the truncated CTFs and the Fourier
12.0 0.768 0.633 0.576 0.442 0.384
method, respectively.

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx 5

Table 2
Simulation results for the wall no. 19.

Pole order Pole ı=1h ı=2h ı=3h

No. coeff. PME% No. coeff. PME% No. coeff. PME%

9 −9.12750 9 0.02068
8 −6.02402 S 0.02067 8 1.39032
7 −3.73320 7 0.23373 7 1.39032 7 1.99745
6 −3.09120 6 94.41167 6 1.38930 6 1.99745
5 −2.96103 5 10213.20 5 2.57568 5 1.99894
4 −1.49474 4 334.63210 4 6.40238
3 −0.41113 3 985.99793
2 −0.31422
1 −0.01923

Brick wall no. 19: F01-M01-F04-I01-M16-F04-G01-F02. No. of poles employed in CTF calculation: 10.

Table 3
Simulation results for the wall no. 19 (linear interpolation).

Pole order Pole ı=1h ı=2h ı=3h

No. coeff. PME% No. coeff. PME% No. coeff. PME%

9 −9.12750 9 0.00460
8 −6.02402 8 0.00460 8 0.14920
7 −3.73320 7 0.23259 7 0.14920 7
6 −3.09120 6 94.41066 6 0.14933 6 0.29545
5 −2.96103 5 10212.97 5 1.21043 5 0.29545
4 −1.49474 4 340.66503 4 8.80457 4 0.29552
3 −0.41113 3 1033.21007
2 −0.31422
1 −0.01923

Brick wall no. 19: F01-M01-F04-I01-M16-F04-G01-F02. No. of poles employed in CTF calculation: 10.

The above defined PME provides an indication of the accuracy of As shown in Table 2 the PMEs decrease with the amount of
the qCTF
i,
, which were calculated by varying the number of poles, CTF coefficients used. On the contrary, a decrease of only two or three
coefficients, time step and the wall thermo-physical characteristics. coefficients gives rise to a dramatic increase in the PME, and a num-
ber of coefficients less than a particular threshold value gives a
completely unreliable result. Furthermore, it is possible to observe
3. Considerations for the simulations performed
a sort of asymptotic trend of the PME. Actually, increasing the num-
ber of coefficients does not increase the quality of the calculation:
The effectiveness of the pth approach can be initially verified
with ı = 1 h, 8 coefficients are enough, while with ı = 2 or ı = 3, 6
by examining the results of the simulation of the test wall no. 19
or 5 coefficients are sufficient. The stabilised value of the PME also
shown in Table 2.
changes with the time step.
In the second column the first nine poles of the CTFs are reported
In the CTF calculation performed with the equations described
in the same order (column 1) in which they are placed starting from
in Appendix A, because of the use of a unit linear ramp input sig-
the origin. In columns 3 and 4, for the time step of 1 h, the PMEs
nal, the sol–air temperature is divided into time steps where the
of the simulations and the number of coefficients used to perform
variation is represented by a linear profile. Therefore, the PME is
each of them are given. The figures in the columns from 5 to 8 have
erroneous because qCTF i,
and qFi, are evaluated with input signals as
the same meaning but are referred to the time steps of 2 and 3 h.
more different the larger the time step is.
Comparing Figs. 2 and 3 it is easy to see that the shape of the
sol–air temperature employed by the Fourier algorithm (spline) is
quite different from the ones used with the CTF calculations (linear
interpolation).
To avoid this mismatch the PME was evaluated again using the
qFi, calculated with the input signals of Fig. 3 obtained by the lin-
ear interpolation of the sol–air temperature values referred to the
selected time step.
As shown in Table 3, the PMEs are now less than the previous
ones, and therefore the stabilised trend in Table 2 means that the
CTF calculation has stably reached its maximum accuracy and that
a further increase in the number of coefficients will not yield any
improvement.
It is important to underline that the procedure for calculating
the CTF coefficients described in Appendix A allows the evaluation
of the nth coefficient only if the nth pole is known. Although the nth
pole slightly affects all the other n − 1 coefficients, the precision of
the simulation mainly depends on the number of coefficients used.
Anyway, if the first n coefficients are evaluated using a number of
Fig. 2. Sol–air temperature. poles greater than n, the accuracy will slightly improve. To obtain

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
6 G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx

Fig. 3. Linear interpolations of sol–air temperature.


Fig. 5. Distribution of p∗n−1 and p∗n poles for all ASHRAE walls and roofs, time
step = 2 h.
the best accuracy it is useless to use a large number of poles because
the maximum can be already reached using n + 1 or n + 2 poles.
Table 4
Tables 2 and 3 show that the stabilised PMEs are reached
The value TH% that makes e−pn ı negligible.
with p∗n = −6.024, p∗n = −3.733 and p∗n = −3.091 for time steps of
1–3 h, respectively. On the contrary, with the same time steps the Time step pth, TH%
PMEs are still not stabilised with p∗n−1 = −3.733, p∗n−1 = −3.091 and 1 5.80 0.30
p∗n−1 = −2.961. 2 2.90 0.30
The above observations have a general validity because they can 3 1.95 0.29

be extended over the whole set of walls and roofs examined. The
results of all simulations are summarised in Tables B1 and B2 of
separates all p∗n−1 from all p∗n of every wall and roof. These val-
Appendix B, which also contain the value of the pole p∗n that gives
ues can be used by inverting Eq. (8) to evaluate for each time step,
the stabilised PMEs and the value of the pole p∗n−1 that precedes p∗n .
and the corresponding values of TH% = 100e−pth ı are reported in
In order to verify that p∗n−1 and p∗n can be considered discrim-
Table 4.
inating parameters and to state how to link the time step ı to an
Table 4 shows that TH% does not change with the time step,
effective pth value, the values of p∗n−1 and p∗n in Tables B1 and B2
and for each wall and roof the value of the pole threshold pth above
were plotted with the relative time steps. The results are shown in
which the poles pn > pth can be neglected is defined by the following
Figs. 4–6.
expression:
Observing Figs. 4–6 it easy to state that, for each time step ı,
there is a value pth, , indicated with a dashed line, that almost ln[0.003]
pth = − (13)
ı

Fig. 4. Distribution of p∗n−1 and p∗n poles for all ASHRAE walls and roofs, time step = Fig. 6. Distribution of p∗n−1 and p∗n poles for all ASHRAE walls and roofs, time
1 h. step = 3 h.

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx 7

Table 5
Simulation results for the wall no. 24 with varying thickness.

Pole order Pole ı=1h ı=2h ı=3h

No. coeff. PME% No. coeff. PME% No. coeff. PME%

13 −7.63211 13 1.19309
12 −6.48414 12 1.19309
11 −5.43164 11 1.19309
10 −4.47429 10 1.19527 10 0.98050
9 −3.61171 9 3.22583 9 0.98050
8 −2.84345 8 212.88994 8 0.98055 8 0.82086
7 −2.16896 7 0.96606 7 0.82086
6 −1.58755 6 18.02936 6 0.82179
5 −1.09839 5 853.00822 5 7.89573
4 −0.70053 4 465.61466
3 −0.39284
2 −0.17417
1 −0.04347

Concrete block wall no. 24. Thickness changed from 0.2032 to 0.68 m; l/FoSie = 1206. No. of poles employed in CTF calculation: 14.

With the aim of checking the validity of Eq. (13) the authors the thickness, the thermal conductivity, the density and the specific
performed several tests on building elements much more massive heat of the slab until 1/(Fo · Sie ) reached the value of 1200. The
than the previously employed ASHRAE walls and roofs (Tables 5–8). following tables report the results of simulations performed with
Following the approach of Li et al. (S) the Fourier number and linear interpolation of the sol–air temperature.
the thermal structure factor for a single-layered slab: Results show a maximum PME of 2.51%, which is much less than
those declared by Li et al. [17], which reached values close to 100%.
ı 1 R Ro
Fo = Sie = + i2 (14) Actually, in [17] it is shown that the error calculated with direct
RC 6 R root-finding (DRF) method is extremely variable when increasing
where R = layer thermal resistance, C = layer thermal capacity, Ri , the quantity 1/(Fo · Sie ) over the value of 600. This could suggest
Ro = indoor and outdoor surface resistances were used to represent that the DRF method is extremely unstable because a small increase
with the quantity 1/(Fo · Sie ) the heavyweight characteristics of a in one of the parameters, like the thickness, can also produce
wall. Referring to the ASHRAE concrete block wall no. 24 that has unreliable results. The authors never came across this unreliabil-
1/(Fo · Sie ) = 102.9, the authors changed separately, one at a time, ity despite thousands of simulations made in order to carry out the

Table 6
Simulation results for the wall no. 24 with varying conductivity.

Pole order Pole ı=1h ı=2h ı=3h

No. coeff. PME% No. coeff. PME% No. coeff. PME%

12 −6.86141 12 2.51244
11 −5.76372 11 2.51244
10 −4.76202 10 2.51329 10 2.21981
9 −3.85619 9 3.00274 9 2.21981
8 −3.04610 8 124.04417 8 2.21981 8 1.96300
7 −2.33164 7 5929.45536 7 2.21650 7 1.96300
6 −1.71269 6 11.75349 6 1.96988
5 −1.18916 5 5.43454 5 590.30395
4 −0.76095 4 366.67044
3 −0.42798
2 −0.19020
1 −0.04755

Concrete block wall no. 24. Conductivity changed from 0.26 to 0.0234 W/mK; l/FoSie = 1202. No. of poles employed in CTF calculation: 14.

Table 7
Simulation results for the wall no. 24 with varying density.

Pole order Pole ı=1h ı=2h ı=3h

No. coeff. PME% No. coeff. PME% No. coeff. PME%

12 −5.88280 12 2.51244
11 −4.89814 11 2.51244
10 −4.00614 10 2.51329 10 2.21981 8 1.96300
9 −3.20654 9 3.00274 9 2.21981 7 1.96300
8 −2.49896 8 124.04417 8 2.21981 6 1.96988
7 −1.88289 7 5929.45536 7 2.21650 5 590.30395
6 −1.35758 6 11.75349 4 366.67044
5 −0.92200 5 5.43454
4 −0.57466
3 −0.31335
2 −0.13457
1 −0.03263

Concrete block wall no. 24. Density changed from 464 to 5440 kg/m3 ; l/FoSie = 1203. No. of poles employed in CTF calculation: 14.

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
8 G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx

Table 8
Simulation results for the wall no. 24 with varying specific heat.

Pole order Pole ı=1h ı=2h ı=3h

No. coeff. PME% No. coeff. PME% No. coeff. PME%

13 −6.98214 13 2.51244
12 −5.90124 12 2.51244
11 −4.91349 11 2.51329
10 −4.01870 10 3.00274 10 2.21981
9 −3.21659 9 124.04417 9 2.21981
8 −2.50680 8 5929.45536 8 2.21981 8 1.96300
7 −1.88879 7 2.21650 7 1.96300
6 −1.36184 6 11.75349 6 1.96988
5 −0.92489 5 590.30395 5 5.43454
4 −0.57646 4 366.67044
3 −0.31433
2 −0.13499
1 −0.03273

Concrete block wall no. 24. Specific heat changed from SSO to 10,285 J/kg K; l/FoSie = 1203. No. of poles employed in CTF calculation: 14.

analysis. It is probable that the cause of the mismatched results using a self-referring validation criterion to check the reliability
declared by Li et al. is in an inadequate choice of the number of of a procedure by comparison with a different one, which is con-
poles and of CTF coefficients. Actually, it is difficult to guess that for sidered more accurate. For this reason the authors compared the
the wall no. 24, with 1/(Fo · Sie ) = 1200, 8–9 CTF coefficients would results obtained from the CTF coefficients, which are based on the
give an error greater than 100%. The previous tables show that by z-transform, with those obtained from the Fourier periodic steady
using one more coefficient, the error of the simulation decreases to state algorithm. They also considered the effect of the interpola-
3%, which is near the best achievable accuracy. A different cause for tion of the input signal due to the presence of the linear ramp in
the mismatched results could be found in the missing poles in the the traditional procedure of CTF coefficients.
numerical search, which is traditionally considered the weak point The pole identification is considered a weak point in the DRF
of the DRF method, though it is possible to verify the loss of a pole method, but, on the other hand, there are known methods to verify
[24], and the search of few poles can be performed very accurately the accidental loss of a pole. Moreover, the knowledge of the value
because of the velocity of modern computers. Moreover, once the of the last pole that has to be found allows the use of a filtering
value of the last pole that has to be found is known via Eq. (13), it is procedure with a very narrow bandwidth.
also possible to perform an extremely selective search of the poles
using a filtering procedure with a very narrow bandwidth. Appendix A.

The one-dimensional heat conduction through a multi-layered


4. Conclusions wall is shown in Fig. A1.
It is described by the following equations [25]:
The ASHRAE Handbook, volume Fundamentals, employs the   Ns    
 Tso (s)    am bm   Tsi (s) 
HB method for sizing heating or air-conditioning systems and for  =     (A1)
exploring the passive thermal behaviour of buildings. The reliabil-  qso (s)   cm dm   qsi (s) 
m=1
ity of the HB method is affected by the correct calculation of the
CTF coefficients. Several studies state that the optimal choice of where Tso (s), Tsi (s) = outside-face and inside-face temperature
the number of poles, time step and number of coefficients is very Laplace transforms (LTs), qso (s), qsi (s) = heat fluxes on outside and
important, especially when the CTF method is applied to a massive on inside-face LTs, s is the Laplace variable, and the elements am ,
wall. There is a clear need for a reliable criterion able to guide the bm , cm , dm of the transmission matrix of the mth layer are
designer in the optimal choice of the above mentioned parameters 
 sinh(Lm s/˛m )
in order to perform the most accurate simulation. am = dm = cosh(Lm s/˛m ) bm = 
In this work a new procedure to evaluate the performance and m s/˛m
the reliability of CTFs was presented. In order to identify this pro-  
cedure the authors calculated the CTFs for all 54 walls and roofs cm = m s/˛m sinh(Lm s/˛m ) (A2)
included in the ASHRAE Handbook, volume Fundamentals, and the
heat fluxes were compared with those ones obtained from the
Fourier periodic steady state solution. Results clearly show that a
threshold value of poles used in the CTFs calculation can be consid-
ered a discriminating parameter, and a rule to identify this value
was extrapolated. To check the validity of this rule the authors per-
formed some additional tests on the ASHRAE wall no. 24 using the
Fourier number and thermal structure factors to vary the thermal
characteristics of the wall.
The analysis of the simulations performed on the modified wall
no. 24 confirmed that the sets of CTF coefficients, related to the
preliminary identification of the threshold pole value, lead to very
reliable results with errors lower than 3%. The presented algorithm
shows that the error of simulations based on the DRF method is con-
tained within an acceptable range of variation if the correct number
of poles and coefficients is chosen. It is also important to avoid Fig. A1. Sketch of the wall.

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx 9

with ˛m = m /m cm and m , m , cm , Lm the layer conductivity, den- If I(t) is a unit linear ramp, then
sity, specific heat and thickness.
C1  rn
Np
If the convective exchanges have to be considered, Eq. (A1) 1 C0
becomes I(s) = O(s) = + + (A6)
s2 s 2 s s + pn
n=1
    Ns      
 To (s)   1 1/ho    am bm   1 1/hi   Ti (s) 
 = ·   · ·  where pn is the nth pole of G(s), rn the corresponding residual, and
 qso (s)   0 1   cm dm   0 1   qsi (s)  C0 and C1 are the residuals related to the origin double pole due to
m=1
    the input signal. Np is the number of poles of the truncated TF.
 A(s) B(s)   Ti (s)  Following the traditional procedure [3], Eq. (A5) can be rewrit-
=  ·  (A3)
C(s) D(s)   qi (s)  ten in the z-transform (ZT) domain:

O(z) = I(z)G(z) (A7)


where To (s), Ti (s) = outside and inside air temperature LTs, ho , hi
convection exchange coefficients on outside and on inside-face. where [26]
Eq. (A1) can be also written
O(z) Z[I(s)G(s)]
G(z) = =
1 A(s) D(s) 1 I(z) Z[I(s)]
qsi (s) = Tso (s) − T (s) qso (s) = Tso (s) − T (s)
B(s) B(s) si B(s) B(s) si  
(A4) z(1 − z −1 )  rn
Np
Num(z)
= C0 + C1 − (1 − z −1 ) =
ı 1 − e−pn ı z −1 Den(z)
n=1
and the quantities 1/B(s), A(s)/B(s), D(s)/B(s) are the wall conduction
(A8)
transfer functions (CTFs).
More generally, the output O(s) of a thermal system solicited by and
an input I(s) is

Np

O(s) = I(s)G(s) (A5) Den(z) = (1 − e−pn ı z −1 )


n=1
where G(s) is the system TF, and I(s) and O(s) are the LTs of the
= 1 + d1 z −1 + d2 z −2 + d3 z −3 + · · · + dNp z −Np (A9)
time-dependent input I(t) and output O(t), respectively.

Table B1
Values of the pole p∗n that gives the stabilised PMEs and of the preceding pole p∗n−1 , for ASHRAE walls.

No. order Walls Wall order ı1 ı2 ı3

p∗n−1 p∗n p∗n−1 p∗n p∗n−1 p∗n

1 1 −7.29506 −4.95040 −4.95040 −1.66598 −4.95040 −1.66598


2 Curtain walls 2 −20.43528 −5.74495 −5.74495 −1.67021 −5.74495 −1.67021
3 3 −6.07057 −1.78678 −6.07057 −1.78678 −1.78678 −1.54226

4 4 −12.56997 −4.48481 −4.48481 −1.71770 −4.48481 −1.71770


5 5 −6.91330 −1.77817 −6.91330 −1.77817 −1.77817 −1.40228
Stud walls
6 6 −10.38324 −2.05113 −10.38324 −2.05113 −2.05113 −1.26071
7 7 −6.84404 −1.79375 −6.84404 −1.79375 −1.79375 −1.37927

8 8 −25.19918 −5.75585 −5.75585 −4.33879 −4.33879 −1.15767


9 Eies 9 −11.52602 −4.27788 −5.58289 −1.82232 −5.58289 −1.82232
10 10 −5.51905 −4.40335 −4.27788 −2.83084 −2.83084 −1.07975

11 11 −9.14807 −5.58289 −4.40335 −2.95687 −2.95687 −1.13099


12 12 −9.35529 −3.54340 −3.54340 −2.29410 −2.29410 −1.64003
13 13 −9.12755 −2.97325 −2.97325 −1.81837 −2.97325 −1.81837
14 14 −9.12385 −4.87847 −2.96474 −1.64761 −2.96474 −1.64761
15 15 −9.16340 −4.75533 −3.12713 −1.79615 −1.79615 −1.47218
Brick walls
16 16 −8.82046 −4.59542 −2.96415 −1.37843 −2.96415 −1.37843
17 17 −8.42089 −2.96901 −2.96901 −2.40719 −2.40719 −0.32379
18 18 −7.03321 −4.30314 −3.15435 −2.95997 −2.95997 −1.71627
19 19 −6.02402 −3.73320 −3.09120 −2.96103 −2.96103 −1.49474
20 20 −7.56203 −3.52780 −3.11225 −1.76379 −3.11225 −1.76379

21 21 −10.52556 −5.53765 −5.53765 −2.04648 −2.04648 −1.72020


22 22 −6.05368 −3.12392 −3.12392 −1.77080 −1.77080 −1.13108
23 23 −8.39239 −4.06176 −4.06176 −1.77561 −1.77561 −1.42062
Concrete block
24 24 −6.73739 −3.67373 −3.67373 −1.57774 −3.67373 −1.57774
25 25 −6.73435 −4.04965 −4.04965 −2.63119 −2.63119 −1.21054
26 26 −6.73314 −5.11792 −3.58392 −2.55310 −2.55310 −1.46107

27 27 −8.13120 −2.83038 −2.83038 −1.91278 −2.83038 −1.91278


28 28 −8.03726 −2.78562 −2.78562 −1.75603 −2.78562 −1.75603
29 29 −7.76847 −2.75945 −2.75945 −2.17291 −2.75945 −2.17291
30 30 −6.11212 −5.54125 −3.59306 −1.72020 −3.59306 −1.72020
Precast and cast in
31 31 −7.43230 −4.44591 −4.44591 −2.24963 −2.24963 −1.75526
place concrete walls
32 32 −6.83070 −4.95326 −4.95326 −3.21608 −3.21608 −0.87838
33 33 −7.88192 −3.73571 −3.73571 −1.76553 −1.76553 −1.18729
34 34 −6.21758 −4.35392 −3.57174 −1.74293 −1.74293 −1.58494
35 35 −6.29498 −3.69600 −3.69600 −1.82520 −1.82520 −0.65898

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
10 G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx

Table B2
Values of the pole p∗n that gives the stabilised PMEs and of the preceding pole p∗n−1 , for ASHRAE roofs.

No. order Walls Wall order ı1 ı2 ı3

p∗n+1 p∗n p∗n+1 p∗n p∗n+1 p∗n

36 1 −5.80032 −3.03587 −5.80032 −3.03587 −3.03587 −1.25655


37 2 −5.99707 −3.97974 −5.99707 −3.97974 −3.97974 −1.84208
38 3 −13.53207 −5.56243 −5.56243 −2.65860 −5.56243 −2.65860
Sloped frame roofs
39 4 −8.87800 −3.50026 −3.50026 −1.30935 −3.50026 −1.30935
40 5 −12.90055 −4.79475 −3.34628 −1.32777 −3.34628 −1.32777
41 6 −8.14235 −3.46636 −8.14235 −3.46636 −3.46636 −1.29734

42 7 −6.59791 −2.84560 −2.84560 −2.01608 −2.84560 −2.01608


43 Wood deck 8 −7.20562 −6.16817 −6.16817 −2.31363 −2.31363 −1.91529
44 9 −8.94976 −4.90029 −4.90029 −1.99377 −4.90029 −1.99377

45 10 −6.97431 −2.26512 −6.97431 −2.26512 −2.26512 −1.22432


46 11 −7.11610 −3.25861 −3.25861 −1.70698 −3.25861 −1.70698
47 12 −9.35859 −4.99825 −4.99825 −2.26469 −2.26469 −1.03591
Metal deck roofs
48 13 −6.76837 −3.96009 −3.96009 −2.48155 −3.96009 −2.48155
49 14 −7.18096 −3.78773 −3.78773 −2.11259 −2.11259 −1.72202
50 15 −7.05511 −3.89141 −3.16379 −1.79336 −3.16379 −1.79336

51 16 −6.99016 −4.18898 −3.61285 −1.94955 −1.94955 −1.71827


52 17 −6.08131 −3.75708 −3.75708 −1.82647 −1.82647 −1.57931
Concrete roofs
53 18 −7.59209 −3.84869 −3.35970 −1.79047 −3.35970 −1.79047
54 19 −5.82047 −5.03113 −5.03113 −2.13223 −2.13223 −1.59966

After some manipulations, the coefficients of Den(z) can be writ- only Np + 1 numerator coefficients, and again the kth pole affects all
ten as the other coefficients.

d1 = −[h1 + h2 + h3 + · · · + hNP ]
Appendix B.
d2 = [h1 h2 + h1 h3 + · · · + h2 h3 + h2 h4 + · · · + hNP −1 hNP ]
Tables B1 and B2.
d3 = −[h1 h2 h3 + h1 h2 h4 + · · · + h2 h3 h4 + h2 h3 h5 + · · ·

+hNP −2 hNP −1 hNP ] Appendix C. Supplementary data

··· Supplementary data associated with this article can be found, in


dNP = (−1)NP [h1 h2 h3 h4 · · ·hNP ] (A10) the online version, at doi:10.1016/j.enbuild.2010.03.012.

where hk = e−pk ı . The evaluation of the generic coefficient dk is pos- References


sible only if the kth pole is known, and this one affects all the other
k − 1 coefficients. [1] ASHRAE Handbook, Volume Fundamentals, 2009.
The TF numerator Num(z) can be calculated with the following [2] D.G. Stephenson, G.P. Mitalas, Cooling load calculations by thermal response
factor method, ASHRAE Transactions 75 (1) (1969) 246–271.
relation: [3] D.G. Stephenson, G.P. Mitalas, Calculation of heat conduction transfer functions
Den(z) z(1 − z −1 ) for multilayer slabs, ASHRAE Transactions 77 (2) (1971) 117–126.
Num(z) = O(z) = (1 + d1 z −1 + d2 z −2 + · · · [4] G.P. Mitalas, J.G. Arsenault, Fortran IV program to calculate z-transfer functions
I(z) ı for the calculation of transient heat transfer through walls and roofs Use of
Computers and Environmental Engineering Related to Buildings, vol. 39, NBS
z(1 − z −1 )
+ dNp z −Np) · O(z) = (1 + d1 z −1 + d2 z −2 + d3 z −3 Building Science Series, Gaithersburg, MD, 1972.
ı [5] D.C. Hittle, C.O. Pedersen, Calculating building heating loads using fre-
quency response of multi-layered slabs, ASHRAE Transactions 87 (2) (1981)
+ · · · + dNp z −Np )(o1 z −1 + o2 z −2 + o3 z −3 + · · · + ok z −k 545–568.
[6] B.A. Peavy, A note on response factors and conduction transfer functions,
+ · · ·) = n0 + n1 z −1 + n2 z −2 + n3 z −3 + · · · + nNp z −Np ASHRAE Transactions 84 (12) (1978) 668–690.
[7] K. Ouyang, F. Haghighath, A procedure for calculating thermal response fac-
+ n(Np +1) z −(Np +1) (A11) tors of multilayer walls state-space method, Building and Environment 26 (2)
(1991) 173–177.
[8] S. Wang, Y. Chen, Transient heat flow calculation for multilayer constructions
where
using a frequency-domain regression method, Building and Environment 38

Np (2003) 45–61.
[9] X. Xu, S. Wang, Y. Chen, An improvement to frequency-domain regression
ok = o(t)t=kı = C1 + C0 kı + rn e−pn kı (A12) method for calculating conduction transfer functions of building walls, Applied
n=1 Thermal Engineering 28 (2008) 661–667.
[10] C. Park, D.R. Clark, G.E. Kelly, HVACSIM + building systems and equipment sim-
Eventually, if the numerator is truncated at the Np -th coefficient, ulation program: Building loads calculation, NBSIR 86-3331, National Bureau
G(z) can be written in the well known form: of Standards (1986).
[11] S.A. Klein, W.A. Beckman, TRNSYS—A Transient System Simulation Program,
Num(z) n0 + n1 z −1 + n2 z −2 + n3 z −3 + · · · + nNp z −Np Solar Energy Laboratory, University of Wisconsin-Madison, 1994.
G(z) = = [12] R.K. Strand, F. Winlelmann, et al., Enhancing and extending the capabilities
Den(z) 1 + d1 z −1 + d2 z −2 + d3 z −3 + · · · + dNp z −Np of the building heat balance simulation technique for use in EnergyPlus, in:
(A13) Proceedings of Building Simulation’99, Kyoto, Japan, 1999.
[13] R.K. Strand, C.O. Pedersen, D.B. Crawley, Modularization and simulation tech-
niques for heat balance-based energy and load calculation programs: the
The coefficients of Num(z) can be also calculated using some experience of the ASHRAE Loads Toolkits and EnergyPlus, in: Building Sim-
ulation 2001, IBPSA, Rio de Janeiro, 2001.
combined expressions [18] like Eq. (A10), which are not reported [14] J.D. Spitler, D.E. Fisher, Development of periodic response factor for use with
here for sake of conciseness. Using Np poles it is possible to evaluate the radiant time series method, ASHRAE Transactions 105 (2) (1999) 491–502.

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012
G Model
ENB-2826; No. of Pages 11 ARTICLE IN PRESS
G. Ciulla et al. / Energy and Buildings xxx (2010) xxx–xxx 11

[15] J. Kosny, E. Kossecka, Multi-dimensional heat transfer through complex build- [20] G. Beccali, M. Cellura, V. Lo Brano, A. Orioli, Is the transfer function method
ing envelope assemblies in hourly energy simulation programs, Energy and reliable in a European building context? A theoretical analysis and a case
Buildings 34 (5) (2002) 445–454. study in the south of Italy, Applied Thermal Engineering 25 (2–3) (2005)
[16] G.H. dos Santos, N. Mendes, Analysis of numerical methods and simulation 341–357.
time step effects on the prediction of building thermal performance, Applied [21] G. Beccali, M. Cellura, V. Lo Brano, A. Orioli, Single thermal zone balance solved
Thermal Engineering 24 (8–9) (2004) 1129–1142. by Transfer Function Method, Energy & Buildings 37 (12) (2005) 1268–1277.
[17] X.Q. Li, Y. Chen, J.D. Spitler, D. Fisher, Applicability of calculation methods for [22] ASHRAE Handbook, Volume Fundamentals, 1997.
conduction transfer function of building constructions, International Journal of [23] M.G. Davies, On the basis of the environmental temperature procedure, Build-
Thermal Sciences 48 (2009) 1441–1445. ing and Environment 13 (1978) 29–46.
[18] C. Giaconia, A. Orioli, On the reliability of ASHRAE conduction transfer function [24] D.C. Hittle, R. Bishop, An improved root-finding procedure for use in calculating
coefficients of walls, Applied Thermal Engineering 20 (1) (2000) 21–47. transient heat flow through multilayered slabs, International Journal of Heat
[19] M. Cellura, L. Giarrè, V. Lo Brano, A. Orioli, Thermal dynamic models using and Mass Transfer 26 (11) (1983) 1686–1693.
Z-transform coefficients: an algorithm to improve the reliability of simu- [25] H.S. Karslaw, J.C. Jaeger, Conduction of Heat in Solids, Clarendon Press-Oxford
lations, in: Building Simulation 2003, IBPSA International Conference and Science, 2003.
Exhibition on Building Simulation, Eindhoven, The Netherlands, 11–14 August, [26] E.I. Jury, Theory and Application of the Z Transform Method, John Wiley & Sons
2003. Inc., 1964.

Please cite this article in press as: G. Ciulla, et al., A criterion for the assessment of the reliability of ASHRAE conduction transfer function
coefficients, Energy Buildings (2010), doi:10.1016/j.enbuild.2010.03.012

Você também pode gostar