A Sensor Network for Existing Residential Buildings Indoor

Environment Quality and Energy Consumption Assessment and

Monitoring: Lessons Learnt from a Field Experiment

Mathieu Bourdeau

1,2

, David Werner

, Philippe Basset

and Elyes Nefzaoui

CAMEO SAS 55, Rue de Châteaudun, 75009, Paris, France

Université Paris-Est, ESYCOM (FRE2028), CNAM, CNRS, ESIEE Paris, Université Paris-Est Marne-la-Vallée,

F-77454 Marne-la-Vallée, France

Keywords: Sensor Network, Energy Monitoring, Building Energy Efficiency, Energy Retrofit.

Abstract: Enhancing residential buildings energy efficiency has become a critical goal to take up current challenges of

human comfort, urbanization growth and the consequent energy consumption increase. In a context of

integrated smart infrastructures, sensor networks offer a relevant solution to support building energy

consumption monitoring, operation and prediction. The amount of accessible data with such networks also

opens new prospects to better consider key parameters such as human behaviour and to lead to more efficient

energy retrofit of existing buildings. However, sensor networks planning and implementation in general, and

in existing buildings in particular, is a particularly complex task facing many challenges and affecting the

performances of such a promising solution. In the present paper, we report on a field experiment of a sensor

network deployment involving more than 250 sensors in three collective residential buildings in Paris region

for the evaluation of a deep energy retrofit. More specifically, we describe the whole process of the sensor

network design and roll-out and highlight the main critical aspects in such complex process. We also provide

a feedback after several months of the sensor network operation and preliminary analysis of collected data.

Reported results path the way for an efficient and optimized design and deployment of sensor networks for

energy and indoor environment quality monitoring in existing buildings.

1 INTRODUCTION

Residential buildings are one of the major energy

consumers and greenhouse gas (GHG) emitters, with

38.1% of the final energy consumption and 36% of the

GHG emissions in Europe (ADEME, 2015). Efforts

have been made to reduce the impact of residential

buildings on energy consumption with various codes,

standards and thermal regulation mandatory

compliance for designs of new buildings (ASHRAE,

2013; HARMONIE, 2017). However, given the slow

turnover of the building stock in European countries,

the effect of these regulations is limited (INSEE,

2017). Hence, existing building retrofit turns into a

priority (ADEME, 2018). On the other hand, ambitious

massive smart sensor networks plans (European

Commission, 2019; European Parliament & European

Council, 2009; French Ministry of Ecological and

Sustainable Transition, 2016) have been launched in

recent years to enhance buildings energy efficiency.

the other hand, sensor networks have been

largely deployed in buildings for energy monitoring

and operation (Fan, Xiao, Li, & Wang, 2018) or

building energy consumption forecasting (Bourdeau,

Zhai, Nefzaoui, Guo, & Chatellier, 2019; Jain, Smith,

Culligan, & Taylor, 2014). Studies have used sensors

to highlight and characterize the link between energy

efficiency and inhabitants’ behaviour (Li & Lim, 2013;

Pisello & Asdrubali, 2014), identified as one of the

main source of energy performance gaps (de Wilde,

2014). Data collection solutions have also been

proposed to supervise building energy retrofits (Calì,

Osterhage, Streblow, & Müller, 2016; CSTB, 2016;

Jankovic, 2019). However, instrumentation solutions

deployment is a complex task, facing many challenges

and difficulties. Thus, it directly impacts on the quality

of the analyses and the efficiency of related energy

savings measures (Calì et al., 2016; Jankovic, 2019;

Pisello & Asdrubali, 2014).

In this context, we present lessons learnd from a

case study of a sensor network deployed in three

existing collective residential buildings in France. The

Bourdeau, M., Werner, D., Basset, P. and Nefzaoui, E.

A Sensor Network for Existing Residential Buildings Indoor Environment Quality and Energy Consumption Assessment and Monitoring: Lessons Learnt from a Field Experiment.

DOI: 10.5220/0008979401050112

In Proceedings of the 9th International Conference on Sensor Networks (SENSORNETS 2020), pages 105-112

ISBN: 978-989-758-403-9; ISSN: 2184-4380

105

instrumentation takes part in a larger project that aims

to study the impact of deep energy retrofit measures on

the studied building energy consumption.

The present paper is organized as follows: we

present the project case study in Section 2, then, we

describe details of the instrumentation solution in

Section 3. A feedback on the main critical points for

sensor networks implementation in existing buildings

is proposed in Section 4. Finally, Section 5 presents

lessons learned from preliminary analysis of data

collected after several months of operation of this

experimental sensor network.

2 PRESENTATION OF THE CASE

STUDY

In the present research project, a group of three existing

social residential buildings – 63 apartments over a

4,600 m² living area – is considered for field

experimentation. Buildings were built in 1974 in Paris

(France) eastern suburb. Shared areas of the three

buildings and a 10-apartment sample are selected for a

three-year instrumentation period, starting in March

2019. The ten apartments comprise surfaces from 50 to

75 m² distributed in the three studied buildings, on

different floors and with different orientations.

Apartment case studies have one to two occupants of

various ages, professional occupation and living habits.

Case studies only had minor retrofit actions. They

are now planned to undergo deep energy retrofit

measures over an eighteen-month period in 2020-2021.

Retrofit measures will be implemented on occupied

site with tenants living in their apartments during the

whole retrofit period.

3 INSTRUMENTATION

METHODOLOGY

3.1 Instrumentation Plan

3.1.1 Purpose of the Instrumentation

The sensor network deployed aims to provide a multi-

scale and multi-targets wireless monitoring solution. It

focuses on data collection at both building- and

apartment-scale for energy consumption, outdoor and

indoor conditions, and inhabitants’ comfort and

behaviour. The installed sensor network needs to be as

non-intrusive as possible. This is essential to ensure

that the participants’ comfort and living habits remain

undisturbed to prevent any bias in the experiment.

The instrumentation solution described in the

present paper is the first step of a larger study. Data

collected through the sensor network should further

serve three consecutive purposes (Figure 1). First, data

analysis should highlight behavioural patterns and

energy drivers. Results should be used to obtain a

calibrated energy model of the buildings prior to

retrofit actions to predict their impact on buildings

energy behaviour and latter identify performance gaps

through a comparison with post-retrofit operation data.

Figure 1: Description of the main steps of the research project

starting with instrumentation planning and deployment.

3.1.2 Sensor Characteristics and Deployment

The sensors selection and deployment are performed in

two different phases, to set a total number of 259

sensors and connected objects (Table 1). The first

deployment phase covers shared areas and entire

building level. Although it involves only 10% of the

total number of sensors for 35% of the total cost. In this

phase, sensors provision, installation and data

communication are entirely managed by a hired

contractor. Four categories of measurements are

targeted:

 overall and detailed electricity demand in shared

areas is monitored on the main electricity meters

and switchboards. A one-minute measurement

time-step is used to capture small electric events

such as for timed lighting and lifts usage;

 building-scale thermal energy consumption is

obtained using ultrasound thermal energy meters

with a five-minute time-step (Figure 2);

 occupancy is assessed using infrared presence

detection sensors positioned at the main entrance

door of the three buildings;

 combined indoor temperature and humidity

sensors are positioned on three floors of each

building – ground floor, middle and last floor.

Since variations of indoor temperature and

humidity are not expected to abruptly change over

time, hourly measurement time-step is set;

SENSORNETS 2020 - 9th International Conference on Sensor Networks

106

 a weather station is positioned on a nearby

university building to monitor local air

temperature, humidity, wind speed, wind

direction, rainfall and solar irradiation at 5-minute

time-step.

This first instrumentation step is now fully

operational, except for the weather station which is

currently being installed.

Figure 2: Installation example of a thermal energy meter for

heating energy demand monitoring (right).

The second part of the sensors focuses on the

characterization of apartments (aside from building-

scale monitoring of domestic hot water consumption

delayed to this second phase). This second deployment

phase should start in November 2019. Categories of

measurements are:

 electric power demand from the main electric

meter and switchboard, and smart-plugs to

monitor all the main appliances of the

instrumented apartments (one-minute time-step);

 hot water consumption assessed using contact-

temperature sensors on hot water pipes (one-

minute time-step) and completed with data from

already-installed remote reading volumetric water

meters;

 heating energy consumption is deduced in a

similar way using contact-temperature sensors

installed on heaters (hourly time-step);

 natural gas consumption, only used for cooking, is

monitored with sensors installed on apartments

gas meter (one-minute time-step);

 indoor conditions and comfort should be captured

using contact-temperature sensors on exterior

walls of the apartments and a sensor combining

indoor temperature, humidity, luminosity, CO

and presence measurements (hourly time-step);

 occupants’ behaviours should mainly be

characterized through occupation data and using

sensors for window opening-closing detection.

As for building-scale measurements, the choice of

time-steps is driven by the events sensors are targeting

and the purpose of data analyses. For electricity, hot

water and gas consumption, usages may only last a few

minutes. Therefore, the smaller time-step the better.

Regarding, heating energy consumption and indoor

environment, they are more likely to change on an

hourly basis. Finally, window-opening detection is

event-driven while occupation detection is linked to

indoor environment monitoring time-step.

3.1.3 Communication Network and Online

Data Collection

Data communication for the present sensor network is

entirely wireless (Figure 3). Several, communication

protocols are available such as Modbus (The Modbus

Organization, 2019), Sigfox (Sigfox, 2019) or

LoRaWan. The latter is currently the most common for

applications such as the one presented in this paper

(Augustin et al., 2016) and most sensors available on

the French market use LoRa technology. Indeed, LoRa

is fit for IoT projects and wireless sensor networks

deployed in smart buildings and smart cities contexts

(Centenaro, Vangelista, Zanella, & Zorzi, 2016;

Pasolini et al., 2018). It is relatively easy to implement,

it provides information on the sensor status and it

allows long-range data transmission even in areas with

difficult access. LoRa technology is also energy-

effective as it relies on radio waves with low

communication rate and then low energy consumption.

Several implementation strategies have been selected

for the two presented instrumentation phases. The first

phase uses an operated LoRa network with two

gateways installed onsite. Onsite LoRa communication

tests led to the conclusion that a single gateway would

be sufficient. However, a second one is installed in case

of reliability. Only the electricity consumption-

dedicated sensors use a GPRS network because of their

acquisition time-step and the limitations of operated

LoRa network in terms of bandwidth usage (Electronic

Communications Committee, 2019).

The second phase is processed differently and

separated into three sub-phases. As for shared areas,

energy monitoring is entirely managed by a contractor.

Other sensors are provided and configured by a

different service company and installed onsite by the

research group in charge of the project.

A Sensor Network for Existing Residential Buildings Indoor Environment Quality and Energy Consumption Assessment and Monitoring:

Lessons Learnt from a Field Experiment

107

Table 1: Characteristics summary of the deployed sensors.

Sensor

Number at

building scale

(per building)

Number

per

apartment

Total

number

Acquisition

time-step

Communication

protocol

Indoor temperature &

humidity

3 / 9 Hourly Operated LoRa

Presence 1 / 3 Hourly Operated LoRa

Combined indoor temperature,

humidity, luminosity, presence

& CO

level

/ 1 10 Hourly Private LoRa

Exterior walls surface

temperatures

/ 2-5 32 Hourly Private LoRa

Window opening detection / 3-5 38 Event-driven Private LoRa

Gas consumption / 0-1 4 One-minute Private LoRa

Electricity demand – main

electric meter

2-3 1 17 One-minute GPRS

Electricity demand –

switchboard

0-1 1 11 One-minute GPRS

Electricity demand –

smartplugs

/ 5-9 67 One-minute Private LoRa

Heating energy demand –

thermal energy meter & pulse

sensor

1 + 1 / 3 + 3 Five-minute Operated LoRa

Heating energy demand –

radiators

/ 3-5 36 Hourly Private LoRa

Domestic hot water energy

demand – thermal energy

meter & pulse sensor

1 + 1 / 3 + 3 One-minute Private LoRa

Domestic hot water energy

demand – apartments

/ 2 20 One-minute Private LoRa

Finally, a ten-sensor sample (those combining

measurement of indoor temperature, humidity,

luminosity, presence and CO

level for apartments) is

entirely set up and installed by the research group.

The used LoRa network is configured as private

network to bypass bandwidth usage constraints

related to data acquisition time-step.

For all sensors, collected data are retrieved on

three FTP servers: a first server used by the contractor

for managing part of the instrumentation solution, a

second sensor for equipment installed by the research

group and a third server dedicated to data retrieved

from the weather station. The final format provides

collected raw data with one csv file for each day and

for each sensor. Information include the sensor

identification and location, the type of measurements,

the timestamp, the measured values and the units.

Figure 3: Simplified diagram of the wireless sensor network.

3.2 Instrumentation Management

Process

Instrumentation overall management has a crucial

impact on the success of the instrumentation and on the

whole project. In the present study, it can be divided

into six different stages over an estimated timespan of

SENSORNETS 2020 - 9th International Conference on Sensor Networks

108

one year and four months (started in October 2018 and

expected to end in January 2020).

First is the accurate definition of the

instrumentation specifications to meet the needs of the

project. In parallel, a review of service providers

offering adapted solutions is performed and apartments

are recruited to participate to the instrumentation study.

Once specifications are organized in a clear

framework, they are submitted to identified

contractors. Several rounds of discussions allow

modifications on the specifications. It aims to adapt to

the reality of the market and for service providers to

refine their offer to meet the needs of the project to the

best of their capabilities. It also helps select the most

relevant contractors. In the present case thirty-three

service providers were contacted and two contractors

were finally selected.

Once a contract is signed, sensors are ordered,

delivered and installed under the supervision of the

research group. Finally, when the sensor network is

functional, data quality is checked. It ensures that all

sensors are properly measuring and transmitting data,

that there are no missing information or other issues

and that the proper data format is displayed for future

applications.

4 IDENTIFTICATION OF

CRITICAL POINTS FOR

SENSOR NETWORK

IMPLEMENTATION

The planning and deployment of the proposed

instrumentation solution has highlighted several

critical aspects in sensor network implementation.

These critical points arose from various issues

encountered during the instrumentation process. These

can be grouped into four categories with i) onsite

installation conditions and environment, ii)

characteristics and purpose of instrumentation, iii)

service provider and iv) project tracking and

management.

4.1 Onsite Installation Conditions and

Environment

Characteristics of the experimentation site have a

significant impact on the feasibility of a monitoring

project. As a first step, it is crucial to identify the

equipment and installations. Equipment conditions are

important, as for security reasons some may not be

monitored. Indeed, over the selected ten-apartment

sample, seven electrical switchboards could not be

equipped because of their obsolescence and should be

replaced for the experimentation. Sensor connectivity

must also be considered for sensors placement and to

choose the relevant number of gateways necessary

since metal stairs and doors disrupt radio waves and

then data transmission. Finally, the installation

environment should be documented. For instance,

several sensors do not use batteries but need a main

power supply – thermal energy meters for instance. In

the buildings, such electrical installations were not

adapted or even not existing where they were needed.

Thus, additional costly actions were necessary to

prepare the installation site.

The present instrumentation solution is also

specific because all buildings remain occupied during

the sensor network deployment and the whole

experimentation. Then, the sensor network must be as

non-intrusive as possible. However, non-intrusive

sensors imply higher costs due to specific technologies.

For instance, it applies with ultrasound thermal energy

meters and with all electricity-related meters that need

to be small, easy movable and without heavy

installation work on the existing equipment.

Also, as the project mostly relies on collected data

from apartments and occupants’ behavior, it is

necessary to recruit participants. Then, recruitment is a

crucial step for the success of the research project, but

is particularly time-consuming, requiring several

rounds of presentations and visits (general

presentation, door-to-door visits, phone calls and

individual meetings). In the project, up to three months

were necessary to recruit ten households. Moreover, it

is crucial that inhabitants find clear benefits in their

participation. In this case it was proposed for

inhabitants to have access to all collected personal data

and summary analysis results for their households.

Finally, as most collected data are personal data, a

specific attention should be given to related regulations

and mandatory administrative actions (French Ministry

of Economy, 2019).

4.2 Characteristics and Purpose of the

Instrumentation Solution

The introduced sensor network serves for a research

project with specific needs. Therefore, it differs from

usual commercial monitoring studies because it aims

to investigate situations and details that would usually

not be considered. More specifically, there are a wide

range of acquisition time-steps (from one minute to one

hour) and a large variety of sensors and measurements

as presented in Table 1.

Consequently, such a complex instrumentation

solution faces several challenges. For instance, sensors

A Sensor Network for Existing Residential Buildings Indoor Environment Quality and Energy Consumption Assessment and Monitoring:

Lessons Learnt from a Field Experiment

109

are designed and manufactured for standardized usage

conditions depending on the environment of

installation and usually with hourly acquisition time-

step. Their internal memory stores up to one day of data

(24 measurements) communicated once a day. Then

using sub-hourly acquisition time-steps requires more

frequent data communication which is energy

intensive and significantly reduces the expected battery

lifetime. However, sensors such as for gas

consumption metering (Zone-atex.fr, 2019) or used

outdoor or in damp and dusty environments are airtight

and cannot be opened once installed. This prevents the

battery to be replaced and induces a costly replacement

of the whole sensor instead.

Sensor technologies and measurement

characteristics (especially the acquisition time-step)

also have an impact on the design of the sensor

network. There is first the choice between the different

existing communication protocols with different

characteristics, advantages and constraints. LoRa is the

prominent solution on the market and was selected in

the present case study but it presents drawbacks. For

instance, operated LoRa networks restrict bandwidth

usage. Then having a one-minute time-step on large

amounts of sensors is hardly possible. Private LoRa

networks offer a bypass solution since they are locally

set networks dedicated to one specific sensor network.

However, they are more complicated to implement.

Other solutions such as GPRS used for some specific

sensors also raise different issues related to wireless

monitoring technologies and electro-sensitivity.

4.3 Service Providers and Contractors

Through this study, service providers have been hired

to provide an instrumentation solution adapted to the

goals of the research project. On the French IoT

market, there are many different contractors. However,

they do not all provide the same delivery. Most

commercial proposals focus on sensor provision but do

not include set up and installation. Therefore, such

contractors must partner with other companies to

match the needs of a project. This often leads to many

communication issues and delays. Furthermore, when

a contract is established, many terms must be carefully

checked to ensure a smooth project process, and more

particularly: i) responsibilities and guarantees about

the sensors/network and about the potential provision

and installation delays, ii) maintenance details and

conditions when included, and iii) what is done by the

contractor with collected data during and after the

contract validity.

It is also necessary to ensure the service quality and

expertise of contractors. Indeed, there is a significant

difference between common “plug and play” sensors –

such as for temperature or humidity sensors – and

energy metering. The latter requires very specific

knowledge on the technologies, installation procedures

and data acquisition checking. Many contractors do not

have such equipment and need to partner with other

manufacturers. This lack of expertise often leads to

many future issues with sensor calibration or

equipment failures.

Finally, the budget management requires a detailed

investigation. Indeed, there can be several unexpected

expenditure items that can significantly increase the

total instrumentation budget: include miscellaneous

accessories, connectivity subscriptions, details of all-

inclusive maintenance contracts, installation and setup,

or site visits. A budget summary from the ten complete

commercial proposals received is presented in Figure

Figure 4: Budget summary for the presented sensor network.

4.4 Project Management

The overall project management, comprising

supervision on the three previous subsections, is the

key to a smooth sensor network deployment and

optimal valorisation. Regarding the instrumentation, a

detailed and precise tracking is mandatory to identify

every sensor with sensor type, measurements,

communication protocol, installation date and location.

Adapted and controlled communication is essential

within the research group to prevent any loss of

information and miscommunication. It also must be

ensured with other actors and particularly with

building occupants. Indeed, it is recommended to avoid

any over-soliciting with inhabitants for obvious

privacy reasons. Finally, as the present research project

is a multi-disciplinary study combining several

research teams, an optimal coordination is needed.

SENSORNETS 2020 - 9th International Conference on Sensor Networks

110

5 PRELIMINARY DATA

ANALYSIS FROM THE FIRST

INSTRUMENTATION PHASE

The first instrumentation phase focusing on building

common areas is fully operational. Thus, a statistical

data analysis is performed on an observation window

from 2019/02/19 to 2019/10/06 to highlight issues in

collected data. Twenty sensors are considered

including data for temperature & humidity (9 sensors),

passage counting (3 sensors), thermal energy (3

meters) and electric power demand monitoring (5

sensors).

The main highlighted issue relates to missing

datapoints. Over the given observation period, 92.7%

of the expected datapoints are collected. Figure 5

compares the evolution of the expected number of

measurements (blue curve) and of the effective

collection results (orange curve). The percentage of

acquired data over time is presented on the green curve.

Figure 5: Evolution of collected datapoints (orange)

compared to expected collected datapoints (blue) and

percentage of acquired data (green) from 2019/02/19 to

2019/10/06.

The major gap is due to a 17-day-long sensor

malfunction in May 2019 for the electrical switchboard

monitoring. Electricity demand data collection is

therefore the category with the least collected

datapoints (on Figure 6: 92.0% of the optimal target),

followed by thermal energy monitoring (94% of data

collected). Indoor environment and passage counting

account for more than 99% of collected data. However,

large amounts of missing energy datapoints should also

be transposed to time-scale analysis. Indeed, electric

sensors provide one-minute time-step information.

Hence, over a 230-day observation period, it would be

equivalent to a loss of less than 19 days of data

concentrated on a specific period for one specific

sensor.

Other observed outliers are abnormal and

additional measurements. Abnormal values are only

found for energy demand monitoring, with overly high

measurements or zero-values. They respectively

represent 0.06% and 6.8% of total. Zero-values are

almost exclusively related to electrical switchboard

monitoring (99.9% of abnormal measures).

Figure 6: Comparison of collected datapoints (orange)

compared to expected collected datapoints (blue) for the

different monitoring categories.

6 CONCLUSIONS AND FUTURE

WORK

This paper presents the planning, deployment and

supervision of a sensor network for existing residential

buildings monitoring. The proposed solution is a multi-

scale sensor network covering thermal and electrical

energy consumption, indoor environment and comfort,

inhabitants’ behaviour and weather conditions, at

various timescales.

The instrumentation solution has been entirely

planned and divided into two implementation phases.

The first phase focuses on the monitoring of building

shared areas and is operational. The second phase

considers a similar approach on a ten-apartment

sample and will be deployed starting from November

2019. Details of both phases are presented with a

description of sensor characteristics, communication

networks and overall project management.

A feedback on the first phase of the sensor network

implementation is provided. It shows the obstacles and

challenges when implementing large sensor networks

with a wide diversity of measurements in existing old

buildings. It highlights several critical points to be

considered for similar future applications. These

critical aspects are grouped in four categories with

onsite installation conditions and environment,

characteristics and purpose of the instrumentation

solution, service provider and contractors and project

management. A preliminary statistical analysis of

A Sensor Network for Existing Residential Buildings Indoor Environment Quality and Energy Consumption Assessment and Monitoring:

Lessons Learnt from a Field Experiment

111

collected data is also provided. It highlights the main

encountered issues in data collection.

In future works, data analysis will be extended to

measurements to the second instrumentation phase

data. Moreover, in order to ensure long-term

measurement accuracy and to prevent future avoidable

data loss and highlighted data collection issues, a long-

term calibration study will be conducted. Finally,

collected data will be integrated in building energy

models to characterize the buildings behaviours and

investigate potential issues in building energy

management.

ACKOWLEDGMENT

This work was supported by the I-SITE FUTURE

Initiative (reference ANR-16-IDEX-0003) in the frame

of the project ANDRE.

REFERENCES

ADEME. (2015). Climat, air et énergie |Chiffres-clés

[Climate, air and energy | Key numbers]. ADEME.

ADEME. (2018). Le plan de rénovation énergétique de

l’habitat (PREH) [The residential energy retrofit plan].

ASHRAE. (2013). Standard 90.1-2013 - Energy Standard

for Buildings Except Low-Rise Residential Buildings (SI

Edition). Atlanta: American Society of Heating,

Refrigerating and Air-Conditioning Engineers.

Augustin, A., Yi, J., Clausen, T., Townsley, W., Augustin,

A., Yi, J., … Townsley, W. M. (2016). A Study of LoRa:

Long Range & Low Power Networks for the

Internet of Things. Sensors, 16(9), 1466.

Bourdeau, M., Zhai, X., Nefzaoui, E., Guo, X., & Chatellier,

P. (2019). Modeling and forecasting building energy

consumption: A review of data-driven techniques.

Sustainable Cities and Society, 48, 101533.

Calì, D., Osterhage, T., Streblow, R., & Müller, D. (2016).

Energy performance gap in refurbished German

dwellings: Lesson learned from a field test. Energy and

Buildings, 127, 1146–1158.

Centenaro, M., Vangelista, L., Zanella, A., & Zorzi, M.

(2016). Long-range communications in unlicensed

bands: the rising stars in the IoT and smart city scenarios.

IEEE Wireless Communications, 23(5), 60–67.

CSTB. (2016). ISABELE mesure la performance

énergétique intrinsèque d’un logement neuf [ISABELE

measuring instrinsic energy performance of new

housings].

de Wilde, P. (2014). The gap between predicted and

measured energy performance of buildings: A

framework for investigation. Automation in

Construction, 41, 40–49.

Electronic Communications Committee. (2019). ERC

Recommendation 70-03 |Relating to the use of Short

Range Devices (SRD).

European Commission. (2019). Smart Metering deployment

in the European Union.

European Parliament, & European Council. Directive

2009/72/CE du parlement européen et du Conseil du 13

juillet 2009 concernant des règles communes pour le

marché intérieur de l’électricité et abrogeant la directive

2003/54/CE [in French] (2009).

Fan, C., Xiao, F., Li, Z., & Wang, J. (2018). Unsupervised

data analytics in mining big building operational data for

energy efficiency enhancement: A review. Energy and

Buildings, 159, 296–308.

French Ministry of Ecological and Sustainable Transition.

Decree 2010-1022 du 31st août 2010 relatif aux

dispositifs de comptage sur les réseaux publics

d’électricité [in French] (2016).

French Ministry of Economy. (2019). Le règlement général

sur la protection des données (RGPD), mode d’emploi

[General regulations on data protection, instruction

manual].

HARMONIE. (2017). 1974-2020 : l’évolution de la

réglementation thermique [1974-2020 : evolution of the

French thermal regulation].

INSEE. (2017). Le parc de logements en France au 1

janvier

2017 [Housing stock in France on January 1, 2017].

Jain, R. K., Smith, K. M., Culligan, P. J., & Taylor, J. E.

(2014). Forecasting energy consumption of multi-family

residential buildings using support vector regression:

Investigating the impact of temporal and spatial

monitoring granularity on performance accuracy.

Applied Energy, 123, 168–178.

Jankovic, L. (2019). Lessons learnt from design, off-site

construction and performance analysis of deep energy

retrofit of residential buildings. Energy and Buildings,

186, 319–338.

Li, B., & Lim, D. (2013). Occupant Behavior and Building

Performance. In Design and Management of Sustainable

Built Environments (pp. 279–304). London: Springer

London.

Pasolini, G., Buratti, C., Feltrin, L., Zabini, F., De Castro, C.,

Verdone, R., & Andrisano, O. (2018). Smart City Pilot

Projects Using LoRa and IEEE802.15.4 Technologies.

Sensors, 18(4), 1118.

Pisello, A. L., & Asdrubali, F. (2014). Human-based energy

retrofits in residential buildings: A cost-effective

alternative to traditional physical strategies. Applied

Energy, 133, 224–235.

Sigfox. (2019). Sigfox - The Global Communications

Service Provider for the Internet of Things (IoT).

The Modbus Organization. (2019). The Modbus

Organization.

Zone-atex.fr. (2019). Normes ATEX et matériel ATEX

[ATEX standards and ATEX equipment].

SENSORNETS 2020 - 9th International Conference on Sensor Networks

112