WK,QWHUQDWLRQDO&RQIHUHQFHRQ5HQHZDEOH(QHUJ\5HVHDUFKDQG$SSOLFDWLRQV 3DULV)5$1&(2FW

Energy Consumption Optimization through Dynamic

Simulations for an Intelligent Energy Management of
a BIPV Building
Ilias Papas, Bruno Estibals, Christelle Ecrepont, Corinne Alonso
Integration of Energy Management Systems
LAAS-CNRS
Toulouse, France
[email protected]

Abstract - Having 6500 integrated sensors and a wide network energy. The development of buildings that aim to minimize
of embedded systems, the ADREAM building can be categorized consumption while maximizing production of energy, such as
as one of the prototype smart buildings of France. As needs for the ADREAM building, promoted the spread of Building
intelligent management of energy are growing, the associated Integrated Photovoltaic (BIPV) systems.
project provides a multidisciplinary platform of experimentation,
developing solutions for efficient Energy Networks, HVAC The constantly increasing focus of the scientific community
Systems, Photovoltaics, and Smart Grids. This paper provides on the optimization of BIPV Systems can be demonstrated in
the overview of the project along with three different modeling [21], [22], [23], [24], while the interest for Building Energy
techniques, illustrating their strengths and limitations on Management Optimization can be seen in [25].
simulating the thermal behavior of the building and the
functioning of the different energy systems. A thermal model of Therefore, the ADREAM project, combining ideas from
the building was developed and calibrated for energy optimized designs of other sustainable buildings, such as the
consumption analysis and prediction, using the software Solaris project in the Paris region [8], and the Edge in
Pleiades+Comfie. A “black box” model was developed using Amsterdam [9], aims to stand as one of the world’s prototype
artificial neural networks for the simulation of energy system models for intelligent energy management. This paper presents
parameters and the exploration of efficient control strategies. the scientific context of the ADREAM building in Section II.
The last method provides the overview of developing an all- The description and analysis of the various modeling strategies
inclusive electrically equivalent physical model with that aim to predict and optimize thermal and electrical
Matlab/Simulink for simulating the global functioning of the consumption is given in Sections III, IV, and V. Section VI
building and its HVAC systems. The conclusion addresses the provides the evaluation of the associated results and concludes
utility of exploring and combining different types of models for on the ongoing research, while it outlines the future work in the
optimizing the energy management of Smart Buildings. domains of thermal and electrical energy consumption
optimization.
Keywords - Smart Building, Building Simulation, Energy
Management, Energy Optimization, Pleiades + Comfie, Model
Calibration, Simulink II. SCIENTIFIC CONTEXT OF THE ADREAM BUILDING
The ADREAM project (French acronym for Embedded
I. INTRODUCTION Reconfigurable Dynamic Autonomous and Mobile
Architectures) is a research program focused on the
Energy management optimization is a growing issue in our
optimization of Smart Grids, Micro-Grids, BIPV systems,
society. As buildings account for about 40% of the global
HVAC (Heating, Ventilation, Air Conditioning) systems, as
energy consumption, the EU proposes a target of 27% more
well as on Ambient Physical Cyber Systems [10]. Figure 1
energy savings by 2030 [1]. The global initiative for energy
shows an aerial view of the building along with its large PV
optimization gave rise to the concept of Zero Energy Buildings
surface. Figure 2 illustrates the various Energy Systems
(ZEB), which was first defined in scientific literature in 2006,
installed in the building.
but it had not been translated concretely into laws and norms.
After the National Renewable Energy Laboratory (NREL) This paper focuses on ADREAM’s objective of developing
proposed an initial definition, Aalborg University produced a a modeling approach to energy management and optimization
state of the art synthesis of ZEB definitions [2] [3]. Thus, the for the totality of the installed systems, integrating all the
research focused on demonstrations over one-year-cycles, entities of production and consumption, by deploying the
which eventually highlighted the fact that energy demand for extensive network of intelligent sensors and regulators. The
heating and electrical consumption needs to be minimized. At presented work provides an overview of three different
the same time, the energy supply should rely completely on a modeling and simulation methods whose aim is to explore
building’s annual production of renewable energy [4] [5]. The applicable solutions for the energy consumption optimization
most common source exploited for energy production is solar objective of ADREAM through an intelligent management of

,&5(5$ 
‹,(((
WK,QWHUQDWLRQDO&RQIHUHQFHRQ5HQHZDEOH(QHUJ\5HVHDUFKDQG$SSOLFDWLRQV 3DULV)5$1&(2FW

Fig. 2. Energy Systems installed in the ADREAM Building

III. DYNAMIC THERMAL SIMULATION

The initial modeling phase of the building ADREAM was
carried out by the software Pleiades + Comfie (P+C) [12]. P+C
is a widely used thermal simulation program developed at the
Ecole des Mines in Paris and supplied by the company Izuba.
As indicated by Peuportier [13], P+C can be considered as a
precise simulation tool for the generation of temperature
evolution and energy needs profiles, as well as for the study of
their sensibility to main thermal parameters.
A Dynamic Thermal Simulation (DTS) serves as a
computer tool that allows the thermal study and electrical
energy assessment of a building from multiple perspectives
(temperature profiles, energy consumption, comfort). The
Fig. 1. Aerial view of the ADREAM Building interest for this type of modeling stems from the motivation to
simulate with high precision the particular architecture of
energy. Thus, for this purpose, the modeling process ADREAM’s structure, along with its thermal properties and the
incorporates and explores the interaction of the thermal and technical characteristics of the systems installed for the purpose
electrical elements in the building. Similar work focusing on of optimizing its thermal and electrical energy consumption.
the modeling process and optimization of HVAC systems and The complete procedure of this modeling algorithm and
energy consumption can be found in [26], [27], [28], [29], and method is described in Papas et al. [14]. Figure 3 shows the 3D
[30]. Model of the ADREAM Building in the software interface.
This study concerns the transfer of energy within the
ADREAM building considering the heating, the air
conditioning, the lighting, and the various electronic devices.
ADREAM is connected to an electrical grid which benefits
from a total of 100kWc of photovoltaic energy. In addition, it
involves the use of geothermal energy through the installation
of three ground source heat pumps (135kW). The hygienic
ventilation system is also linked to a ground heat exchanger to
minimize the heating/cooling energy needs of the air. The
notion of a ZEB is examined extensively through the
ADREAM research program as the electrical energy surplus
from the photovoltaic production, combined with the optimized
consumption of high performance HVAC (Heating, Fig. 3. 3D Model of the ADREAM Building
Ventilation, and Air Conditioning) systems, can be transported
to the rest of the buildings composing the LAAS-CNRS The P+C model was calibrated after a series of validated
laboratory. The principal objective of the ADREAM project is estimations and hypotheses, which were based on in situ
to give rise to a building of optimized energy consumption and measurements, sensor data, and information extracted from a
production, while achieving periodical states of positive survey to the ADREAM’s occupants. Table 1 shows the
energy, without compromising user comfort. Figure 2 shows evolution of the statistical indicators (NMBE (Normalized
the interconnection of the various energy systems installed in Mean Bias Error) and CVRMSE (Coefficient of Variation of
the building. The specific details of the ADREAM Energy Root Mean Square Error)) from the initial to the final
Network, the Database Supervision and Control System are simulation, while Figure 4 illustrates the fit between measured
given in Papas et al. [14]. data (T REAL) and simulated data (T INITIAL
SIMULATION) and (T FINAL SIMULATION) for the
evolution of the interior temperature of an office [14].
TABLE 1: Statistical indicators for the initial and final simulations

Statistical T° INITIAL SIMULATION T° FINAL SIMULATION

Indicator
NMBE 4.57% 0.48%
CV(RMSE) 5.51% 2.84%

,&5(5$ 
WK,QWHUQDWLRQDO&RQIHUHQFHRQ5HQHZDEOH(QHUJ\5HVHDUFKDQG$SSOLFDWLRQV 3DULV)5$1&(2FW

behind the use of these variables has to do with the fact that
they all depend on each other. At the same time, by studying
the evolution of the water temperature, the estimation of the
total electrical power consumption can be deduced and
consequently predicted.
A dynamic non-linear filtering was implemented for the
prediction of the return water temperature based on its past
outputs and on the produced water temperature inputs. The
method used is that of the Non-Linear Autoregressive Model
with External Inputs (NARX), where an output y(t) is predicted
for a given d amount of past y(t) and x(t) values. Equation 1
represents the NARX model applied.

Fig. 4. Example of model calibration for an office during a 1-month period

(1)
The final step for carrying out a complete simulation Where:
including the building’s thermal behavior analysis, as well as y(t) = Return Water Temperature [°C]
its electrical consumptions, involved the entry of all the HVAC x(t) = Produced Water Temperature [°C]
systems’ parameters. That is, in order to launch a complete d = Sample data
simulation the required information concerned all the technical
parameters of: a) the three heat pumps, b) the central AHU and As it can be observed in Figures 4 and 5, the fit between
the five individual AHU’s, c) the ceiling fan coils, and d) the model and measured data falls into an accepted level. The two
radiators. As shown by Papas et al [14], several strategies for figures compare the ANN model output with the measured data
reducing the energy consumption of ADREAM were evaluated for the produced water and return temperature of the heat
with the potential of reducing the total electrical energy pumps. The statistical indicators corresponding to this
consumption by 20%. Most notably, after the validation of the correlation are, for Figure 4: NMBE = 0.23% and CV(RMSE)
complete model, performing a parametric variation on the = 1.93%, and for Figure 5: NMBE = -0.62% and CV(RMSE) =
functioning parameters of the building’s water distribution 1.90%. The development of such a precise model is due to the
system produced very promising results for the total electrical fact that the algorithm is improved through the “feeding” of the
consumption. Consequently, by replacing the constant flow past outputs during the training of the network [20].
circulation pumps with variable flow ones the building’s
energy gain was close to or superior to the simulated gain, as
shown in Table 2.
TABLE 2: Comparison of the building’s electrical consumption between 2014
and 2017 + 2018 following the implementation of optimization strategies

HVAC HVAC
Consumption Consumption
2014 [MWh] 2017 [MWh]
Jan.-Mai + Sep.- 83,61 70,57
Dec. Energy Gain 16%
HVAC HVAC
Consumption Consumption
2014 [MWh] 2018 [MWh]
Jan. - June 44,87 34,669
Fig. 5. Produced water temperature [°C]
Energy Gain 23%

IV. BLACK BOX MODELING WITH ARTIFICIAL NEURAL

NETWORKS
For very complex systems and interactions, such as the
ones corresponding to the building ADREAM, it is often useful
to use a “Black Box” modeling method through the
exploitation of Artificial Neural Networks (ANN) [17]. The
“Black Box” process approximates the physical equations
governing the interaction of energetic systems by applying
linear differential equations which associate a specific output to
a finite number of past inputs and outputs [14]. In the case of
the ADREAM building, the modeling of the HVAC systems
was pursued for the purpose of their performance optimization
and their overall regulation. For the modeling of the heat
pumps the input values used were the Electrical Power of the
Heat Pumps, and the Return Water Temperature, while the Fig. 6. Return Water Temperature [°C]
output used was the Produced Water Temperature. The purpose

,&5(5$ 
WK,QWHUQDWLRQDO&RQIHUHQFHRQ5HQHZDEOH(QHUJ\5HVHDUFKDQG$SSOLFDWLRQV 3DULV)5$1&(2FW

Since the purpose of this study is the development of a VI. CONCLUSION

global model that has the ability to predict power consumption This paper explored three different modeling
through valid simulated data, independent of the registered
methodologies for the optimization of the thermal and
data, it is essential to produce the input of the return
temperature through a model. Thus, a global model combining electrical consumption of the ADREAM building. The first
ANN and physical elements was developed with Simulink for method involved the calibration of a Dynamic Thermal
the purpose of simulating the interaction of all the systems with Simulation for the purpose of analyzing the thermal behavior
each other and with the building. However, simulating the of the building and reducing its electrical consumption. The
interaction between ANN and physical elements has an impact second one consisted of the “Black Box” modeling method
on the robustness of the ANN models. For this reason, a more through the exploitation of Artificial Neural Networks (ANN)
robust physical model comprising the totality of the systems with the objective of exploring optimized control strategies.
and the buildings was explored. The third method presented a global model developed in
Simulink incorporating all systems of consumption in
V. PHYSICAL MODELING IN SIMULINK interaction with the building. All modeling methodologies
proved to be very precise tools according to the calculation of
The complete physical model of the ADREAM building, statistical indicators, such as the RRMSE and NMBE.
along with all the HVAC systems installed is developed
However, for the purpose of advancing further the objectives
through the use of Matlab and the Simulink interface.
Exploring this methodology allows the detailed mathematical of the ADREAM project to be part of the ZEB research
description of each physical entity and system through a model community, the last model will be explored more thoroughly
which can simulate their interaction. At the same time, the in combination with electrical network models for the
simulation outputs are equivalent to those of an electrical development of an efficient strategy involving energy
network model such as an energy production network or an production, consumption, and storage. The final aim of this
electrical storage network. project involves the exploration of more efficient control
strategies for the optimization of the thermal and electrical
For the development of the Simulink model, the physical energy consumption through an intelligent management of
equations of the HVAC systems were either manually entered energy, providing a prototype model which can serve as an
or modeled through physical and thermal elements of the
example for the regulation and controls strategies of a
Simscape library. The main model components involved the
sustainable building.
Heat Pumps System, the Geothermal System, the Water
Storage, the Fan Coils, and the Building. Figure 7 shows the ACKNOWLEDGMENT
complete model of the ADREAM Building and its interaction
with the Heat Pump, the Geothermal System, and the Water This project has been financed by the University of
Storage Units. Simulating the energy needs of the building Toulouse III – Paul Sabatier in collaboration with the research
according to its temperature set-points for a given period, the association Neocampus.
model can produce the thermal energy consumed. Knowing the
value of the Heat Pumps’ Coefficient of Performance (COP)
Fig. 7. Global Model of Building + HVAC Systems in Simulink
for each season, allows the calculation of the total electrical
consumption.

,&5(5$ 
WK,QWHUQDWLRQDO&RQIHUHQFHRQ5HQHZDEOH(QHUJ\5HVHDUFKDQG$SSOLFDWLRQV 3DULV)5$1&(2FW

REFERENCES >@ A.T. Nguyen, S. Reiter, An investigation on thermal performance of a low

cost apartment in hot humid climate of Danang, Energy and Buildings 47
(2012) 237-246
>@ http://ec.europa.eu/energy/en/topics/energy-strategy/2030-energy- >@ System Identification Toolbox – ARX Models.
strategy <http://www.rohan.sdsu.edu/doc/matlab/toolbox/ident/ch2gui20.html>.
(accessed 21.01.2017). >@ T. Salque, D. Marchio, “Development of a Neural Network-based Building
>@ P. Torcellini, S. Pless, and M. Deru, D. Crawley, “Zero Energy Model and Application to Geothermal Heat Pumps Predictive Control”,
Buildings: A Critical Look at the Definition”. Preprint NREL ACEEE SIMUL 2012: The Fourth International Conference on Advances in System
Summer Study Pacific Grove, California August 14−18, 2006. Simulation (2012) 24-29.
>@ Anna Joanna Marszal, Per Heiselberg « A literature review of Zero >@ G. Xu, A. Kusiak, “Modeling and optimization of HVAC systems using a
Energy Building (ZEB) definitions ». December 2009© Aalborg dynamic neural network”, Energy 42 (2012) 241-250.
University. ISSN 1901-726X. DCE Technical Report No. 78. >@ https://fr.mathworks.com/help/nnet/ug/design-time-series-narx-feedback-
>@ A.J. Marszal, P. Heiselberg, J. S. Bourrelle, E. Musall, K. Voss, A. neural-networks.html.
Napolitano. « Zero Energy Building – A review of definitions and >@ S. Kirmani, et al., “Effective low cost Grid-Connected Solar Photovoltaic
calculationmethodologies » Energy and Building (2011) System to Electrify the Small Scale industry/Commercial Building”,
>@ L Wang, J Gwilliam, P Jones, “Case study of zero energy house design International Journal of Renewable Energy Research, Vol. 7, No 2, 2017.
in UK”, Energy and buildings 41 (2009) 1215-1222
>@ Isa, N. B. M., Wei, T. C., & Yatim, A. H. M., Smart grid technology: >@ M. Elloumi, at al., “Multi Agent System Design for PV Integrated Home
Communications, power electronics and control system, Sustainable Management”, International Journal of Renewable Energy Research,Vol.
8, No 1, 2018.
Energy Engineering and Application, International Conference (2015)
10-14. >@ A. Kane, et al., “Performance Enhancement of Building Integrated
>@ Mohamed, M. A., Eltamaly, A. M., Farh, H. M., & Alolah, A. I., Energy Photovoltaic Module using Thermoelectric Cooling”, InternationalJournal
management and renewable energy integration in smart grid system. of Renewable Energy Research, Vol. 3, No 2, 2013.
Smart Energy Grid Engineering (SEGE), 2015 IEEE International >@ E. Tarigan, “Simulation and Feasibility Studies ofRooftop PV System for
Conference University Campus Buildings in Surabaya, Indonesia”, International
(2015) pp. 1-6. Journal of Renewable Energy Research, Vol. 8, No 2, 2018.
>@ Solaris Positive Energy, Solaris Concept: The first reproducible green >@ Abhinaya K.S. et al., ”Assessment and Remodelling of a Conventional
office building with positive energy, (2012) <http://www.solaris-energie- Building Into a Green Building Using BIM”, International Journal of
positive.com/solaris-positive-energy-saving-positive-energy- Renewable Energy Research, Vol. 7, No 4, 2017.
building.php>. >@ J. De Leon-Ruiz, et al., “Mathematical Thermal Modelling of a Direct-
>@ Randall, T., The Smartest Building in the World: Inside the Connected Expansion Solar-Assisted Heat Pump Using Multi-Objective Optimization
Future of Architecture, (2015) <http://www.bloomberg.com/ Based on the Energy Demand”, Energies, 11, 1773, 2018
features/2015-the-edge-the-worlds-greenest-building/>.
>@ https://www.laas.fr/public/en/adream. >@ S. Chokchai, et al., “A Pilot Study on Geothermal Heat Pump (GHP) Use
>@ PCVue Solutions, <http://www.pcvuesolutions.com>. for Cooling Operations, and on GHP Site Selection in Tropical Regions
>@ Izuba Energies, Pleiades+Comfie, <http://www.izuba.fr/logiciel/ Based on a Case Study in Thailand”, Energies, 11, 2356, 2018.
comfie>. >@ W. El-Baz, et al.,“Experimental Study and Modeling of Ground-Source
>@ B. Peuportier, Bancs d’essais de logiciels de simulation thermique, Heat Pumps with Combi-Storage in Buildings”, Energies, 11, 1174, 2018.
Journée Thématique SFT-IBPSA (2005). >@ A. Arteconi, et al., “Assessing the Demand Side Management Potential and
>@ I. Papas, C. García Rodríguez, B. Estibals; C. Ecrepont, C. Alonso, the Energy Flexibility of Heat Pumps in Buildings”, Energies, 11, 1846,
“ADREAM:Energy Consumption Optimisation through Dynamic 2018.
Energetic Simulations for an Intelligent Management of Energy”, 2016 >@ S. Staub, et al., “Reversible Heat Pump–Organic Rankine Cycle Systems
Intl IEEE Conferences on Ubiquitous Intelligence & Computing, for the Storage of Renewable Electricity”, Energies, 11, 1352, 2018.
Advanced and Trusted Computing, Scalable Computing and
Communications, Cloud and Big Data Computing, Internet of People,
and Smart World Congress, p 975 –983.
>@ ASHRAE, ASHRAE Guideline 14-2002, Measurement of Energy and
Demand Savings, ASHRAE, Atlanta, 2002.

,&5(5$