Wireless Sensing For The Built Environment: Enabling InnovationTowards Greener, Healthier Homes

Elena I. Gaura*, James Brusey*, Ross Wilkins*, John Barnham**

* Cogent Computing Applied Research Centre, Faculty of Engineering and Computing,Coventry University, Coventry, CV1 5FB{e.gaura.j.brusey,wilkin24}@coventry.ac.uk

** Orbit Heart Of England, 10 Greenhill Street, Stratford-upon-Avon, Warwickshire, CV37 6LGjohn.barnham@orbit.org.uk

Abstract

Worldwide carbon reduction targets for the built en-vironment are staggeringly ambitious. If they are to beachieved, orders of magnitude performance increases arerequired from HVAC systems, construction techniquesand insulating materials. Given the limited understand-ing of many of the newer materials and techniques, ob-jective measurement is fundamental to meeting thesetargets in time. This paper presents the case for a holis-tic approach to measurement within the built environ-ment and shows how Wireless Sensor Networks (WSNs)are a prime candidate technology to support such anapproach.

WSNs are readily enablers of understanding in do-mains characterised by spatio-temporal, multivariate com-plexity. Simple, portable and non-intrusive WSN sys-tems, deployed for weeks or years, are powerful tools forempirical environmental and energy performance eval-uation of occupied dwellings. Coupled with structureddeployment processes and novel empirical evaluation met-rics, WSNs enable, for old stock, focused actions towardsreduced energy consumption, improved internal environ-ment, lower maintenance costs and maintaining the costviability of the building asset. They are equally valuablein the context of new builds, for generic and apportionedenergy consumption evaluations against the deliveredenvironmental quality and design expectations.

Keywords: WSN, Built Environment, Energy Perfor-mance, Occupant Comfort

1 Introduction

The industry's perception of wireless, embedded net-worked sensing systems (WSNs) as reliable, available,usable and a�ordable instruments is changing. To many,they are no longer the next wave in engineering andcomputing, but have become today's reality in assess-ment: be that of people, their well-being and actions,environments and their impact, machines and their per-formance. When widely adopted, this mighty technol-ogy, supported by miniature sensors, cheap computingpower and recent advances in wireless communications,is likely to induce a new information revolution.

It is apparent that WSNs could have a signi�cantrole to play in the built environment. Understanding,assessing and controlling buildings may be the killer ap-plication for networked wireless sensing, which will en-able the transit of this technology from research to wideadoption in a series of related industries. Applicationsof WSNs within the built environment industry havestrong business cases (albeit made at present by theWSN community) derived from:

1. A relatively mature base of WSN o� the shelfhardware which meets most of the requirementsfor measurement in the built environment.

2. Excellent economies of scale, commensurate withthe size of the industry and thus promising a dras-tic reduction in commercial WSN systems costs.

3. Strong political drive and ambitious decarboni-sation agendas which would clearly bene�t frommeasurement and quanti�cation: current carbonreduction targets are daunting for many nationalgovernments. In the UK for example, 80% reduc-tion is envisaged before 2050 and an aspirationhas been set for all new developments to be zerocarbon by 2016 [4]. Achieving such targets withbounded investment implies the need for continu-ous insight into actual performance of new builds,that of refurbishments and also evidence-based eval-uations of stock in general.

Within this context, the paper reports on the resultsobtained and insight gained so far by the authors duringtheir ongoing collaboration towards:

� Demonstrating the usefulness and appropriatenessof WSNs as tools for the built environment.

� Examining some of the conceptual and practicalissues to be overcome by the computer science andmodern built environment communities towardsexploiting fully the power of wireless sensing sys-tems and the data in which they collect.

� Establishing design and deployment guidelines forempirical buildings evaluation.

Clean Technology 2011, www.ct-si.org, ISBN 978-1-4398-8189-7 367

The paper is structured as follows: Section 2 discussesthe need for sensing in the built environment, within thecurrent political climate; Section 3 describes and evalu-ates the authors experience with WSN based assessmentof residential buildings. Section 4 highlights some openresearch and practical issues uncovered during �eld de-ployments that will gain importance in the transition ofWSN based systems from the research to the commer-cial domain. Finally, Section 5 concludes the paper.

2 Why sense the built environment?

The building and construction industry is practisedriven and consequently less prone to re�ection, dissem-ination of good practise and post-factum evaluation ofits �nal products or their components. Evaluation ofa building's performance, post-construction, is not com-monly part of a building's construction and commission-ing cycle. The inherent assumption is that a building,as constructed (and, following on, as occupied) will per-form according to the design speci�cation.

Although the number of studies carried out world-wide to empirically assess the performance of buildingsis dwarfed by the number of buildings constructed, the�ndings overwhelmingly show that, when evaluated �asoccupied� a building's carbon footprint: i) is heavily de-pendent on the occupants (with up to 200% energy con-sumption variations in identical residential buildings [5])and ii) rarely aligns with design expectations of stock ingeneral (usually performing worse) [9].

With reference to the residential sector in particular,the industry's over reliance on predictions from designmodels and estimations (based on nominal rather thanactual performance of materials) thus: i) precludes theaccurate quanti�cation of a building stock's carbon foot-print; ii) precludes evidence-based analysis of new buildsand refurbishments value for money and the assessmentof various technologies viability .

The sheer cost, scale and scope of the current decar-bonisation exercise, its timeliness (one building needs tobe refurbished every minute to meet UK carbon emis-sion targets [7]) imposes rates of change far greater thanthose customary to the construction industry (renownedfor its conservative attitudes [1]). This makes necessaryrapid and accurate feedback on the real performance ofnew housing developments and refurbishments. A ro-bust, evidence-based learning process, enabled by postconstruction and post occupancy monitoring, will: i)allow informed design and refurbishment choices whichmaximise the cost and performance bene�ts of usingnovel low carbon technologies, and ii) set realistic per-formance expectations from the use of both conventionaland innovative building techniques, technologies and ma-terials.

If deployed at large scale, Empirical Environmentaland Energy Performance Evaluations (EEEPE) will: i)

considerably enhance the industry's know-how on e�ec-tive building decarbonisation; ii) enable energy bench-marking of occupied buildings and pro�ling the carbonfootprint of real rather than modelled buildings; iii)identify mismatches between design expectations, build-ing performance and occupant perceptions, thus drivedesign re�nements and support informed techniques andmaterial choices; iv) enable the apportionment of fabric,HVAC systems and occupant behaviour onto a build-ing's energy pro�le, thus drive appropriate measures forenergy reduction; v) support evidence based strategiesfor investment in both new builds and refurbishments.

The need for EEEPE is beginning to be recognised byvarious regulatory bodies (see guidance from the Homesand Communities Agency, UK [4]), although standardsfor this method of evaluation are yet to emerge.

E�ective EEEPE is realised through continuous en-ergy and environmental monitoring of a building overa set period of time (variable depending on the goalsof the EEEPE) and inference of information from thedata gathered, through data mining and / or statisticalapproaches.

Advances in wireless sensor networks (WSNs) allowfor EEEPE to be deployed at relatively low cost perbuilding, with minimal specialist infrastructure costsand minimal disruption to occupants. WSN systemscan be designed to deliver a large variety of data types,but most commonly, EEEPE is concerned with accuratemeasurement of temperature, humidity, light, air qual-ity, occupancy, energy and water consumption. This setof measurands could evaluate a building's performancein a holistic manner.

However, for EEEPE to become a widespread tool,a clear de�nition of informational outputs to be derivedfrom the data is called for. At a practical level, to besuccessful beyond pilots, the need for and value addedby EEEPE will need to be fully understood by the builtenvironment practitioners.

3 Empirical evaluation of residential

buildings�Experiences

Over the past two years, the authors team here (with3 computer sciences researchers specialising in WSN de-sign and one built environment practitioner experiencedin asset management) have worked towards designingbespoke end-to-end, WSN based informational systemsfor EEEPE evaluation of residential buildings. They fur-ther deployed the instrumentation in 15 homes owned byOrbit Heart of England (OHE) Housing Association1,gathering data equivalent to approximately 800 moni-toring days. From the 15 deployments (performed with4 identical sets of instrumentation), i) the �rst 3 (win-

1OHE own around 14000 properties in the Midlands, England,

which are provided as social housing.

Clean Technology 2011, www.ct-si.org, ISBN 978-1-4398-8189-7368

Figure 1: ArchRock Node(left) with integrated CO2Module(right)

ter of 2008/2009) were pilots, aiming to evaluate theperformance of the instruments against their technicalspeci�cation (wireless network connectivity, yield, ro-bustness, data quality and battery life); ii) the following5 (winter of 2009/2010 and summer 2010) were used toiteratively re�ne and validate the systems and the de-ployment processes, de�ne holistic and occupant drivenmetrics, and uncover operational issues (leading to anumber of deployment policies) and iii) the most recent7 deployments (winter 2010/2011) were used to validatethe informational approaches developed and their abilityto generate knowledge when used by built environmentpractitioners; these deployments were well aligned withOHE's operations.

Two WSN system usage scenarios were tested: i)routine winter performance evaluation and energy con-sumption / home �health� benchmarking and ii) inves-tigation into causes for excessive energy consumption insome dwellings (following tenant reports). Both scenar-ios were supported by 2 weeks per home deployments,and the success of the WSN based EEEPE evaluationapproach was measured based on:

1. The ability of the instrument to gather data ofsu�cient quality.

2. The ability of themetrics and associated visualisa-tions to convey knowledge to building specialists.

3. The match between the knowledge inferred fromthe EEEPE and the surveyor's subjective assess-ment .

The instrumentation is built around the ArchRocksystem [8] and comprises of a number of sensing nodes,routers and a server. All nodes (based upon the TelosBplatform [6]) sense temperature and humidity (batterypowered) at 5 minute interval. Some nodes have an in-tegrated CO2 gas sensing module (ELT B-530) to mea-sure the air quality (5 minute sampling) and have beeneither battery powered (early work) or mains powered(later prototypes) (Figure 1). The nodes form a 6LoW-PAN network based upon the IEEE 802.15.4 standard.Electricity usage is recorded via a meter-�tted CurrentCost device [2] which has been interfaced to the serverand records the power consumption every 6 seconds.

A typical deployment would include air quality nodesin the bedrooms and living rooms and a standard ArchRock

Figure 2: Typical Deployment

node in each of the other rooms / de�ned spaces (see Fig-ure 2 for a typical apartment deployment). The largestdeployment was 30 nodes per dwelling and the smallestwas 10. An outdoor node was also deployed every time.

Several house and family archetypes were monitored,which had a variety of heating systems (air source, gascentral heating, electric storage heaters and ground sourceheating).

3.1 Results and Evaluation

From an engineering and computer science viewpoint,the deployments were considered a success given:

� High data yield and excellent radio coverage�Yieldaveraged 95.1% over the whole set of deployments,with a minimum of 81.1% (due to a faulty routercausing a system failure) and a maximum of 99.9%(achieved in 4 homes). 11 deployments were exe-cuted with a server per home, thus distances be-tween neighbouring nodes were relatively short (10mor less). In 4 deployments, however, we success-fully connected 2 neighbouring dwellings to a sin-gle server, maintaining a yield of 99%. This showsthat instrumentation costs for large monitoring ex-ercises (whole streets or geographically compactnew built sites) can be contained by using fewservers and dense networks of nodes.

� Satisfactory robustness of the instrumentation�The hardware (4 identical sets) survived with mi-nor failures due to children playing with the nodesand breaking the connected CO2 module and wa-ter damage on the outdoor nodes.

� Acceptable battery life�The nodes containing CO2sensors had a battery life of approximately 10 days,whilst the rest lasted on average for 9-10 months.To eliminate the need for battery changes within

Clean Technology 2011, www.ct-si.org, ISBN 978-1-4398-8189-7 369

deployments, the CO2 nodes were adapted for mainspower.

� High data quality�Few nodes delivered erroneousdata, and this was always due to low battery lev-els; a range based �lter was used during the pre-processing stage to clean the data.

� Low number of home interventions needed for main-taining the instrumentation�Only 3 deploymentsneeded to be recon�gured.

Despite the excellent performance, the instrumentation,as deployed, had two drawbacks: i) high cost; ii) shortshelf-life, as the Arch Rock components became obsoleteafter the �rst year of use, thus limiting the value of theinstrumentation considerably.

3.1.1 Metrics and Visualisation

As opposed to many other WSN applications reportedin the literature, in this work, designing and successfullydeploying well performing data acquisitions systems isonly a �rst step towards e�ective EEEPE. The follow-ing research question emerged once pilot deploymentswere performed and data was attempted to be analysedjointly (and independently) by our computer science andbuilding specialists team:

What knowledge generation strategies are needed tobring meaning to environmental and energy data? Inother words, a) what are the metrics and processes tobe applied to monitoring data sets in order to bridge thedata to information gap?; b) what knowledge is desiredto be derived from the information?; and c) how shouldthe information be represented to end-users to enabledecision-making?

Given the wealth of data generated by monitoringsystems and the relatively low informational contentper data byte, empirical metrics are needed to con-vert data to information; the metrics should i) be easyto understand by building practitioners, ii) be weatherand energy source independent, iii) be accompanied byeasy to interpret visualisations, iv) integrate well withcommon surveying practise as well as observational andsimulation-based assessment, v) function at several lev-els, from holistic to occupant focused assessments andfrom global evaluation to itemised apportioning of re-sources consumption; vi) enable knowledge generationby the user of the information (designers, contractors,asset managers, etc).

Based on their deployments experience, the authorspropose that:

� Energy/�oor area/degree day is an appropriate holis-tic metric, which allows for benchmarking as wellas absolute evaluations. When coupled with an en-vironmental quality metric (such as Expected Com-fort or Exposure metrics [3]), and when applied to

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.11

0.12

0.13

0.14

0.15

0.16

0 1 2 3 4 5 6 7 8 9 10

kW

h/m

2/D

eg

ree

Da

y

Expected Comfort

Primary Energy Load for Passivhaus:Current Specification as of 2010

Average from all monitored homes

I

J KL

G

ED

F

C

B

A_Feb09

A_Feb10A_Dec10

H

Air Source HeatingElectric Storage Heating

Gas HeatingGround Source Heating

Figure 3: Comparing homes using energy and comfort

a large number of homes, this metric allows for:trends to be inferred; generic conclusions to bedrawn regarding relative performance of speci�ctypes of heating systems and building archetypes;quanti�cation of refurbishment e�ects; and quali-�cation of occupant choices and behaviour impacton the homes' energy consumption. Examples ofapplication for this metric are shown in Figure 3,for 14 homes. Whilst the number of deploymentsis yet too low to allow all the assessments above,note that for example, both properties featuringstorage heaters have higher consumption than av-erage consumers and exhibit low comfort levels.Home B is displaying high levels of comfort yetthere is an extortionate amount of energy beingused. It was found that the occupants left theheating on at all times and had the thermostat sethigh.

� Holistic pro�ling of a home should contain statis-tical information on Temperature, Humidity, AirQuality, Comfort and Energy; a �healthy, e�cienthome� template can be created as a baseline forevaluations, or, for new zero carbon buildings, the�passive house� design speci�cation template canbe used. Examples of holistic pro�ling are givenin Figure 4. This easy to interpret visualisationnot only allows for the �health� of a home to beassessed but also highlights the home's departurefrom ideal behaviour and points out its causes.Home B is very warm but is not perfectly comfort-able. At the same time, it consumes considerableenergy. House A_Dec10 on the other hand is inmuch better health, average temperature/humidityis as to be expected, carbon dioxide levels are low.The �nal spidergram shows expected values for aPassivehaus home. The home is comfortable interms of temperature, humidity and air qualityand the home is using a very low amount of energy.

The metrics and visualizations above have been de-rived iteratively through consultation with surveyors. In

Clean Technology 2011, www.ct-si.org, ISBN 978-1-4398-8189-7370

Figure 4: Spidergrams

AllRooms

Main Bedroom & Livingroom

Livingroom MainBedroom

10

12

14

16

18

20

22

24

Tem

pera

ture

Figure 5: Multiple sensing point vs reduced sensing

their current form, they were deemed as �t for purpose.A further evaluation of the correctness of the informa-tion delivered by the metrics and their robustness wascarried out. The study showed that:

� Deployment density can a�ect the information ac-curacy. Analysis of 6 deployments highlighted, forexample, the need for sensors to be �tted in allrooms to enable accurate pro�ling. Figure 5 illus-trates how measuring temperature in only some ofthe rooms would yield a poor estimate of the dis-tribution of temperatures throughout the house.

� A two weeks monitoring period was found to besu�cient for holistic assessment and benchmark-ing. For new builds, a two year monitoring pe-riod is recommended to allow for both the build-ing and the occupier to achieve stable performanceand bedding-in respectively.

� The Energy / unit area / degree day and Com-fort metrics appear to be robust with regard todeployment duration.

3.1.2 Matching surveyor subjective assessment

This evaluation exercise consisted in comparing and con-trasting EEEPE reports on 4 properties as produced byquali�ed surveyors on the one hand and the authors on

the other hand. The team's task was to provide evi-dence and con�rm whether the heating systems in theselected homes were functioning as speci�ed; if not, toidentify the likely cause. The �ndings were delivered in-dependently by the 2 teams, in a face to face meeting.The surveyors used observation, experience and predic-tive tools to assess and diagnose the homes. The authorsused the metrics and visualizations above, together within-house algorithms for heating energy apportionment.The two teams arrived at similar conclusions regardingthe homes performance but the authors had the advan-tage of being able to rank the properties and quantifythe energy wasted through heating system malfunction-ing.

4 EEEPE beyond pilots�Generic

�ndings and Open challenges

Besides the results obtained and evaluated in the pre-vious section, insight was also gained by the team intoa number of operational and deployment issues. Theseissues only became apparent once the transition of themonitoring instrumentation was made from the researchdomain (pilots) to its operational use within OHE. Itwas found that:

� A good understanding of practical and proceduralissues is needed when deploying systems in occu-pied homes. As an outcome, stock owner proce-dures were developed for i) gaining access to prop-erties, ii) communicating the purpose of measure-ment to occupants, iii) ensuring the systems arenot tampered with and, iv) providing technical in-ductions to tenant liaison teams. It has becomeclear that �buy-in� for assessment needs to be se-cured from both tenants and tenant liaison sta�,with the latter playing a critical role in EEEPEsuccess.

� Deployment and remedial protocols needed to beestablished to minimise occupant disturbance. A

Clean Technology 2011, www.ct-si.org, ISBN 978-1-4398-8189-7 371

deployment of 12 nodes and associated server nowtakes approximately one hour, which includes �ll-ing in bespoke proformas to gather relevant basicoccupant and home information, audit meters andbrief tenants on how to identify and report anysystem malfunction.

� The development of a well organized deploymentdatabase, was required to log observations, trackdeployments and allow the authors team to inter-act and plan remotely.

The work has also highlighted a variety of open chal-lenges. They are listed below (organized into two cate-gories), for the bene�t of researchers and practitionerswho intend to engage with the study and applicationof EEEPE. These questions (and probably more) willneed to be fully answered before the built environmentindustry will fully adopt WSNs as empirical evaluationtools.

4.1 System design challenges

� Who should be deploying and exploiting the WSNbased systems: the computer scientists or the end-user?

� How does one design EEEPE systems, holistically,to respond to the diverse end-user base of contrac-tors, surveyors, political bodies, occupants, build-ing stock owners, etc?

� How low is the �low instrumentation cost� aimedat by the built environment industry to fully adoptEEEPE and how can systems be delivered withinthe given cost bracket?

� What level of hardware/software integration andautomation is necessary to allow delivery to theend-user of the desired informational content inan acceptable form?

� What is the right balance between instrument �ex-ibility (to allow diverse evaluation types), perfor-mance, battery life, integration level and cost?

4.2 Deployment and operational issues

� What are the operational costs associated with de-ployment of EEEPE and how could these costs becontained?

� What are the most e�ective approaches to ten-ant liaison in order to ensure tenant acceptanceof monitoring systems?

� What are the direct bene�ts to tenants followingEEEPE assessment and how should these be ex-plained?

� What are the perceived barriers for EEEPE to be-come operational tools for classically trained, ob-servation driven surveyors?

5 Conclusions

The authors proposed and con�rmed the hypothesisthat WSNs: i) allow the built environment industry toreduce reliance on subjective surveyor's interpretationof a home environment or predictive energy consump-tion and performance models, ii) o�er added value tocompared with traditional evaluations and iii) enablea thorough analysis of the home's environment and en-ergy consumption with little disruption to the occupant.The results of a sustained interdisciplinary research ef-fort were presented, consisting on a validated approachto EEPE and associated instrumentation. Furthermore,remaining open challenges were listed, looking forwardto the adoption of WSNs in the built environment.

REFERENCES

[1] Wren & Bell. Comely green �nal report. Technicalreport, Wren & Bell, 2001.

[2] CurrentCost. Currentcost envi. available from http:

//www.currentcost.com/product-envi.html,2011.

[3] Elena I. Gaura, James Brusey, Ross Wilkins, andJohn Barnham. Inferring knowledge from buildingmonitoring systems: The case for wireless sensing inresidential buildings. In Proceedings of Clean Tech-nology 2011, 2011.

[4] Homes and Communities Agency. Monitoring guidefor carbon emissions energy and water use. avail-able from http://www.homesandcommunities.co.

uk/carbon_challenge, May 2010.[5] Erik Lundstöm. Occupant in�uence on energy con-

sumption in single-family dewllings. Technical re-port, Swedish Concil for Bulding Research, 1986.

[6] Memsic. Telosb datasheet. available from http:

//www.memsic.com/support/documentation/

wireless-sensor-networks/category/

7-datasheets.html?download=152%3Atelosb,2010.

[7] The Chartered Institution of Building Ser-vices Engineers. One building a minute �the great refurbishment challenge. avail-able from http://www.cibsetraining.co.uk/

images/National_Conference_2010/website%

20national%20conference.pdf, 2011.[8] Arch Rock. Arch rock o�cial site. available from

http://www.archrock.com/, 2009.[9] Jürgen Schnieders. Cepheus�measurement results

from more than 100 dwelling units in passive houses.In European Council for an Energy E�cient Econ-omy: Summer Study 2003, 2003.

Clean Technology 2011, www.ct-si.org, ISBN 978-1-4398-8189-7372