Energy and Buildings 126 (2016) 477–484

Contents lists available at ScienceDirect

Energy and Buildings

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

Evaluating the reliability of multi-energy source buildings: A new

analytical method for considering the dynamic behavior of thermal
loads
Mohammad-Hossein Shariatkhah a,∗ , Mahmoud-Reza Haghifam a ,
Mohesn Parsa-Moghaddam a , Pierluigi Siano b
a
Faculty of Electrical and Computer Engineering, Tarbiat Modares University, Tehran, Iran
b
Department of Industrial Engineering, University of Salerno, Fisciano, Italy

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

Article history: Developments in energy conversion technology and multi-carrier energy systems have led to higher
Received 20 March 2016 flexibility of energy systems, thus improving the reliability of supply and decreasing the energy costs.
Received in revised form 13 May 2016 Usually in previous studies on reliability evaluation of Multi-Energy Source Building (MESB), the required
Accepted 14 May 2016
energy for supplying the loads is considered to be fixed during each time interval and the ability of the
Available online 18 May 2016
system to supply the loads is evaluated on the basis of this assumption. In this paper, due to the high
amount of the thermal loads in a building’s energy portfolio, the flexibility of these loads is addressed and
Keywords:
an analytical method is presented to model the dynamic behavior of thermal loads in reliability analyses of
Reliability
Multi-energy source building
MCEBs. The proposed method is based on Markov chain concepts integrating thermodynamic equations.
Thermodynamic loads In addition, in order to evaluate the validity of the proposed method, a Monte-Carlo simulation method
Markov model is employed. The methods are simulated on test systems and the results are presented and discussed.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction suggested for installing in energy system [7]. In Ref. [8] the results
of a technical and of a cost analysis of two different types of ther-
The most demanded energy by consumers is in one of the forms mal energy storage systems (hot water thermal energy storage
of electricity, gas or heat. Traditionally, the required infrastructures systems and latent heat thermal energy storage) for residential
to supply each type of energy were built and utilized separately [1]. micro-CHP plants are presented. In addition, developing technol-
Environmental constraints, the power market restructuring, the ogy for converting different forms of energy into others can increase
need to improve the system efficiency and the demand for high the flexibility of an energy system if the interaction between differ-
reliability supply have increased the utilization of distributed gen- ent forms of energy in the studies is considered [9]. On this basis, the
eration sources [2]. Developing new technologies such as fuel cell, concept of multi-carrier energy systems and energy hubs has been
combined heat and power (CHP) and micro-turbines have increased introduced. The presented studies in this field have tried to model
the use of small-scale distributed generation sources and devices the interactions taking place “inside” the multi-carrier energy sys-
that can convert one form of energy into electricity [3,4]. Reference tems as well as the interactions with the “outside” world, at both the
[5] deals with the effects of the heat dumping on the operation of operational and the planning stage [10]. Reference [11] presented
four different residential micro-CHP systems. In [6] optimal sizing a model that avoids some limitations identified for the conven-
of distributed energy resources in medium voltage or low volt- tional energy hub model in which a mathematical model has been
age micro-grids according to different criteria is carried out. In formulated with the use of graph and network theory.
order to accommodate high penetration rate of non-dispatchable A Multi-Energy Source Building (MESB) represents an energy
renewable energy sources, different type of energy storages are system where energy carriers at the input are converted into dif-
ferent forms of energy by converters to supply loads at the output.
The converters include combined heat and power resources, trans-
∗ Corresponding author.
formers, batteries, heaters and coolers and so on [12–15]. A MESB,
E-mail addresses: m.shariatkhah@modares.ac.ir (M.-H. Shariatkhah),
due to the possibility of the substitution of the energy may increase
haghifam@modares.ac.ir (M.-R. Haghifam), parsa@modares.ac.ir the reliability and efficiency [16,17]. Previous researches on MESB
(M. Parsa-Moghaddam), psiano@unisa.it (P. Siano). mainly concentrated on the modeling and on the evaluation of the

http://dx.doi.org/10.1016/j.enbuild.2016.05.043
0378-7788/© 2016 Elsevier B.V. All rights reserved.
478 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484

employed to find the optimal power dispatch in smart grids with

Nomenclature distributed energy resources. In [24], the authors have addressed
the flexibility of thermal loads and, by employing Monte Carlo sim-
Indices
ulation method, the impact of the loads dynamic behavior on the
˛, ˇ,ω,  Indices for the input and output energy carrier,
system reliability indices has been shown.
respectively
On the basis of previous researches, in this paper, a novel
i, j, k Indices to show the order of system states
approach for evaluating the reliability of a MESB due to the flex-
iter Index to show the order of iterations
ibility of the thermodynamic loads is presented and discussed. The
n Index to show the order of the system elements
novel contributions of the proposed method are presenting a new
(converters)
analytical method based on Markov chain is presented in order to
sys Index for system
model the system reliability considering a high portion of ther-
D Index to show order of time
mal and cooling loads in a building energy demand. The proposed
method is based on basic theoretical equations and can be devel-
Water Heater Parameters
oped for modeling the reliability of other energy systems.
a Thermal resistance of tank walls (kW/◦ C)
In order to demonstrate the effectiveness of the proposed
ce Specific heat constant of water (kWh/m3 ◦ C)
method, the presented method is simulated on two test systems
Ce Tank thermal capacity (kWh/◦ C)
and the results are compared to the results of Monte Carlo method.
q Hot water rate of extraction (m3 /h)
The rest of the paper is organized as follows: in Section 2, the
Ta Ambient temperature (◦ C)
dynamic behavior of loads is presented and the necessity of carrying
TT (t) Water temperature at time t (◦ C)
out the proposed research is discussed. In Section 3, the modeling
Td Desired temperature for outlet water (◦ C)
approach of different parts of the problem is presented and the
Tin Inlet water temperature (◦ C)
proposed method is detailed by combining thermodynamic equa-
Tout Outlet water temperature (◦ C)
tions with the Markov chain concepts. In Section 4, Monte Carlo
Tmin /Tmax Minimum/maximum limitation of water temper-
simulation method is briefly presented as a method to evaluate
ature (◦ C)
the proposed method. Simulation results on some test systems are
PWH Power supply of the water heater (kW)
shown in the Section 5, and finally in Section 6 some conclusions
PWHreq Required power to maintain water temperature at a
are presented.
determined temperature (kW)

Energy hub and Reliability Modeling Parameters 2. Problem description

˛ˇ Capacity of the connection from ␣ to ␤
ki Transition rate from state k to i 2.1. Investigating the inflexible and flexible loads
n Failure rate of connection n (failure per year- f/y)
n Repair rate of connection n (failure per year- r/y) For several loads in a building the energy demand should be sup-
n,i , ¯ n,i Status of connection n plied by various carriers, such as electricity, gas and heat. Therefore,
mi Mean time of staying in state i a desirable performance is defined for each of these loads con-
pi Probability of state i sidering their characteristics. The loads can be divided into two
ps Ratio of success categories: inflexible loads and flexible loads. Inflexible loads refer
Fki Frequency of entering into each state k from a state to the loads in which, along with energy connection (or disconnec-
i tion), their performance changes instantaneously from a desirable
to an undesirable status (or vice versa) [24]. Lighting loads are an
N number of the system elements (converters) example of inflexible loads (Fig. 1). In computing the reliability
Rn , Qn Availability and unavailability of connection n indices, for any period of time a fixed amount of energy is con-
C Coupling matrix sidered that if it is supplied, the status of the system is supposed as
E Input power vector success and if it does not, the status is supposed as failure.
L Output power vector In contrast, flexible loads refer to the loads in which, along with
energy disconnection (or connection), their performance does not
change simultaneously from a desirable to an undesirable condi-
tion (or vice versa). Thermal loads are the most important example
economic advantages of these systems [12–17]. In [18], the exergy of this kind of loads which may act such an energy storage due to
principles in the context of supply systems are applied to achieve its thermodynamic equations. For example, in the case of a water
rational use of energy resources in a MESB by taking into account heater, the water temperature is defined within a specified range
the different quality levels of energy resources as well as those of
building demands. As the advantage related to their possibility to
increase the reliability of energy systems has not been completely
studied in previous researches it is discussed in the following.
In [1], a method for assessing the reliability of a multi-carrier
energy hub has been provided. Moreover, in [19], the system effi-
ciency and safety have been considered in the planning study of
an energy hub. In these two references, the system loads have been
assumed fixed during the time intervals and, based on that, the reli-
ability indices have been calculated. This is while, the flexibility of
loads is another parameter in energy systems which can improve
the reliability and reduce the system costs. Many studies have
considered the loads flexibility (especially thermal loads) and the
supply costs have been minimized by scheduling and controlling
these loads [20–22]. In [23], a new formulation of shiftable loads is Fig. 1. The performance response of inflexible loads to energy supply varying.
 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484 479

Fig. 2. The performance response of flexible loads (thermal load) to energy supply
varying.

of desirable performance. As can be seen in Fig. 2, if the water

temperature is ideal, after a system element failure and supply
Fig. 3. Each state of Markov chain can include both the desirable performance (rep-
interruption, it will take some time for the water temperature to resented by red color in above of each state) and the undesirable performance
violate its desirable bandwidth. Considering this time as a failure, (represented by blue color in below of each state). (For interpretation of the ref-
leads to a pessimistic assessment of the system reliability. On the erences to colour in this figure legend, the reader is referred to the web version of
this article.)
other hand, if, due to the supply interruption, the water tempera-
ture is not within the limited range, after the interruption, it will
take some time for the system to get its desirable performance
(specified temperature range) (Fig. 2).
Since a large part of the load of buildings consist of thermody-
namic loads, it is required to model the effect of the flexible loads
performance on the reliability indices.

2.2. Significance of the study

Markov chain method is a common method to assess the reli-

ability of engineering systems. In application of this method for
evaluating the energy systems, the amount of available energy in
each of the Markov chain states is determined based on the oper-
ating or failed status of system elements. Using this method is
sufficient to evaluate the reliability of inflexible loads; the summa- Fig. 4. The general form of multi-energy resource building.
tion of the probability of the states with an available energy higher
than the required energy, is equal to the availability of the supply
system.
However, this method is not sufficient to evaluate the avail- in this state (Fig. 3). Therefore, it is necessary to calculate the ratio
ability of flexible loads’ performance in which the availability (or success and failure for each state of Markov chain.
unavailability) of energy is not equal to the availability (or unavail- In this paper a new method is presented for evaluating the
ability) of load performance. To consider the dynamic behavior of energy supply systems of MESB which is based on the performance
the flexible loads, beside the investigation of the available energy, it of the system loads and also considers their dynamic behavior.
is also required to consider the relationship between the load per-
formance and the supplying energy. In other words, as in Markov
model, the loads performance does not necessarily coincide with 3. Modeling approach and the presented method
the amount of supplied energy in each state, the system perfor-
mance in each state of the Markov chain can include both the 3.1. Multi-energy resource building modeling
success or failure status.
For example, in the case of hot water tank, after an element Fig. 4 demonstrates the general model of a general multi-carrier
failure and energy supply interruption, it takes some time until that energy hub, which can also represent a building. A MERB includes
the water temperature violates the desirable range of temperature. several energy carriers at the input to supply the loads at the output
Therefore, the transition of the system from a normal Markov state through corresponding energy converters (e.g. electric heater for
to a state with interrupted supply does not necessarily results in converting electricity to heat) within the building.
system performance failure. Accordingly, although the new state of In optimization analyses usually a coupling matrix is used to
the Markov chain based on the amount of available energy supply relate the loads at the output to the energy carriers at the input [15].
is considered as a failed state, based on the desired performance However, the coupling matrix is not adequate to be used in these
(here acceptable temperature) the system may experience both a analyses since in order to investigate the possibility of loads’ supply
desirable and an undesirable performance during its staying time in reliability studies, the capacity of elements and their operating
480 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484

status should be also determined. Matrix ␨ is defined as presented

in Eq. (1) to represent the capacity of the converters.
⎡ ⎤
˛˛ ˇ˛ ... ω˛
⎢ ⎥
⎢ ˛ˇ ˇˇ ... ωˇ⎥
⎢ ⎥
⎢ . ⎥
␨=⎢ . . . ⎥ (1)
⎢ . ⎥
⎢ . . . ⎥
⎣ . . . .
⎦
˛ω ˇω ... ωω

Moreover, matrices  and M are defined (with the same dimen-

sion as matrix ␨) in order to represent the failure rates and repair
rates of the converters [1].

3.2. Thermodynamic loads modeling

Regarding the nature of flexible loads, different formulations

can be used to describe the relation between their energy supply Fig. 5. The connection of each state with other states.
and their performance. In this paper, thermodynamic equations of
a hot water tank are used as a complicated sample of flexible loads. output power of each state i is determined based on the following
It is assumed that the desirable performance of a water heater equation:
is shown by a temperature bandwidth as follows:
i,ˇ

ω

Tmin ≤ TT (t) ≤ Tmax (2) PWH = ˇ,i .ς ˇ

(5)
=˛
The basic model is based on a linearized energy balance anal-
ysis and has been employed in many studies [21,22] to model the If the system has also inflexible loads, the amount of energy
variation of water temperature at each time (TT (t)). The model can power required to provide the flexible loads is equal to the output
be formulated as follows [21]: power of Eq. (5) minus the power required from inflexible loads.
Moreover, the probability of each state i is calculated as follows
dTT [26]:
Ce = −a(TT (t) − Ta ) − ce q(Tout − Tin ) + PWH (3)
dt

N
This model indicates that the variation of water temperature pi = (n,i .Rn + ¯ n,i .Qn ) (6)
during the time depends on the rate of energy entering into the
n=1
system and the rate of energy leaving the system. For this system,
the leaving energy includes the first term (a(TT (t) − Ta )) and the sec- In the above equation, n,i is a binary variable and its value is such
ond term (ce q(Tout − Tin )). Where, the first term indicates the heat that it is 1 in the state i if the converter n is operating, while, in case
exchange with the environment and depends on the thermal resis- of a failure, this value is zero. In addition, Rn is the success proba-
tance of the tank wall; and the second term indicates the missed bility for a converter n that is determined based on its failure rate
energy regarding the inflows of cold water. Moreover, the entering (n ) and repair rate (n ). Moreover, Qn,i is the failure probability of
energy includes the injected power (PWH ) for heating the water. In the converter n.
this model it is assumed that the water in the tank is perfectly mixed Step 2) For each state i, the mean time interval of staying in the
at all times and thus it can be characterized by a single temperature state is shown with mi and calculated as follows:
variable. 1
Based on Eq. (3), if the rate of entering energy is equal to the mi =  (7)
rate of the leaving energy, the variation of temperature will be zero. ik
Therefore, it is possible to calculate the required power for keeping k=1:N

the temperature in a specified amount as follows: Moreover, the frequency of entering into each state k from a
state i (Fki ), is determined based on Eq. (8).
PWHreq = a(Td − Ta ) + ce q(Td − Tin ) (4)
Fki = Pk .ki (8)
If it is assumed that Td = Tmin , the minimum energy power for
keeping the temperature at the minimum acceptable temperature Step (3) For each state, an initial performance status is consid-
will be determined. ered. For instance, in the case of the hot water tank, for each state
a specified temperature is assumed as the initial point of study.
3.3. The proposed method for modeling the dynamics of flexible Step (4) For each state, it is necessary to determine the per-
loads in the reliability analysis: an analytical method based on formance status of the system at the time of entry. To determine
repetition this, the entering frequency from other states and the performance
status of the system at those states (e.g. here water temperature)
To solve the problem, a modeling procedure based on Markov should be considered. For example, as in the case of the water tank
chain theoretical concepts [25] integrating thermodynamic equa- and for a system with three elements, the entrance to each state
tions of hot water tank is proposed. Similar to the steps of this will be possible through three states (Fig. 5). Therefore, based on
problem, other flexible loads can be considered in reliability eval- the entering frequency from each of these three conditions and on
uation problems. the water temperature in each of those states, the expected tem-
Step 1) First, the Markov chain is established based on the perature of the water at the moment of arrival to a desired state
operating or failed status of the system elements (here energy is determined. It should be considered that since in this step the
converters). Then, considering the capacity of each converter, the expected performance status is calculated, some input data may be
 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484 481

lost; Thus, based on the nature of each system, the accuracy of this In this equation, TT (t) is:
process should be investigated. The expected temperature of the i,

water at the moment of arrival to a state i can be determined by: aTa + ce qTin + PWH
TT (t) = max{
  a + ce q
iter iter−1
TT,i =( Fki TxT,k )/ Fki (9) i,

aTa + ce qTin + PWH

k=1:N k=1:N −( − TT (in))e−(a+ce q)×(t)/Ce , Tmax } (17)
a + ce q
Step (5) For each state, based on the mean time interval of stay-
ing in that state and the entering status, the expected status of the Step (6) Since the leaving temperature of states in step 5 has
system at the leaving time should be determined. For example, for been calculated based on the assumed initial temperature in step 3,
the case of hot water tank, the presented Eq. (3) is used for deter- the third through fifth steps are repeated for the updated situation
mining the variation of water temperature. If it is assumed that the in the fifth stage and this is done until the solution converges.
temperature of water in all the volume of the tank is the same and it Step (7) After that the expected temperature at the leaving time
is equal to the temperature of outlet water, Eq (3) can be rewritten of each state is determined, for each state i, the success and failure
as follows: ratio of the state can be determined. To do this, for each state i,
dTT i,
the entrance scenario from any of the other states j is checked. The
Ce = −a(TT (t) − Ta ) − ce q(TT (t) − Tin ) + PWH (10)
dt success ratio of a state i with the assumption of entering from a
To solve this equation, the general solution is determined based state j is calculated from one of Eq. (18). This equation is based on
on the roots of the homogeneous equation: the water temperature at the time of leaving the state j and entering
into the state i (TT (tD )), and it also considers if the output power of
i,

state i (PWH ) is higher or lower than the needed power PWHreq .

⎧ i,

⎪
⎪ if : TT (tD ) ≥ Tmin &PWH ≥ PWHreq → then : psji = 1
⎪
⎪
⎪
⎪
⎪
⎪
tfs
∞
⎪
⎪
⎪
⎪
i,

if : TT (tD ) ≥ Tmin &PWH ≤ PWHreq → then : psji = out,i e out,i dt + (tfs /t)out,i e−(out,i t) dt
−( t)
⎨
0 tfs (18)
⎪
⎪
∞
⎪
⎪
⎪
⎪ i,

if : TT (tD ) ≤ Tmin &PWH ≥ PWHreq → then : psji = ((t − tfs )/t)out,i e−(out,i t) dt
⎪
⎪
⎪
⎪
⎪
⎪ tfs
⎩ i,

if : TT (tD ) ≤ Tmin &PWH ≤ PWHreq → then : psji = 0

In Eq. (18), tfs is the time when from the arrival to the system
Ce S + a + ce q = 0 (11) in state i takes to reach the minimum allowable temperature limit
Td = Tmin . It is obtained as follows.
S = −(a + ce q)/Ce (12)
i,

aTa + ce qTin + PWH

And the particular solution is determined by putting the general tfs = (Ce /(−(a + ce q))) × ln(Tmin − )/
a + ce q
solution of the heterogeneous equation in the main equation and
considering the boundary conditions, as follows: aTa + ce qTin + PWH
i,

(−( − TT (tD ))) (19)

if : (tD+1 − tD ) = 0 → then : TT (tD+1 ) = TT (tD ) (13) a + ce q
i,

axa + ce qxin + PWH Thus, if a state i is connected with N other states, these calcu-
TT (t∞ ) = (14)
a + ce q lations are done for all N cases, and then, based on the entering
frequency of the state, the success ratio is calculated from the fol-
Therefore, the temperature of water after a period of time is
lowing equation:
determined as follows:
 
aTa + ce qTin + PWH
i,
psi = ( Fki × psji )/ Fki (20)
TT (tD+1 ) =
a + ce q k=1:N k=1:N

i,
Step (8) Having the probability of each state multiplied by its
aTa + ce qTin + PWH
−( − TT (tD ))e−(a+ce q)×(tDn+1 −tD )/Ce (15) success ratio, the probability of success or availability of the entire
a + ce q system according to Eq. (21) is determined.

It is worth noting that, for each of the flexible loads, the changes
N

in the system performance over a certain period of time and the pssys = pi × psi (21)
amount of energy supplied in each state of the Markov chain will i=1

be calculated by the same process.

To determine the expected temperature at the leaving time of 4. Sequential Monte Carlo method
each state s, it is required to multiply the probability of leaving the
state after duration time t (out,i e−(out,i t) ) by the system’s temper- Monte Carlo simulation is a method that models the complex
ature at that moment (TT (t)) and then, calculate its integral for a systems and considers the uncertainties [27,28], and sequential
long time interval: Monte Carlo method is appropriate to simulate the behavior of a

∞ system over time [29]. In this method, according to the probability
functions of the system’s parameters, the random behavior of the
TT s = out,i e−(out,i t) TT (t)dt (16)
system during time intervals is simulated. If the number of sim-
0 ulation time intervals is sufficiently large, the mean behavior of
482 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484

Table 1
Water tank characteristics.

Surface area (m2 ) Volume (m3 ) Weight (kg) Thermal resistance (J/min ◦ C) Power rating (kW)

Hot water tank [21] 2.8 0.304 304 24.51 5

Table 2
The results of evaluating the system based on two different approaches.

Evaluation based on performance unavailability Evaluation based on energy unavailability

Scenario 3 Scenario 2 Scenario 1 Scenario 3 Scenario 2 Scenario 1

␭ = 10 = 365 ␭ = 5 = 182.5 ␭ = 2.5 = 91.25 ␭ = 10 = 365 ␭ = 5 = 182.5 ␭ = 2.5 = 91.25
Proposed analytical method 0.0145 0.0217 0.0247 0.0267 0.0267 0.0267
Monte-Carlo 0.0159 0.0225 0.0250 0.0259 0.0265 0.0268

Table 3
Each convertor’s capacity, and their failure and repair rates in base scenario (b ,
Mb ).

Converter Capacity ␭b ␮b rb = 1/ ␮b

elec to elec 5000 2 182 48

elec to gas 0 – – –
elec to heat 2000 5 365 24
gas to elec 2500 4 365 24
gas to gas 1500 5 365 24
gas to heat 6000 4 182 48
heat to elec 0 – – –
heat to gas 0 – – –
heat to heat 2500 4 365 24

the system during the simulation time can be used as a reliable

sample of the actual behavior of the system [27]. In this paper,
to evaluate the validity of the proposed method, a Monte-Carlo
simulation method is employed. In this method, the studied time
horizon is divided into many small intervals and the variations of
temperature during each interval is determined based on the ran-
dom status of system elements. The modeling approach based on
sequential Monte Carlo simulation is summarized in different steps Fig. 6. Flowchart of Monte Carlo simulation model [24].
as described in the following [24]:
Step (1) Establish the Markov states and calculate the available
energy supply for each state using Eq. (5). and evaluate the behavior of different variables. Then, a complete
Step (2) For all states determine the probability of remaining in system with three carriers at the input and three at the output,
that state or of a transition to other states after a specific time slice including inflexible and flexible loads, is investigated.
t based on the transition rates between the states.
Step (3) Assume a state as the initial state. Then, generate a ran-
dom number between zero and one. Determine whether for the 5.1. Case 1
next period the system remains in that state or moves to another
state. A converter with a failure rate of 5 failures per year (␭ = 5 f/y),
Step (4) Assume an initial temperature for the initial state and a repair time of 48 h (repair rate  = 182.5 r/y) and a capacity of
calculate the variation of the temperature during each time slice 2000 W is used to heat a hot water tank with the specifications of
taking into account the power supply of each state. Table 1. Given the existence of just one converter, the state space
Step (5) Repeat steps 4 and 5 for a large enough number of of the Markov chain consists of two states: success, with a capacity
time intervals and record the variation in temperature during time of 2000 W, and failure, with 0 W. The proposed analytical method
intervals. (based on the performance of loads) is implemented in accordance
Step (6) Determine the period during which the temperature with the contents described in Section 3 and is compared with
has been in an acceptable or unacceptable range; then, calculate the Monte Carlo method described in Section 4. The results of the
the availability and unavailability indices. system evaluation are shown based on the availability of the allow-
The steps of the Monte Carlo simulation model are shown in able temperature (acceptable performance) in Table 2. Moreover,
Fig. 6. in order to show the need to consider the dynamic behavior of the
thermal load, the system is also evaluated based solely on the avail-
5. Simulation results ability of energy (neglecting the loads performance as the inflexible
loads in conventional studies). To do this, first, based on Eq. (4) the
A hot water tank with the characteristics shown in Table 1 is energy required to maintain the temperature at 65 ◦ C is calculated
studied as a flexible load. Allowable temperature bandwidth for this − which is 368 W- then, since this amount of energy cannot be
tank is considered between 50 and 80 ◦ C and the intended function provided in the second state (failure state), the probability of the
of the thermostat is set from 55 to 75 ◦ C. second state represents the unavailability of energy supply.
Initially, a very simple system is considered consisting of a con- Moreover, assuming that the failure rate and repair rate are
verter and a flexible thermal load (hot water tank) to observe double or half, the simulations are repeated.
 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484 483

Table 4
The obtained unavailability of loads in test system 2.

Inflexible Inflexible gas Inflexible heat Flexible heat Flexible heat

electric loads loads loads loads (based on loads (based on
energy performance
unavailability) unavailability)

 = 0.5b , M = 0.5Mb Proposed method 0.019849 0.021649 0.021505 0.010840 0.010869

Monte-Carlo 0.019850 0.021640 0.021489 0.011961 0.011109
 = b , M = Mb Proposed method 0.017823 0.021649 0.021505 0.010840 0.010870
Monte-Carlo 0.018162 0.021832 0.021670 0.010964 0.010170
 = 2b , M = 2Mb Proposed method 0.013264 0.021649 0.021505 0.010840 0.010870
Monte-Carlo 0.013476 0.021258 0.020988 0.010166 0.011647

of energy leakage occurrence, first the inflexible loads are supplied

and then the rest of energy is associated to flexible load.
Two proposed methods were implemented on the test sys-
tem and the results are given in Table 4. As can be seen, if the
hot water tank is considered as an inflexible load with a constant
energy demand, the unavailability of energy supply is obtained
nearly about 0.022, while regarding the thermodynamic behavior
of water and evaluating the system based on the desirable temper-
ature, an unavailability about 0.018 is obtained. Again, comparing
the obtained results with the results of Monte Carlo simulation for
this system the validity of the proposed method is achieved.
The simulations are also repeated for two new scenarios: halving
and doubling the repair rates and failure rates. The results indicate
that in evaluating the system based on availability of energy, the
variation of failure and repair rates is not sensitive. However, the
Fig. 7. Water temperature while leaving the second state (failure state) based the results demonstrate that the proposed method can take the impact
stay time in that state. these variations into account. The best indices are for the case with
the largest failure and repair rates (the shortest repair time).
As seen, based on the assessment of the availability of energy, by
making the failure and repair rate half or double, as the probability 6. Conclusion
of the second state based on Eq. (6) does not change, the results
have not changed. However, while doing the evaluation based on Thermal loads constitute a considerable proportion of energy
the proposed methods, given that the constant time of water tem- demand of buildings and accurate modeling of them is necessary in
perature in thermodynamic relations has not changed, but the time reliability studies. This paper addressed the impact of flexible loads
of failure or repair was doubled or halved, the results are different. in reliability analyses and presents an analytical method based on
In Fig. 7, for scenario 2, based on the results of the Monte Carlo basic concepts of Markov chain model to consider the dynamic
simulations water temperature graph when leaving the second behavior of thermal loads. To validate the proposed method, dif-
state (failure state) is shown according to the duration time of stay- ferent scenarios of the test system have been considered and the
ing in that state. As can be seen, in cases where the system in less results have been compared to those obtained by carrying out a
than 12 h was repaired (and left the second state), no violation of Monte Carlo simulation. Although comparing the results of the sim-
the allowable temperature water (65 ◦ C) was observed. Therefore, ulated test systems indicates the accuracy of the analytical method,
in these cases, even the unavailability of supply is occurred, the the execution time of the analytical method is much shorter if
availability of desirable temperature is still maintained. compared with the execution time for Monte Carlo method. The
Comparing the results of the analytical method and Monte Carlo numerical study shows that by simultaneously varying the failure
method indicates the accuracy of the analytical method for the sim- and repair rates of the system components, the indices calculated
ulated system. Simulations were performed using Matlab software. based on system performance (acceptable temperature) are better
Time needed to run the analytical method was less than a second, in the case with larger rates. However, for the indices calculated
while for Monte Carlo method, it was about an hour on a personal based on the availability of supply, the difference between these
computer, Core(TM) i3-2330 M CPU 2.20 GHz, RAM 4 GB. simulated cases cannot be distinguished.
The proposed method presented in this paper can be used in the
5.2. Case 2 planning studies of MCEBs to design the suitable size of components
of the energy supply system. The proposed modeling procedure
In this section, a test system with three energy carriers at input, can be used to exactly consider the flexibility of thermal loads (e.g.,
and three at output is considered (similar to Fig. 4). The input and water heater and HVAC systems), and a similar approach can also
output carriers include electricity, heat and gas. The capacity of be developed to model the flexibility of other flexible loads (e.g.,
each converter and the failure and repair rates of them are given electric vehicle and washing machine).
in Table 3. The output loads respectively are: the electrical load
(inflexible), 2000 W, the gas load (inflexible), 1000 W, and the heat References
load including both inflexible load, 2250 W, and flexible load with
the characteristics shown in Table 1. Considering that, the cutting [1] G. Koeppel, G. Andersson, Reliability modeling of multi-carrier energy
systems, Energy 34 (2009) 235–244.
off the inflexible load means the service is not available. Since with [2] L. Carradore, F. Bignucolo, Distributed multi-generation and application of the
the interruption of energy supply the performance of inflexible energy hub concept in future networks, Universities Power Engineering
loads is interrupted simultaneously, it is assumed that in the case Conference 2008. 43rd International, September 1–4 (2008) 1–5.
484 M.-H. Shariatkhah et al. / Energy and Buildings 126 (2016) 477–484

[3] S.P. Borg, N.J. Kelly, High resolution performance analysis of [17] T. Krause, F. Kienzle, S. Art, G. Andersson, Maximizing exergy efficiency in
micro-trigeneration in an energy-efficient residential building, Energy Build. multi-carrier energy systems, IEEE Power Energy Soc. Gen. Meeting 25-29
67 (2013) 153–165. (2010) 1–8.
[4] H. Cho, A.D. Smith, P. Mago, Combined cooling, heating and power: a review [18] M. Di Somma, B. Yan, N. Bianco, G. Graditi, P.B. Luh, L. Mongibello, V. Naso,
of performance improvement and optimization, Appl. Energy 136 (2014) Operation optimization of a distributed energy system considering energy
168–185. costs and exergy efficiency, Energy Convers. Manage. 103 (2015) 739–751.
[5] L. Mongibello, N. Bianco, C. Caliano, G. Graditi, Influence of heat dumping on [19] A. Shahmohammadi, M. Moradi-Dalvand, H. Ghasemi, M.S. Ghazizadeh,
the operation of residential micro-CHP systems, Appl. Energy 160 (2015) Optimal design of multi-carrier energy systems considering reliability
206–220. constraints, IEEE Trans. Power Delivery 30 (2) (2014) 878–886.
[6] M.L. Di Silvestre, G. Graditi, E.R. Sanseverino, A generalized framework for [20] P. Morales, A. Flores-Tlacuahuac, V.M. Zavala, Analyzing the effects of comfort
optimal sizing of distributed energy resources in micro-grids using an relaxation on energy demand flexibility of buildings: a multiobjective
indicator-based swarm approach, IEEE Trans. Ind. Inf. 15 (10) (2014) 2–162. optimization approach, Energy Build. 85 (2014) 416–426.
[7] G. Graditi, M.G. Ippolito, E. Telaretti, G. Zizzo, An innovative conversion device [21] J.C. Laurent, R.P. Malhame, A physically-based computer model of aggregate
to the grid interface of combined RES-based generators and electric storage electric water heating loads, IEEE Trans. Power Syst. 9 (3) (1994) 1209–1217.
systems, IEEE Trans. Ind. Electron. 62 (4) (2015) 2540–2550. [22] Q. Hu, F. Li, Hardware design of smart home energy management system with
[8] L. Mongibello, M. Capezzuto, G. Graditi, Technical and cost analyses of two dynamic price response, IEEE Trans. Smart Grid 4 (4) (2013) 1878–1888.
different heat storage systems for residential Micro-CHP plants, Appl. Therm. [23] G. Graditi, M.L. Di Silvestre, R. Gallea, E.R. Sanseverino, Heuristic-based
Eng. 71 (2) (2014) 636–642. shiftable loads optimal management in smart micro-grids, IEEE Trans. Ind. Inf.
[9] F. Brahman, M. Honarmand, S. Jadid, Optimal electrical and thermal energy 11 (1) (2015) 271–280.
management of a residential energy hub, integrating demand response and [24] M.H. Shariatkhah, M.R. Haghifam, M. Parsa-Moghaddam, P. Siano, Modeling
energy storage system, Energy Build. 90 (2015) 65–75. the reliability of multi-carrier energy systems considering dynamic behavior
[10] P. Mancarella, MES (multi-energy systems): an overview of concepts and of thermal loads, Energy Build. 103 (2015) 375–383.
evaluation models, Energy 65 (2014) 1–17. [25] R. Billinton, R.N. Allan, Reliability evaluation of engineering systems, 2nd
[11] J. Wasilewski, Integrated modeling of microgrid for steady-state analysis edition, Plenum Press, New York, London, 1992.
using modified concept of multi-carrier energy hub, Int. J. Elect. Power Energy [26] M.H. Shariatkhah, M.R. Haghifam, G. Chicco, M. Parsa-Moghaddam, Modelling
Syst. 73 (2015) 891–898. the operation strategies of storages and hydro resources in adequacy analysis
[12] M. Geidl, G. Andersson, Operational and structural optimization of of power systems in presence of wind farms, IET Renewable Power Gener.
multi-carrier energy systems, Euro. Trans. Electr. Power 16 (2006) 463–477. (2016), http://dx.doi.org/10.1049/iet-rpg.2015.0443.
[13] E. Fabrizio, M. Filippi, J. Virgone, An hourly modelling framework for the [27] R. Billinton, W. Wangdee, Predicting bulk electricity system reliability
assessment of energy sources exploitation and energy converters selection performance indices using sequential Monte Carlo simulation, IEEE Trans.
and sizing in buildings, Energy Build. 41 (2009) 1037–1050. Power Delivery 21 (2006) 909–917.
[14] A. Sheikhi, A.M. Ranjbar, H. Oraee, Financial analysis and optimal size and [28] R. Hemmati, R. Hooshmand, A. Khodabakhshian, Reliability constrained
operation for a multicarrier energy system, Energy Build. 48 (2012) 71–78. generation expansion planning with consideration of wind farms
[15] M. Geidl, G. Koeppel, Energy hubs for the future, IEEE Power Energy Mag. 5 (1) uncertainties in deregulated electricity market, Energy Convers. Manage. 76
(2007) 24–30. (2013) 517–526.
[16] M.C. Bozchalui, S. AhsanHashmi, H. Hassen, C.A. Cañizares, K. Bhattacharya, [29] J.C.O. Mello, M.V.F. Pereira, A.M. Leite da Silva, Evaluation of reliability worth
Optimal operation of residential energy hubs in smart grids, IEEE Trans. Smart in composite systems based on pseudo-sequential Monte Carlo simulation,
Grid 3 (4) (2012) 1755–1766. IEEE Trans. Power Syst. 9 (3) (1994) 1318–1326.