Energy and Buildings 71 (2014) 155–167

Contents lists available at ScienceDirect

Energy and Buildings

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

Managing energy Smart Homes according to energy prices: Analysis

of a Building Energy Management System
Rim Missaoui a,b , Hussein Joumaa a,∗ , Stephane Ploix a , Seddik Bacha b
a
G-SCOP Laboratory, 46, Avenue Felix Viallet, 38031 Grenoble Cedex 01, France
b
G2ELAB Laboratory, BP 46, 38402 Saint Martin d’Hres, France

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

Article history: The Demand-Side-Load Management will change the way people behave. Different authors have pro-
Received 1 July 2013 posed energy management algorithms for Smart Home that either integrates or not renewable energy. All
Received in revised form these researches have the same general objective: minimizing the daily energy cost without affecting the
26 November 2013
comfort of occupants. This paper deals with the performance analysis of a Global Model Based Anticipa-
Accepted 13 December 2013
tive Building Energy Management System (GMBA-BEMS) managing household energy. This GMBA-BEMS
is able to optimize a compromise between user comfort and energy cost taking into account occupant
Keywords:
expectations and physical constraints like energy price and power limitations. To validate the GMBA-
Building
Modeling BEMS, the model of a building has been developed in MATLAB/Simulink. This work analyzes GMBA-BEMS
Scheduling application that manages appliances such as heating, washing machine and dishwasher from a grid point
Satisfaction of view.
Energy price © 2013 Elsevier B.V. All rights reserved.
Load management
Optimization
Building Energy Management System
Smart Home
Smart grid

1. Introduction given criteria. In [5], an analysis of the load management techniques

Local energy production and consumption means in a Smart
Home can be managed by a Building Energy Management System
(BEMS). Advanced BEMS makes it possible to deploy new kinds of
energy management strategies that may change the way of con-
suming and producing energy by supporting occupants to reach a
better energy performance and comfort.
A Smart Home is a residential dwelling equipped with sensors
and possibly actuators to collect data and send control according to
occupants’ activities and expectations [1,2]. Potential applications
for Smart Homes are described in [3]. The goal of these applications
is to improve home comfort, convenience, security and entertain-
ment [3].
Thanks to this communication network, a load management
mechanism has been proposed in [4]. Since then, several studies
have been conducted in order to design an optimized electric BEMS
able to determine the best energy assignment plan according to a
∗ Corresponding author. Tel.: +33 622220526.
E-mail addresses: Rim.Missaoui@g2elab.grenoble-inp.fr (R. Missaoui),
Hussein.Joumaa@imag.fr (H. Joumaa), Stephane.Ploix@grenoble-inp.fr (S. Ploix),
Seddik.Bacha@g2elab.grenoble-inp.fr (S. Bacha).
is detailed. According to [6], Energy Management System contains
methods that coordinate the activities of energy consumers and
energy providers in order to best fit energy production capabili-
ties with consumer needs. With such solutions, electricity can be
reduced to support the grid.
During the last 2 years, many research projects focused on
demand side management and loads control of domestic Smart
Grid technologies for many reasons. First, energy use in buildings
currently account for about 32% of global total final energy con-
sumption in the world. In terms of primary energy consumption,
buildings represent around 40% in most IEA (International Energy
Agency) countries [7] and 65% of the total electric consumption
in [8]. Buildings are also responsible for 36% of the EU CO2 emis-
sions [7]. Not only Energy performance, but also load management
in buildings is a key issue to achieve the EU Climate & Energy
objectives, namely the reduction of a 20% of the Greenhouse gases
emissions by 2020 and a 20% energy savings by 2020 [9].
These technologies may modify the domestic energy use (elec-
tricity and heat) and adjust the electricity consumption/production
in dwellings [10,11]. These researches can be divided into two com-
plementary categories: predictive energy management and real
time control. This control uses prediction model in addition to
measured data in order to forecast the optimum control strategy

0378-7788/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.enbuild.2013.12.018
156 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167
that will be implemented. Similar researches have been carried out
on predictive controllers using stochastic models [12]. Both short-
term (10–20 min) and long-term (days) prediction errors lay within
acceptable ranges both in terms of temperature and humidity level.
The second category of research uses also the predictive con-
trol but it introduces real time control algorithms in order to give
more benefits contrary to [13,14] which do not study the price pre-
diction. Most of these researchers studied the real time electricity
pricing environments to encourage users to adjust load peaks for
two goals: reducing their electricity bill and reducing the Peak-to-
Average Ratio (PAR) in load demand [15,16].
The BEMS are usually based on simple models because it is diffi-
cult to determine the parameters of detailed models that fits actual
measurements. BEMS has to be “appropriate” to detailed models.
The problem of the evaluation of the degree of “appropriation” and
then the evaluation of the proposed solutions by the BEMS is rarely
treated as a research problem. This work deals with an analysis of
a Global Model Based Anticipative Building Energy Management
System (GMBA-BEMS) managing household energy.
Most anticipative approaches of energy management problem
focuses on specific appliances like electrical water heater in [16]
and HVAC (heating, ventilation, and air conditioning) system in
[15]. HVAC is the technology of indoor and vehicular environmen-
tal comfort. HVAC system design is a subdiscipline of mechanical
engineering, based on the principles of thermodynamics, fluid
mechanics, and heat transfer.
According to [17], the approach of GMBA-BEMS called G-
homeTech used in this paper is general enough to handle a large
set of electric appliances: electrical heater, washing machine,
dishwasher, fridge, etc. It represents 80% of the total residential
consumption [18].
The rest of this paper is organized as follows: The following sec-
tion states the objective of the work. The solution proposed for the
analysis of a BEMS is presented in Section 3. In Section 4, the settings
and the problem solved by the GMBA-BEMS are detailed. Analysis
of a GMBA-BEMS is presented in Section 5.
2. Paper objective
2.1. Problem statement
As detailed below, BEMS are based on simple models because
it is difficult to determine the parameters of detailed models that
fits actual measurements, BEMS has to be “appropriate” to details
models: it requires validation scenarios and a building simulator con-
nected to a BEMS.
Regarding the proposed testbench, the energy management
strategy aims to minimize the household’s electricity cost taking
into account price signals from the grid by optimally scheduling
the operation and energy consumption of each appliance accord-
ing to user comfort expectations. As in [15], a time varying curve of
electricity price is used. The household load management is based
on price and consumption forecasts considering users’ comfort
to meet an optimization objective which compromises minimum
payment and maximum comfort. Real time adjustments are then
done according to real time electricity market prices actual inter-
est. However, in this study, it is done according to the total available
power: the PV production (according to solar radiation curves) and
the power limitation (subscription), to the electricity market prices,
and the power consumption of the other appliances (Fig. 1). Notes
that PV is a method of generating electrical power by converting
solar radiation into direct current electricity using semiconductors
that exhibit the photovoltaic effect. Photovoltaic power generation
employs solar panels composed of a number of solar cells contain-
ing a photovoltaic material.
Sensors External data, weather and
prices
and distribution module in the
Dwelling power generation
- temperature
- light levels
- consumed energy
input
Predictive and
reactive optimizer
Actuators
output
- switches
- valves
- dampers… HMI
Simulator BEMS
Fig. 1. Virtual co-simulation general schema for BEMS validation.
The validation testbench is not only concerned with the heating
control such as in [19,20], but also with the electrical appliances
making the problem more complex. It aims to introduce a real time
energy management-decision based on both reactive and anticipa-
tive global algorithms [21] contrary to [22,23] where the authors
use predictive control to anticipate a solution for heating systems.
[17] and [24] detail the BEMS algorithm selected for the proposed
analysis. The anticipative layer assigns energy references by tak-
ing into account predicted events. Concerning the reactive layer, it
intervenes when the anticipative plan cannot be followed because
of unforecasted events and it decides whether some appliances
have to be switched ON or OFF.
On the other hand, to validate a BEMS, two parts must be pre-
sented: the simulator and the energy management algorithms (Fig. 2).
In the BEMS, the multi-layers algorithms are in interaction with
external data that comes from the weather, the energy marketer,
the Human Machine Interface (HMI), and the real time simulator.
The HMI can be used by the occupant to provide instructions to the
BEMS.
Simulators replace a dwelling and its HVAC systems to simu-
late their response to the BEMS as described in [20]. They are used
to improve product development, to train BEMS operators, to tune
actuators, and to simulate faulty situations [25]. Then, The vali-
dation of a BEMS should be done through a simulator model. The
simulation models includes in addition to the HVAC many electri-
cal appliances such as lighting, flaps, washing machine, dishwasher
and fridge.
Some simulators are presented in the literature. For example,
the software “PME Comfort” is used to simulate the thermal com-
fort of dwelling [26]. “Solene” [27], simulates the sunshine, light
and radiation. “ESP-r” [28] and “FLOVENT” [29] simulate the move-
ment of air in dwellings. In [30], there is a comparison between two
environments for dynamic simulations of dwellings “CODYRUN”
[31] and “TRNSYS” [32]. Because of the complexity of usage for
such advanced tools, new more accessible simulation tools have
been developed including “COMFIE/ PLEIADES” [33,34] and “Ener-
gyPlus” [35,36] offering the ability to model dwelling and HVAC
system.
SIMBAD [37] (SIMulator for Buildings And Devices) is one of
the recently developed simulator that may easily take into account
control systems in MATLAB/Simulink environment. It uses both the
“TRNSYS” and “HVACSIM+” [38] simulation software. Johnson Con-
trols and the National Institute of Standards and Technology (NIST)
in the US have developed a low cost PC-based emulator [39]. The
company is now using it for the purpose of testing new control
products.
 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167 157

Fig. 2. Exchanged command from BEMS and data from appliances.

The BEMS is fed up with simplified models compatible with a
mixed integer linear programming formulation [40]. But the behav-
ior of a real dwelling is much more complex. Therefore, simulation
requires finer dwelling models than the BEMS.
To summarize, the objective of the paper is to analyze a BEMS
in a context of variable pricing. To do this, three steps must be per-
formed: choose building simulator, configure the GMBA-BEMS and
analyze the results.
The MATLAB/Simulink simulator “SIMBAD” [41,42] (SIMulator
for Buildings and Devices) environment is chosen to be used in
this work because of its capability of modeling local control system
(SIMBAD is available at the address http://software.cstb.fr/).
3. Validation context
In this paper, the application of BEMS choosed is G-homeTech,
developed in GSCOP Lab and now marketed by Vesta-System com-
pany, whose energy management mechanism and methodology
are detailed in the paper [17]. The validation of this BEMS through
a finer simulation model of a building takes into account energy
prices and power limitation. Strengths and weaknesses of the
GMBA-BEMS are discussed later.
3.1. Validation system
In order to validate the optimization algorithms of the GMBA-
BEMS, a reference virtual dwelling model developed on SIMBAD
has been used. This dwelling named Gershwin has been modeled
on MATLAB/Simulink by the CSTB (Scientific and Technical Building
The case study considers only the energy management of the
ground floor of the Gershwin individual house. The ground floor of
the house was equipped with three heaters (one in each room: the
living room (3 kW), the kitchen (1.5 kW) and the room 4 (1.5 kW)).
Furthermore, the house is equipped with some appliances like a
washing machine (2 kW), dishwasher (2.2 kW), that are considered
as controllable loads (Fig. 3 and Table 1). Note that occupants may
be involved in the control loop. The rest of the power consumption,
like the lighting or the flap is considered as not controllable. They
are modeled by a daily consumption profile that described in the
last section.
Let now introduce the concept of service, a service with index i,
transforms energy in order to meet a user’s need via one or several
appliances. A service is qualified as permanent if its energy con-
sumption/production covers the whole time range of the energy
assignment plan such as heating service, otherwise, the service
is referred to as a temporary service such as cooking or washing
service.
A temporary service is characterized by the duration and desired
end time of the operation. The flexibility of this service comes from
the possibility of shifting its operating time it, i.e. bringing forward
or delaying the service.
A permanent service is characterized by a quantity of energy
consumed or produced. The flexibility of this service comes
from the possibility of modifying the energy quantities con-
sumed/produced throughout all the periods (decrease or increase
in energy consumption or production at a given time).
Table 1
Modeled appliances.

Centre) [43]. This building model is a classical recent individual low Name Power max (kW) Control
consumption house of 150 m2 , inhabited by a family of 4 persons. Heater living room 3.00 Set point temperature
Software engineering techniques such as co-simulation could be Heater room1 1.50 Set point temperature
implemented to validate a BEMS. The virtual co-simulation com- Heater kitchen 1.50 Set point temperature
bining SIMBAD and the GMBA-BEMS (G-homeTech) is one solution Washing machine 2.00 starting time
Dishwasher 2.20 starting time
for the heterogeneity issue.
158 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167
Fig. 3. The first floor of the virtual house.
The studied BEMS classifies these loads according to the pro-
vided service to the occupant. For example the heater provides
permanent services that depend on temperature set points. The
washing machine and dishwasher provide temporary services that
depend on ending times [17]. For each service, both simulation
model in Gershwin house and management model used by G-
homeTech are presented.
The house is supplied by the power grid and by a local
production that comes from solar panels. The grid power sup-
ply is modelled by a simple controllable voltage source and an
over-current switch (limited to the value imposed by the utility
subscription Pgrid = 6.5 kW). The PV system is 8 kW peak and its effi-
ciency is  = 10% and the capture area is S = 50 m2 . Hence the PV
model is written as follow:
PPV = .S.Psol (1)
where Psol is the solar power collected by 1 m2 of PV panel.
3.2. Building simulator
The principle of GMBA-BEMS validation is based on two types
of model: simulation model and energy management model. For
the simulation model, Gershwin house is used to test G-homeTech
behavior.
This model provides the weather at Trappes; a city near Paris
and the Matlab-Simulink model contains 4 blocs:
• The controllers block contains the controls of the equipment of
the house, heat pump, boiler, DHW Tank, and HVAC system. G-
homeTech manages the consumption of the heater by sending the
set point temperature and ON/OFF controls.
• The system includes models of the house equipment including
the thermal model which is a black box model. The electric appli-
ance models are introduced in this bloc and communicate with
G-homeTech by sending the power consumption and receive the
set point temperature and the ON/OFF controls and the startup
time for some appliances.
• The multi-zone building contains the description of the house
and calculates temperature, CO2 and lighting in each room with
an S-function performed by the CSTB.
• The sensor provides the temperature indicated by a sensor in each
room that is sent to G-homeTech.
As already described, the permanent service is related to the
heater. Concerning the thermal model for validation is a black box;
just the input and the output are known which are in communica-
tion with the manager (Fig. 4.).
Temporary appliances such as washing machines and dishwash-
ers are introduced in Gershwin house. Their simulation models are
based on active power measurements obtained from the real appli-
ances (Fig. 5).
When a user presses the ON/OFF button of washing machine
for example, it does not start operating immediately; it waits the
starting order from the BEMS.
4. Configuration of the BEMS
4.1. Configuration of permanent services
Thermal model used for simulation presents some differences
with the management model existing in G-homeTech [17]. Man-
agement thermal model does not take into account the temperature
of the adjacent rooms, the ground and wall temperature and the
heat flow of the occupants. The objective is to develop a simple
thermal model that represents the thermal behavior of the living
room of Gershwin house and compute the inside temperature from
the other data (temperature outdoor, the solar flux, etc.). The time
resolution for the GMBA BEMS is 1 h.
4.1.1. Choosing the right thermal model
Four models have been compared to make a better parameter
estimation of the thermal phenomena of the room.
 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167 159

Fig. 4. Washing machine and dishwasher active power consumption and device description.

• Model 1 is a first order model with three temperatures. This is
the simplest model because the inertia of the walls is globalized
(a unique capacity for all walls).
• Model 2 is a second order model which adds a second capaci-
tor Co in model 1 that represents the capacitance of surrounding
furniture.
• Model 3 is a first order model that adds a capacitance for each
wall.
• Model 4 corresponds to model 1 where three coefficients for
internal gains (lighting, heating and occupancy) have been added
to simulate losses for internal gains (lighting, heating and
occupation).

Fig. 5. Four compared electric analogies of a thermal model.

160 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167

Table 2
The parameter of the thermal model

Rmin Rmax RV Cm Rroom Rground Tground (◦ C)

0. 2501 608 7.067 1398600 0.380 3130 15

The model of the living room in the first floor is:

dT W Tin Tout Trooms Tground

= + + +
dt CW Rmin CW Rmout CW Rrooms CW Rground
 
TW 1 1 1 1
− + + + (2)
CW Rmin Rmout Rrooms Rground

Rmin Rv
 TW T out

Tin = light + heat + occ + + (3)
Rmin + Rv Rmin Rv

Same kinds of models are used for the other rooms. It has five
input variables.
Fig. 6. The equivalent circuit model of the living room thermal model. A parametric estimation has been performed to determine the
parameter of the chosen model. By using the model 1, the results
are presented in Table 2.
The idea is to match the temperature calculated by the ther-
mal model with the temperature given by the living room 4.1.2. Configuration of the thermal comfort used by GMBA-BEMS
of the Gershwin simulation. For this purpose, an parameter Permanent services manage a physical variable: the tem-
estimation process has been used to determine the optimal perature. Therefore, when occupants are present, the occupant
parameters of the electrical model to minimize the difference satisfaction can be assessed by the difference between the preferred
between the internal temperature computed by the model and value Topt and the actual temperature value. The proposed BEMS
the simulated one. In this case, it is to define RWin , Rwout , Rv , calculates the discomfort index D(i, k) for a particular permanent
Rground , Rroom , Cw and Tground . The resistance Rv models the ven- service i, corresponding to a thermal zone, at each period k using
tilation and the permeability of the building. The walls are the following equation:
characterized by a pair of resistance: the indoor wall resistance ⎧ T (i, k) − T (i, k)
(RWin ) and the outdoor wall resistance (Rwout ) and a capacitance ⎪ opt in
⎨ T (i, k) − T (i, k) if Tin (i, k) ≤ Topt (i, k)
opt min
CW . D(i, k) = (4)
⎪
⎩ Tin (i, k) − Topt (i, k) if Tin (i, k) > Topt (i, k)
Tmax (i, k) − Topt (i, k)
• light : Light power supplied to the room by lighting in [W]
• heat : Heating electrical power in [W]. where the Topt (i, k) is the preferred temperature at each anticipa-
• occ : Heat generated by the occupants in the room in [W]. In tive period and Tmin (i, k) and Tmax (i, k) stand respectively for the
this model, each occupant produces 80W of heat. So, occ =80 x minimum and maximum acceptable temperatures.
occupant’s number. In our case three thermal zones are configured: the first zone is
• Tout : outdoor temperature (◦ C) the living room (i = 1), the second zone is the kitchen (i = 2) and the
• Trooms : average temperature of the adjacent rooms (◦ C) third zone is the room4 (i = 3).
• TW : temperature of the wall (◦ C) ⎧
⎪ 21 − Tin (i, k)
• Tground : temperature of the ground (◦ C) ⎨ if Tin (i, k) ≤ 21
21 − 19
D(i, k) = (5)
⎪
⎩ Tin (i, k) − 21 if Tin (i, k) > 21
Note that the optimization may lead to a local mini- 24 − 21
mum. Thus, depending on the location of initial parameters
With i ∈ 1, 2, 3.
(Fig. 6).
The NELDER MEAD geometrical optimization algorithm
• The preferred anticipated temperature chosen by the occupant
has been used because it can handle nonlinear optimiza-
tion problem. A large number of optimizations have been is fixed at Topt (i) = 21 ◦ C. Tmin (i) = 19 ◦ C and Tmax (i) = 24 ◦ C for the
performed for each model (1000 optimizations) span large three rooms. The maximum power consumed by the first heater
number of random possible initializations to avoid local min- is 3 kW; the second and the third heater consume both 1.5 kW.
ima. The results for all parameter optimizations are given in
Fig. 7. 4.2. Configuration of temporary services
Based on this analysis, model 1 appears to be more advanta-
geous because it is the simplest even though it leads to similar • The fourth service (i = 4) corresponds to a washing machine. It is
optimization error as others but with better convergence proper- considered as a temporary service. It can be shifted providing that
ties. It is implemented and integrated into the G-homeTech energy the following comfort are satisfied: fmin (4) = 2 am, fmax (4) = 3 pm
manager. The number of parameters is less important than in other and fopt (4) = 10 am. The maximum power consumed is 2 kW.
model: it facilitates the optimization and increases the chances of • The fifth temporary service (i = 5) corresponds to dishwasher.
finding a global minimum. Based on these results, model 1 leads to Its fmin (5) = 2 pm, fmax (5) = 11 pm and fopt (5) = 10 pm. The maxi-
a global minimum error (Fig. 7) for a large number of initial random mum power consumed by this service is 2.2 kW. The potential
parameters. consumption duration d(i, k) for the two services is 2 h.
 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167 161

Fig. 7. Results for parameter optimizations.

According to [17], G-homeTech models temporary service is as where fopt (i) stands for the requested ending time and fmin (i) and
follows: fmax (i) stand respectively for the minimum and maximum accept-
able ending times.
d(i, k) = min(f (i), (k + 1)) − max(f (i) − d(i), k) (6) Two temporary services are configured: the washing machine
(i = 3) and dishwasher (i = 4).
where d(i, k) is the potential consumption duration of the loads ⎧
corresponding to service i during a sampling period [k, (k + 1)]. ⎪ f (3) − 10
⎨ if f (3) > 10
 is the sampling period of the anticipative layer. f(i) the ending 15 − 10
D(3) = (9)
time is the decision variable associated to the temporary services. ⎪
⎩ 10 − f (3) if f (3) ≤ 10
Therefore, the energy consumed by the loads E(i, k) related to 10 − 2
service i during d(i, k) is given by (7): ⎧
⎪ f (4) − 22

⎨ if f (4) > 22
E(i, k) = d(i, k)P(i, k) if d(i, k) > 0 23 − 22
(7) D(4) = (10)
E(i, k) = 0 otherwise ⎪
⎩ 22 − f (4) if f (4) ≤ 22
22 − 14
where P(i) is the power consumed sampling period of the service i.
The BEMS manages the energy Smart Home consumption according
The GMBA-BEMS uses a model with a single operating cycle
to energy price and power limitations with respect to weighted
(average power) instead of three that is used by the validation
sub-criteria: the comfort and the cost.
model (Fig. 5).
The objective of the BEMS is to answer to the occupants’ require-
ments, i.e. reaching the best compromise between thermal comfort, 4.3. Configuration of cost criterion used by GMBA-BEMS
ending time of temporary services and energy cost.
When the available energy is sufficient for all the temporary In the dwelling integrated German photovoltaic price model,
services requesting energy, then the ending time f(i) for each tem- the total energy cost depends on three different prices of the kWh:
porary service i is equal to the requested ending time fopt (i). In the self-consumption cost Cself when the energy produced locally is
contrast, when the available energy is not sufficient for all the ser- consumed, the bought cost Cbought when the energy from the grid is
vices and/or the energy price is very expensive at that time, some of bought and the sold cost Csold when the energy is selled to the grid.
the loads should be shifted earlier or delayed. Therefore the actual The energy cost criterion Ctota of the grid connected PV system is
ending time f(i) differs from fopt (i). This phenomenon decreases the written (Table 3):
occupant satisfaction but may improve the economic criterion. The where EPV stands for the energy produced locally and Econs for
dissatisfaction criterion for the temporary service i is expressed as the total consumed energy.
follows:
⎧ f (i) − f (i) Table 3
⎪ opt
⎨ f (i) − f (i) if f (i) > fopt (i) German photovoltaic prices model
max opt
D(i) = (8)
⎪
⎩ fopt (i) − f (i)
CselfEuro/kWh CboughtEuro/kWh CsoldEuro/kWh
if f (i) ≤ fopt (i)
fopt (i) − fmin (i) 0. 2501 Is variable 0. 4301
162 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167

criterion results from the aggregation of two types of criteria: the

economic criterion, determined by the total cost of energy, and the
user discomfort criterion function (the deviation from the user’s
optimal thermal preferences and the shifting of the ending times
of temporary services):


T
ˇ  
T
J= C(j, k)E(j, k) +  ˛(i) D(i, k) (12)
i∈
˛(i)
j∈S k=1 i∈ k=1

The parameter ˛(i) depicts the priorities in between the con-

sumer services and the parameter ˇ depicts the relative importance
given by the occupants to the energy cost and the comfort.
Two cases have been studied in this paper. The first one is
related to when the energy cost is considered as a constraint in the
optimization. The occupants in this case accept to decrease their
comfort level in order to reduce the energy bill. In the second case,
Fig. 8. Considered predictions: (a) uncontrollable power consumption profile, (b)
only the comfort criterion is considered. The BEMS should select
the purchase energy price, (c) PV production and (d) the real and predicted Tout .
the best solutions based on the consumer preferences.

The energy used is modeled by E(5, k). The maximum subscribed

power is Egrid (5) = 6.5 kW. The consumption/production balance
leads to:
5. Analysis of the impact of a BEMS

4
Econs (i, k) ≤ E(5, k) + EPV (6, k) (11)
Let now remind the case study and the objective. The first floor
i=1 of a virtual house called Gershwin has been taken as an example to
with Econs , the total load energy consumption, E(5), the power analyse a GMBA-BEMS. The map of this floor given in Fig. 3 contains
drawn from grid and EPV (6), the power locally produced. a living room, a kitchen, and a room 4 that is equipped with some
controllable and uncontrollable electrical appliances (Table 1). Four
4.4. Objective function used by GMBA-BEMS persons occupy this floor and depending on the room they have
different occupation profiles during the day (Fig. 11). The objective
The BEMS performs optimizations where objective is a com- is to analyse the strategy of the BEMS and compare it with the choice
promise weighted energy cost and occupant comfort. The higher of the occupant and the most important is to check if there will be
cost should correspond to the peak consumption periods. The final money earned in a German energy price context.

Living Room Kitchen Room4

22 22
Temperature (°C)Temperature (°C) Temperature (°C)

22
Occupant

21.5 21.5 21.5

solution

21 21 21

20.5 20.5 20.5

22 22
Comfort-prefer-
red solution

21 21 21

20 20 20

19 19 19

22 22 24
23
21 21
Cost-prefer-
red solution

22
20 20 21
20
19 19 19
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
Set point temperature sent by the BEMS Room temperature
Heater Power

4 2 2
3 1.5 1.5
(kW)

2 1 1
1 0.5 0.5
0 0 0
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
The heater power consumption in comfort – preferred solution The heater power consumption in cost – preferred solution

Fig. 9. Thermal behavior of three thermal zone without reactive control.

 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167 163

Fig. 10. Occupation and lighting power of the three thermal zones.

5.1. Scenarios of the case study
In order to show the impacts of the proposed algorithm, two
main scenarios have been considered: analysis without and with
BEMS.
In the occupant solutions (without BEMS), the occupant man-
ages his own appliance without receiving the information from
the BEMS. The only information available for him is the weather
forecast and the subscription for the next day. The occupant is
assumed to seek to satisfy its need for comfort without considering
the cost. So, he selects the desired temperature for the three rooms
Topt (1) = Topt (2) = Topt (3) = 21 ◦ C (occupant solutions in Fig. 9, where
only living room is presented) during the day. For the same day, he
starts the washing machine at 8 am, the dishwasher at 8 pm. For the
other appliances are considered uncontrollable equipment. Just a
daily consumption profile is considered (see Fig. 8a).
conversely, the BEMS controls the appliances shown in Fig. 1
according to occupant requirements, the weather forecast data, the
consumption and the electricity price. GMBA-BEM is supposed to
shift electricity demand to more beneficial periods (e.g., nights or
when electricity is cheap).
The same demands as the occupant solutions scenarios have been
considered for comparison. Furthermore, weather data (outside
temperature, solar radiation), energy prices and subscription are
known 24 h in advance.
In this scenario, two cases are considered: in the first one
(comfort- preferred solution) only the comfort criterion is consid-
ered and in the second case (cost-preferred solution) only the cost
criterion is used. For the first case, the BEMS is supposed to follow
the temperature set point of the permanent services (3 heaters)
and to start temporary services as close as possible to expected
ending time. With this solution, the occupants should pay more
than where comfort is preferred. To minimize the cost, occupants
have to reduce their comfort.
Section 5.2 focuses on anticipation mechanism and Section 5.3
considers both reactive and anticipative mechanisms.
5.2. Analysis of the anticipative mechanism

3500 The weather forecast, the electricity price and anticipate uncon-
Anticipated consumed power Real consumed power
trollable consumption are given in Fig. 8.
3000 The results given in Fig. 9-occupant solution clearly show the
influence of lighting power and the number of occupants on the
Consumed Power (W)

2500 indoor temperature Tin of the three rooms (see Fig. 10-occupation
and -lighting power). Indeed, in the peak period of lighting power
2000
and when the room is occupied, the inside temperature does not
follow the temperature set point set by the occupant or computed
by the BEMS.
1500
In the solution based on the BEMS the set point tempera-
ture is given by the BEMS in the comfort-preferred solution and
1000
cost-preferred solution. The difference between the two cases
(Fig. 16-comfort- preferred solution and cost-preferred solution) can
500 be noticed by the temperature of the rooms at 8 pm when the price
is very high.
0 Fig. 8 shows that there are two price peaks during the day and
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours) the highest price of energy is at 8 pm with 0.26D/kWh. These pro-
files are used in order to control energy Smart Home through a
Fig. 11. Real and anticipated consumed power of the heater. variable purchasing price which is the second scenario.
164 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167

3000 Preferred occupant 3000

Optimal solution
solution (comfort)
Optimal solution regarding cost
regarding cost
Power (W)

Power (W)
2000 electricity price
electricity price
2000 Preferred occupant
solution (comfort)

1000
1000

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hour) Time (hour)

Fig. 12. Temporary services behavior: dishwasher. Fig. 13. Temporary services behavior: washing machine.

1,4 Anticipated global cost (cost – preferred solution)=1,86€

Indeed, with the comfort-preferred solution the BEMS preserves
1,2 Anticipated global cost (comfort – preferred solution)=2,7€
the temperature set point equal to requested one despite the high
energy prices at this time. In this case, the temperature set point 1
sent by the BEMS is equal to T(op) = 21 ◦ C when the room is occu- 0,8
pied and equal to the Tmin = 19 ◦ C when there is no one. But for the

Global Cost €
0,6
cost-preferred solution, the BEMS at 8 pm, reduces the set point tem-
0,4
perature of the three rooms during 1 h to Tmin = 19 ◦ C chosen by the
0,2
resident for reducing the consumed energy by the heaters.
To better analyze the impact of G-homeTech, especially the ther- 0
mal model that has been used, expected and actual data must be -0,2
compared. For actual example the cost-preferred solution, the antic- -0,4
ipated and the power consumption of the living room heater (the
-0,6
first service) are compared in Fig. 11.
1 3 5 7 9 11 13 15 17 19 21 23
In Fig. 11, slight differences between expected and actual results
Time (hours)
can be observed. These differences are not significant. It is related
to the 1 h step computation of the anticipated consumed power. Fig. 14. The global energy cost in the three scenarios.
Figs. 12 and 13 show the difference between the occupant solu-
tions (without BEMS), comfort-preferred solution and cost-preferred
solution, concerning temporary services. The occupant solutions are to the requested one fopt (4) = 10 am and fopt (5) = 10 pm. However,
the same as the comfort-preferred solution (because the BEMS trans- in the cost-preferred solution, the BEMS shifts the two appliances
mits the set point preferred by the consumer). into off-peak hours reducing in this case the comfort of the user
The BEMS set the washing machine and the dishwasher respec- and increasing the energy cost (Fig. 14).
tively at 8 am and 8 pm in the first and second case to promote the Fig. 15 presents the energy price of the comfort-preferred solution
comfort criterion despite the energy price is very high at these two and cost-preferred solution. If the comfort of the consumer is pre-
times of the day. So, it keeps the ending time of the two loads equal ferred, the global energy cost during 24 h is 2.7D but if the inhabitant

Fig. 15. Results with reactive control.

 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167 165

chooses to reduce his electricity bill, then he saves about 1D/day and
he just pays 1.86D.
In this section, the power constraint described in (8) is satis-
fied. The total consumed power does not exceed the anticipated
available power that is the sum of the subscription and the local
production of the PV. The real parameters as the PV production
power, no-controllable consumption and the outside temperature
are kept the same as the anticipated one. In this case, the set
point temperature calculated by the anticipative layer was trans-
mitted without modification of thermal simulator model. The total
anticipated consumed power by the heater of the heater of the
living room for example (24.7 kWh/24 h) is higher than the real
one ((22.6 kWh/24 h). This difference is related to the 1 h step
consumption of the anticipated consumed power. The manager
reserves energy on hour step time which explains the 8.5% differ-
ence between the two consumptions. Concerning the temporary
service (washing machine and dishwasher) always the solution
Fig. 16. (a) The instantaneous total power consumption with and without BEMS,
given by the anticipative layer is transmitted to the simulator what-
(b) the average total power consumption with and without BEMS.
soever in the comfort-preferred solution cost-preferred solution.

5.3. co-simulation validation process with reactive control

The impact of the reactive mechanism has not been tested yet. In
the previous paragraph, only the anticipative mechanism has been
tested.
Till now, the obtained simulation results for real conditions are
equal to the predicted one. Actual values usually differ from predic-
tions. In the following, the role of the reactive layer will be tested
by changing the actual solar radiation (the PV power produced
locally) and the uncontrollable power consumption profiles (see
Fig. 15). Because of these changes, the constraint described by Eq.
(8) is not satisfied during same time periods and the consumption
exceeds the available power in during same periods. The PV power
production is reduced between 10 am and 12 am. Furthermore, the
non-controllable real power is different from the predicted one and
it exceeds the subscription Pgrid = 6.5 kW at 8 pm (see Fig. 15).
Fig. 15 summarizes the results of this last scenario. When the
power production is reduced, the total consumption exceeds the
available power and the OFF control is sent by the BEMS to the
heater of room 4 shows the cut off of the heater twice a day:
between 11 am and 12 am when the PV production is not sufficient
and the total consumption exceeds the available power. The BEMS
chooses to turn off only the heater of the room 4 because it is not
equipped in this period and it is also a good solution to return to the
energy balance. The second controls disable are sent between 8 pm
and 9 pm when the consumption exceeds 6.5 kW. The BEMS control
for the first time slot shows that there is a swinging phase when the
BEMS changes quickly the control and oscillates between enable
and disable. Indeed, every 1 mn (the sampling time), the BEMS reac-
tive control compares the amount of consumed power which is
available. So when the heater is disabled, the consumed power is
less than the subscription so it sends the enable. For this period,
the BEMS computes again the sum of needed powers exceeds the
limits of the subscription so it sends disable and so on. To solve
this oscillation problem, reactive mechanism should consider time

Total consumptionprofile withoutEMS Anticipatedpurchaseprice (a)

12
Total consumptionprofile withEMS Subscription
10
Power (W)

0
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (Hours)
10
Avreage total power consumption without EMS (b)
8 Avreage total power consumption without EMS

6
Power (KW)

0
0 5 10 15 20 25
Time (Hour)

Fig. 17. Global energy cost with and without BEMS.

166 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167
horizon and react to problematic situation by taking into account
past decisions.
For the second violation of the anticipative plan, when the
non-controllable power exceeds the subscription, the BEMS sends
disable controls for all the three heaters (Fig. 16) at 8 pm. At this
time the indoor temperature don’t follow the requested tempera-
ture given by the BEMS.
Reactive mechanism operates to satisfy the power constraint
described in (8) while keeping a high satisfaction level. The total
anticipated consumed power by the heater of the living room for
example (24,7 kWh/24 h) is higher than the real one with reactive
control (20,3 kWh/24 h). The appearance of the reactive control
increases the difference between the two consumptions (17.8%).
Concerning the temporary service the same results as before is
found because this service in not interruptible. Fig. 16 shows the
instantaneous and the average total power consumption with and
without BEMS.
A last simulation has been done with the three scenarios
seen before: occupant solution, comfort-preferred solution and cost-
preferred solution in the context of reactive control. Fig. 17 presents
the energy price in these three cases. If the comfort of the con-
sumer is preferred, the global energy cost during 24 h is 6.43D but
if the inhabitant chooses to reduce his electricity bill, then he saves
about 1D/day and he pays just 5.9D. The highest cost paid is when
the occupant chooses to manage alone his appliances consumption
that is 7.3D. The reactive control allows not only to save money,
but also not to exceed the available power and maybe avoid over-
consumption penalty that currently exists in the industry and not
in the residential sector.
6. Conclusions
The BEMS are usually based on simple models because it is diffi-
cult to determine the parameters of detailed models that fits actual
measurements. BEMS has to be “appropriate” to detailed models.
The problem of the evaluation of the degree of “appropriation” and
then the evaluation of the proposed solutions by the BEMS is rarely
treated as a research problem.
In this paper an analysis of an advanced Global Model Based
Anticipative Building Energy Management System (GMBA-BEMS)
from an electrical point of view has been performed.
A case study based on varying electricity price demonstrates
the benefits of the proposed algorithm to reduce the electricity bill
approximately 20.
Simulation results show that the combination of the proposed
energy scheduler design and the price predictor leads to signifi-
cant cost reduction about 1 per day (365 per year). This encourages
the occupants to participate in a residential electrical load control
program.
Acknowledgments
This work has been performed in the scope of the ReactivHome
project accredited by the National Agency for Research (ANR) and
benefits also of the HOMES project. The authors want to acknowl-
edge all the project partners especially the industrial partners:
Schneider Electric, CSTB and Vesta-system as well as the LIG Lab-
oratory and Ophelie Painchault for her contribution to parameter
estimation part.
References
[1] L. Jiang, H. Liu, B. Yang, Smart home research, in: Proceedings of International
Conference on Machine Learning and Cybernetics, vol. 2, August, Shanghai,
China, 2004, pp. 659–663.
[2] V. Ricquebourg, D. Menga, D. Durand, B. Marhic, L. Delahoche, C. Loge, The smart
home concept: our immediate future, in: 1st IEEE International Conference on
E-Learning in Industrial Electronics, December, Hammamet, Tunis, 2006, pp.
1–5.
[3] S. Meyer, A. Rakotonirainy, A survey of research on context-aware homes, in:
Proceedings of the Australasian Information Security Workshop Conference on
ACSW Frontiers 2003, vol. 21, Darlinghurst, Australia, 2003, pp. 159–168.
[4] K. Wacks, The impact of home automation on power electronics, in:
Proceedings of Applied Power Electronics Conference and Exposition, March,
1993, pp. 3–9.
[5] Z. Paracha, P. Doulai, Load management: techniques and methods in electric
power system, in: Proceedings of IEEE Energy Management and Power Delivery,
vol. 1, March, 1998, pp. 213–217.
[6] K. Wacks, Utility load management using home automation, IEEE Transactions
on Consumer Electronics 37 (1991) 168–174.
[7] OECD/IEA, International Energy Agency, 2013, http://www.iea.
org/aboutus/faqs/energyefficiency/.
[8] Adem. Centre de ressource, http://www.pcet-ademe.fr/domaines-actions/
batiments/contexte-et-enjeux
[9] RTE: réseau de transport d’électricite, Le bilan électrique francais 2010, 2011
http://www.rte-france.com/
[10] J. Scott, P. Vaessen, F. Verheij, Reflections on smart grids for the future, in: Dutch
Ministry of Economic Affairs, April, 2008.
[11] Distributed generation in liberalised electricity markets, 2002, http://gasunie.
eldoc.ub.rug.nl/FILES/root/2002/3125958/3125958.pdf.
[12] R. Negenborn, M. Houwing, B. De Schutter, H. Hellendoorn, Adaptive prediction
model accuracy in the control of residential energy resources, in: Proceedings
of IEEE International Conference on Control Applications, September, San Anto-
nio, USA, 2008, pp. 311–316.
[13] F. Wolak, Residential customer response to real-time pricing: the Anaheim
critical-peak pricing experiment, in: Center for the Study of Energy Markets,
May, 2006, Working Paper 151.
[14] A.H. Mohsenian-Rad, V. Wong, J. Jatskevich, R. Schober, Optimal and
autonomous incentive-based energy consumption scheduling algorithm for
smart grid, in: Innovative Smart Grid Technologies (ISGT), January, Gothenburg,
Sweden, 2010, pp. 1–6.
[15] A. Mohsenian-Rad, A. Leon-Garcia, Optimal residential load control with price
prediction in real time electricity pricing environments, IEEE transactions on
Smart Grid 1 (2010) 120–133.
[16] D. Pengwei, L. Ning, Appliance commitment for household load scheduling,
IEEE transactions on Smart Grid 2 (2011) 411–419.
[17] D. Long Ha, H. Joumaa, S. Ploix, M. Jacomino, An optimal approach for electrical
management problem in dwellings, Energy and Buildings 45 (2012) 1–14.
[18] RTE: réseau de transport d’électricite, Bilan prévisionnel de l’équilibre offre-
demande d’électricite en france, 2011 http://www.rte-france.com/
[19] B. Paris, J. Eynard, S. Grieu, T. Talbert, M. Polit, Heating control schemes
for energy management in buildings, Energy and Buildings 42 (2010) 1908–
1917.
[20] J. Clarke, J. Cockroft, S. Conner, J. Hand, N. Kelly, R. Moore, T. Brien, P. Strachan,
Simulation-assisted control in building energy management systems, Energy
and Buildings 34 (2002) 933–940.
[21] D.L. Ha, S. Ploix, E. Zamai, M. Jacomino, Tabu search for the optimization of
household energy consumption, in: IEEE International Conference on Infor-
mation Reuse and Integration, Waikoloa, Hawaii, USA, September 2006, pp.
86–92.
[22] F. Oldewurtel, A. Parisio, C. Jones, D. Gyalistras, M. Gwerder, V. Stauch, B.
Lehmann, M. Morari, Use of model predictive control and weather forecasts
for energy efficient building climate control, Energy and Buildings 45 (2012)
15–27.
[23] R. Balan, J. Cooper, K. Chao, S. Stan, R. Donca, Parameter identification and model
based predictive control of temperature inside a house, Energy and Buildings
43 (2011) 748–758.
[24] D. Long Ha, S. Ploix, M. Jacomino, M. Hoang Le, Home energy management prob-
lem: towards an optimal and robust solution Energy Management, INTECH,
2012, pp. 77–105 (Chapter 5).
[25] G.E. Kelly, W.B. May, J.Y. Kao, C. Park, Using emulators to evaluate the per-
formance of building energy management systems, ASHRAE Transactions:
Symposium 100 (1994) 1482–1493.
[26] P.O. Fanger, A. Stberg, A.G.McK. Nicholl, N. Breum, E. Jerking, Thermal comfort
conditions during day and night, European Journal of Applied Physiology and
Occupational Physiology 33 (1974) 255–263.
[27] P. Ineichen, Quatre années de mesures d’ensoleillement à Genève: 1978–1982,
Mesures d’ensoleillement à Genève, Université de Genève, Groupe de physique
appliquée, Section de physique, 1983.
[28] D. Thevenard, K. Haddad, Ground reflectivity in the context of building energy
simulation, Energy and Buildings 38 (2006) 972–980.
[29] Refrigerating American Society of Heating and Air-Conditioning Engi-
neers, Ashrae standard: Ventilation for acceptable indoor air quality, 1989
http://books.google.fr/books?id=ZsoLAQAAMAAJ
[30] F. Lucas, T.A. Mara, F. Garde, H. Boyer, A comparison between CODYRUN and
TRNSYS, simulation models for thermal buildings behaviour, in: Word Renew-
able Energy Congress, Florence, Italy, 1998.
[31] A. Lauret, T. Mara, H. Boyer, L. Adelard, F. Garde, A validation methodology aid
for improving a thermal building model: case of diffuse radiation accounting
in a tropical climate, Energy and Buildings 33 (2001) 711–718.
[32] F. Kuznik, J. Virgone, K. Johannes, Development and validation of a new
{TRNSYS} type for the simulation of external building walls containing {PCM},
Energy and Buildings 42 (2010) 1004–1009.
 R. Missaoui et al. / Energy and Buildings 71 (2014) 155–167
[33] E. Kruger, B. Givoni, Predicting thermal performance in occupied dwellings,
Energy and Buildings 36 (2004) 301–307.
[34] S. Thiers, B. Aoun, B. Peuportier, Experimental characterization, modeling and
simulation of a wood pellet micro-combined heat and power unit used as a heat
source for a residential building, Energy and Buildings 42 (2010) 896–903.
[35] D. Crawley, L. Lawrie, F. Winkelmann, W.F. Buhl, Y. Huang, C. Pedersen, R.
Strand, R. Liesen, D. Fisher, M. Witte, J. Glazer, EnergyPlus: creating a new-
generation building energy simulation program, Energy and Buildings 33
(2001) 319–331.
[36] P. Ellis, P. Torcellini, Simulating tall buildings using energyplus, in: IBPSA Inter-
national Conference, Montreal, Canada, 2005.
[37] F. Chlela, A. Husaunndee, C. Inard, P. Riederer, A new methodology for the design
of low energy buildings, Energy and Buildings 41 (2009) 982–990.
167
[38] P. Cui, H. Yang, J. Spitler, Z. Fang, Simulation of hybrid ground-coupled heat
pump with domestic hot water heating systems using hvacsim+, Energy and
Buildings 40 (2008) 1731–1736.
[39] G. Decious, C. Park, G. Kelly, A low cost building/hvac emulator, in: HPAC
Heating/Piping/Air-Conditioning, vol. 2, January 1997, pp. 188–193.
[40] C. Clastres, T. Ha Pham, F. Wurtz, S. Bacha, Ancillary services and optimal house-
hold energy management with photovoltaic production, Energy 35 (2010)
55–64.
[41] F. Chlela, A. Husaunndee, C. Inard, P. Riederer, A new methodology for the design
of low energy buildings, Energy and Buildings 41 (2009) 982–990.
[42] P. Riederer, From sizing and hydraulic balancing to control using the simbad
toolbox, in: IBPSA International Conference, Eindhoven, The Netherlands, 2003.
[43] http://www.cstb.fr/http://www.cstb.fr/