Virtual sensors for estimation of energy consumption and thermal comfort in buildings with underfloor heating

Advanced Engineering Informatics 25 (2011) 688–698

Contents lists available at ScienceDirect

Advanced Engineering Informatics journal homepage: www.elsevier.com/locate/aei

Virtual sensors for estimation of energy consumption and thermal comfort in buildings with underfloor heating Joern Ploennigs a,⇑, Ammar Ahmed b,1, Burkhard Hensel a,2, Paul Stack b,1, Karsten Menzel b,3 a b

Dresden University of Technology, Institute of Applied Computer Science, D-01062 Dresden, Germany University College Cork, Department of Civil and Environmental Engineering, College Road, Ireland

a r t i c l e

i n f o

Article history: Available online 17 August 2011 Keywords: Energy efficiency Building performance analysis Virtual sensors Hybrid HVAC systems

a b s t r a c t Evaluating a building’s performance usually requires a high number of sensors especially if individual rooms are analyzed. This paper introduces a simple and scalable model-based virtual sensor that allows analysis of a buildings’ heat consumption down to room level using mainly simple temperature sensors. The approach is demonstrated with different sensor models for a case study of a building that contains a hybrid HVAC system and uses fossil and renewable energy-sources. The results show that, even with simple sensor models, reasonable estimations of rooms’ heat consumption are possible and that rooms with high heat consumption are identified. Further, the paper illustrates how virtual sensors for thermal comfort can support the decision making to identify the best ways to optimize building system efficiency while reducing the building monitoring cost. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction Buildings contribute to 42% of western countries energy consumption [1,2] and it is important to reduce their energy consumption and their related carbon foot-print. A building’s energy-efficiency depends on two aspects [3]. First, the design of a building defines its minimal reachable carbon foot-print. It can be improved by design changes such as adding insulation or making use of renewable energy sources. Second, it is the building’s operation and maintenance that defines how this potential is utilized. Several studies demonstrated that the energy consumption of identical houses may vary by more than a factor of two depending on the occupants’ behavior and the buildings’ operation [3,4]. Optimal building operation requires identifying the best tradeoff between its operating costs, occupant comfort and its energyefficiency [5,6]. Most important for a company managing a building are the operational costs such as costs for heating, cooling, hot water, electricity, and maintenance. Next comes occupant comfort, which needs to be provided at a sufficient level. Least important is often the building’s energy efficiency that targets a minimal energy consumption and carbon foot-print. These three objectives are not necessarily related, i.e. it might be cheaper to heat a building over night ⇑ Corresponding author. Tel.: +49 351 463 38066; fax: +49 351 463 38460. E-mail addresses: [email protected] (J. Ploennigs), a.ahmed@ student.ucc.ie (A. Ahmed), [email protected] (P. Stack), [email protected] (K. Menzel). 1 Tel.: +353 21 420 5453; fax: +353 21 427 6648. 2 Tel.: +49 351 463 38376; fax: +49 351 463 38460. 3 Tel.: +353 21 490 2523; fax: +353 21 427 6648. 1474-0346/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.aei.2011.07.004

at low energy prices to avoid heating during the day, even if demand-oriented heating during the day would involve less heat transfer losses and provide more constant comfort to users. The first step in optimizing a building’s operational consumption is a continuous monitoring of the building’s consumption and conditions. This provides the necessary data for building performance analysis, continuous commissioning, optimizing the operation and users guidance [5,7,8]. The large influences of occupant behavior and room usage show the need to analyze a building’s energy consumption down to consumers at room level, in order to identify energy leaks and optimize the building’s performance. New functional buildings are usually fitted with large building automation systems (BAS) that contain several hundreds of devices controlled by a central building management system (BMS) [9]. The market is currently undergoing a change especially in central Europe and most construction contracts focus nowadays on retrofitting existing buildings instead of new construction. Wireless sensor networks are ideally fitted for this market as they allow for easy installation without design changes [10]. This benefit in installation costs is often compensated by higher device costs [11]. Analyzing a building’s energy consumption down to room level requires a large amount of measurement equipment, such as sensors and meters, in each single room. Therefore, detailed building performance analysis is often hampered by high monitoring cost, and practitioners often request alternative cost-efficient methods [12,13]. Virtual sensor and actuator approaches can reduce equipment requirements. They use a mathematical model of the process to

689

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

Cooling Circuits P-12

Solar Thermal Array

Gas-fired Boiler

HotWater

Underfloor Heating Pump

HeatPump

M

MeterHP

TIN

Meter FH

M

Underfloor Heating Manifolds

Tout

Water from culvert

to river Fig. 1. High level schematic of the mechanical system for the ERI.

estimate a projected sensor value from other measurements, or convert a virtual control value into another actuator command. Virtual sensor is an established term in process modeling and control for a long time [14,15]. Virtual sensors are usually used when the targeted monitoring or control value is not directly or only expensively measureable (e.g. hostile environments), or only measureable with large delays (e.g. dead-time processes). This paper develops an estimation algorithm for a virtual sensor in Section 3 to analyze a building’s energy consumption down to room level. The algorithm is adjustable in its information intake to different sensing and metering equipment available in a building. Therefore, the algorithm offers a flexible usage with rough estimations using few sensors to detailed computations using a high density sensing deployment. The algorithm can be used for offline data analysis and for online monitoring to estimate the heat flows in rooms. Section 4 compares different algorithm variances and discusses their outcome for a real building. The archived user comfort level is also compared, such that a broad set of performance metrics is provided to support building performance analysis, monitoring, and optimization. The approach is validated using the Environmental Research Institute (ERI) building, located on the campus of the University College Cork, as an existing low-energy building with a hybrid HVAC (heating ventilation air conditioning) system that is introduced in the next section.

2. The Environmental Research Institute building Environmentally-friendly buildings combine concepts for energy-efficient buildings with the usage of renewable energy sources to create the energy needed for heating, cooling or domestic hot water [16]. This enables the buildings to operate energyefficiently and aids in the efforts to preserve a clean environment. In Europe such buildings first need to meet the energy-efficiency standards implementing the European Union’s EPBD (Energy Performance of Buildings Directive) [17]. Additionally, environmentally-friendly buildings utilize renewable energy resources [16,18] to minimize the usage of fossil fuels and related air pollution. Renewable energy can be regenerated by natural conservation processes [19] from wind, sunlight, and geothermal heat. These resources are often not always available, e.g. when there is insufficient sunlight for solar collectors on a cloudy winter day. Therefore, regenerative systems are usually backed-up by conventional units such as gas boilers, which results in hybrid systems.

However, conventional back-up systems also reduce the building’s environmentally-friendliness as they burn fossil fuels. The Environmental Research Institute (ERI) is a three-storey, 4500 m2 low-energy building. The building is used as ‘‘Living Laboratory’’ by the Informatics Research Unit in Sustainable Engineering (IRUSE) of the University College Cork and the Irish strategic research cluster ITOBO [20] to serve as a full-scale test bed for Intelligent Buildings demonstrating building performance concepts [21]. The building is equipped with a wireless sensor network of about 100 devices and a wired Building Management System (BMS) consisting of about 180 sensors and meters that monitor indoor and outdoor conditions. The ERI covers various HVAC requirements for laboratories, clean rooms, cold stores, offices, open offices, and seminar rooms used by multiple research groups from biologists, chemists, engineers, and computer scientists. The building is heated by an underfloor heating system that is primarily supplied by a geothermal heat pump that taps into a water supply fed from a culvert running adjacent to a nearby river. The water is preheated by heat recovered from the cold stores and heat generated by the solar thermal array. The underfloor heating operates at a maximum temperature of 40 °C. A condensing gas boiler is sized to act as a complete back-up system. The hot water is provided by an 84 m2 solar thermal collector array, while the rest of the requirement is provided by the boiler. See Fig. 1 for a schematic of the mechanical system. The heat pump and boiler are separately controlled. Both have timetable dependent setpoints to schedule their activity. They ensure that heating and domestic hot water are provided over office hours.

3. Estimating the room heat consumption 3.1. Assumptions about the room heat consumption and discussion To identify rooms with unusual high heat usage the room’s heat consumption needs to be analyzed. But, instead of installing heat meters in each room to measure the individual heat consumption, the room’s heat intake is estimated from the overall heat consumption of the underfloor heating system by creating a virtual sensor model using knowledge about the room heating controls. In the case of the ERI, each room has an individual temperature closed loop control consisting of a temperature sensor, a controller and one to four on/off-valves operating separate underfloor heating circuits. The temperature values are logged by the BMS and

690

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

exported to a data warehouse for analysis [22]. However, exact information about the control values at the valves is not available in the BMS system. This is not uncommon as often only important measurements at time of deployment are logged, but it is the control values that define the room’s and building’s heat consumption. The closed loop controls of the room’s air temperature are activated at all times in the ERI; but, the central heating system provides heat only at specific times. For comparing rooms’ consumption, a thermal room model is needed. Such models are typically based on conservation of energy [23–26]. That means that the change of the heat Q_ store stored in the room’s air and envelopes equals the difference between the heat flow brought into the room by heating Q_ in , the heat Q_ exch exchanged with neighbor rooms, and the heat which leaves the room through its envelope Q_ out .

Q_ in þ Q_ exch  Q_ out ¼ Q_ store :

ð1Þ

Several heat emitters account for the heat consumption Q_ in such as heating, sun light, persons, and equipment such as computers. The heat losses Q_ out are mainly losses via the room’s envelope such as the walls, windows, ceilings, and floors. The heat exchanged with neighboring rooms Q_ exch depends on the size and heat transfer coefficients of the walls and the temperature difference of both rooms. As the building is naturally ventilated there is also a consistent heat transfer through air exchange. The capacity of storable heat Q_ store in the room’s air, walls, and equipment depends mainly on the material used, especially if high capacitive materials such as Phase Change Materials [27] are used. Due to the high number of influences on the room heat, the detailed computation of the room heat would now require a complex simulation model [28], that models the rooms not only individually, but considers heat exchange with adjacent rooms down to modeling the air flow and the heating system on computational fluid dynamics (CFD) level. Simulation tools which can be used for such an analysis are broadly available and an overview can be found in [29]. Basically, it is possible to create virtual sensor models from such simulations. This is usually done by training Artificial Neural Networks (ANN) to create abstracted predictors from the simulation results [30,31]. Unfortunately, the effort is very high to create the simulation models manually, parameterize and run them in various configurations to train the ANN models. This is not efficient for practice, especially, since much information about the system is often unknown, beginning with the pipe layout to the materials used for walls, floor, pipes, and windows. The situation may change in future with the increasing support of the simulators for the import of available design information from Building Information Models (BIMs) [26,32,33]. BIMs are designed as exchange format for the design information relevant to a building’s life cycle. BIMs are currently gaining momentum as standard exchange format for new buildings, but they are often not available for older buildings [34–36]. As long as BIMs are not yet broadly available, the usage of simulations to create virtual sensor models requires too much effort. Virtual sensor models that are created from data mining [37,38] or self-train from online data are easier to [39,40], but require some reference measures for modeling. The approach followed in this paper is to create virtual sensor models from statistical calculation models. Statistical models are simplified models used in building design for estimating the energy demand or for dimensioning components such as the underfloor heating system [41]. The approach, therefore, is related to approaches for building performance assessment [26,42] with the different goal of creating a virtual sensor model to estimate a room’s heat consumption. The simple virtual sensor model is cre-

ated based on simplification of the building physics using the assumptions: (A1) The heat metered centrally is consumed by the rooms and not lost to other reasons such as leaky pipes. (A2) The rooms consume heat if a positive temperature difference exists between the heating system medium and the room’s air temperature. (A3) The rooms consume heat only if the valves are open and a water flow exists. Based on these assumptions, the room’s heat consumption Q_ in can be estimated in a simplified way, without modeling in detail the whole building and heating system. Various effects are neglected that mainly define the balance between Q_ out and Q_ store . A heating system can never be ideally isolated so that the contained hot water will always loose heat on its way through the pipes. Assumption A1 results, therefore, in incorrect estimations, as heat is also lost to rooms that are passed by supply pipes such as service rooms in the basement. This loss is negligible as supply pipes are usually well isolated nowadays. Assumption A2 leads to an important simplification as it allows neglecting the heat storage in and heat exchange between rooms. Heat storage basically results in a delayed behavior between heat consumption and heat loss that continually occurs. Based on assumption A2, it does not matter when the heat is lost, but when it is consumed by the room. The ERI building is a good example to illustrate this difference, as it is mainly heated over night and uses the storage effect over the day. The heat is produced in early morning hours and stored in the building’s concrete mass. The building will release the stored heat over the day which results in a more slowly degrading room temperature on a cold day. The large time difference between heat consumption at night and the heat loss during the day is now neglected, assuming that the room that consumed the heat also releases it. This has relevance in conjunction with the heat exchange between rooms, which is also neglected focusing again on the heat consumption of a room, not its heat loss. A room that is centrally located in a building without exterior walls gives a good example to point out the difference. As the central room has no exterior walls it has basically no heat loss to the outside, but it still consumes heat that can only be released via heat exchange with the neighboring rooms. Assumption A2 considers this heat exchange indirectly as this results in a higher room temperature of the adjacent rooms, while again storage and time delay effects are neglected. Also, for the same reason, other heat sources, besides the heating, such as persons, sun light, and equipment can be neglected based on assumption A2. If they exist they influence the room temperature and are indirectly considered. Assumption A3 simplifies effects in the heating system such as that the room may still be heated after closing the valves until the water in the pipe has cooled down to room temperature. This, however, is a storage behavior and neglected as discussed for A2. 3.2. Estimating the room heat consumption Based on the assumptions, a room’s heat consumption is estimated, which does not necessarily have to equal the room’s heat loss. Let Q_ FH be defined as the absolute heat flow of the underfloor heating system that is centrally measured by a heat meter connected to the underfloor heating system’s inlet and outlet. This is, for example, the heat meter FH in Fig. 1 in the case of the ERI. The heat flow is computed in the heat meter from the mass flow _ FH in the pipes, the specific heat capacity of the water cw rate m and the temperature difference of the inlet and outlet of the

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

_ FH cw ðT in; FH  T out; FH Þ. The mass flow underfloor circuits as Q_ FH ¼ m _ FH is controlled in the case of the ERI by the central underrate m floor heating pump in Fig. 1. The temperature difference depends on the heat consumption (flow) of the rooms. Let Q_ r be the heat consumption of any room r in the set of rooms R connected to the heating system. Based on A1 the underfloor heating system has no further losses beside in the rooms and the overall heat flow Q_ FH ðkÞ in time step k 2 Nþ 0 is the sum of all room’s heat flows Q_ r ðkÞ and each room takes a percentaged share pr ¼ Q_ r =Q_ FH in the overall consumption, thus

_ FH ðkÞcw ðT in; FH ðkÞ  T out; FH ðkÞÞ  Q_ FH ðkÞ ¼ m ¼

X

X

Q_ r ðkÞ

r2R

pr ðkÞQ_ FH ðkÞ:

ð2Þ

r2R

Due to assumptions A3, a room’s heat consumption depends on the valve opening 0 6 v r 6 100%. This is insofar correct, as the valve _ r in the underflow ciropening limits the individual mass flows m cuits of each room. Hence, if the valves in a room are closed (v r = 0%) then mass flow and heat consumption will be zero, as _ r cw ðT in; r  T out; r Þ. If the valves are open (v r > 0%) and a flow Q_ r ¼ m _ r > 0 exists, then the room may consume heat. The issue is that the m _ r and the individual temperatures T in; r ; T out; r of precise mass flow m each room are unknown. As a result, the share pr of a room in the building’s heat consumption is not computable via the first part of Eq. (2). Therefore, it is assumed that for each room a relative heating coefficient4 C r can be defined, which can be interpreted as an estimation of the room’s heating capacity in relation to other rooms for the case that the valves in all rooms are open. For example, this can be the sizes of the room’s underfloor heating circuits. Let the actual heating coefficient C FH ðkÞ of a building be the sum of all room’s coefficients linearly depending on their valve opening v r ðkÞ so that

C FH ðkÞ ¼

X

v r ðkÞ  C r ðkÞ

ð3Þ

r2R

C FH ðkÞ can then be interpreted as a measurement of the building’s actual active heating capacity. For example, if C r represents the room’s underfloor heating circuit’s size, then C FH ðkÞ is the current size of active circuits in time step k. As C r is defined as the room’s heating capacity in relation to other rooms, Eq. (3) also allows estimation of the percentaged share pr of a room in the overall heat flow from Eq. (2) so that

v r ðkÞ  C r ðkÞ Q_ r ðkÞ ¼ pr ðkÞ  : _ C FH ðkÞ Q FH ðkÞ

ð4Þ

This equation allows the estimation of the room heat flow from the relative heating capacity of a room. For example, this will mean that a room utilizes the same proportion of the current heat flow as its heating circuit size is in relation to the building’s active circuits C FH ðkÞ in the abovementioned example of using the underfloor heating circuit size. In addition, it easily scales to different levels of available knowledge by modeling v r ðkÞ and C r ðkÞ with available knowledge as demonstrated in the next sections. Summarizing, a building’s room heating consumption can be estimated in each time step k 2 Nþ 0 by the following algorithm: (1) Compute the valve opening v r ðkÞ for each room r 2 R based 4 The name relative heating coefficient was chosen to emphasize that it is a coefficient with a relative association. It is related to the room’s heating capacity, but does not have to represent its real value. It will later on be defined by different measurements and different units, which is not an issue as it is used as coefficient. It is named relative because it only has a valid meaning in relation to the other room’s coefficients.

691

on the control value ur ðkÞ. (2) Compute the building’s actual heating coefficient C FH ðkÞ with Eq. (3) from v r ðkÞ. (3) Estimate the room’s heat flow Q_ r ðkÞ for each room r 2 R via Eq. (4) from v r ðkÞ, and C FH ðkÞ. 3.3. Restore the control value The valve opening v r depends on the control values ur computed by the room temperature controller. If a detailed model is necessary then the specific characteristic curves of the valves can be used. In the case of the ERI, simple on/off-valves are used and it can be simplified that v r ¼ ur 2 f0; 1g. The proposed algorithm uses the control value ur for each room’s individual heating control. As information about the control value is often not available, like in the investigated ERI case, the control value ur can also be computed by remodeling the control algorithm. In the case of the ERI, the valves are individually controlled for each room by a bounce control with a hysteresis hr ðkÞ of 0.3 K around a temperature set-point wr ðkÞ. This means, that if the temperature T r ðkÞ in a room drops below ðwr ðkÞ  hr ðkÞÞ the valves are opened ðv r ¼ ur ¼ 1Þ and the room is heated until the room temperature reaches ðwr ðkÞ þ hr ðkÞÞ at which point the valves are closed ðv r ¼ ur ¼ 0Þ. Thus,

8 1; T r ðkÞ 6 wr ðkÞ  hr k; > > < ur ðkÞ ¼ 0; T r ðkÞ P wr ðkÞ þ hr k; > > : ur ðk  1Þ; otherwise:

ð5Þ

If more advanced PID controllers (Proportional, Integral, Derivative) are used, the control value can be recomputed in a comparable way if the controller parameters are known. If the set-point wr ðkÞ is not known, it can be partly reconstructed from the step response of the room temperature [43] if the quality of control is good enough. 3.4. Computing the relative heating coefficient With a detailed model of a building’s heating system, the individual consumption of one room can be computed quite precisely. The main reason for defining a relative heating coefficient C r and the rather simple relative equation (4) is that such detailed models require a lot of specific information about the heating system and building beginning with the layout and types for pipes, pumps, and valves, to the materials used for walls, floor, and windows. Unfortunately, in practice such information gets often lost after a few years of operation or can only be extracted laboriously from outdated documentation. BIMs are opening new perspectives in this regard, but are not yet available for all buildings as discussed above. The relative heating coefficient C r can therefore be individually adapted to the available information and allows many options for simplifications. The next subsections introduce several coefficients that will be later compared in Section 4 for the ERI building. 3.4.1. Relative heating coefficient based on the room size A simple estimation is to use the room size Ar as an indicator of the relative heating capacity. The estimation is based on Newton’s law of cooling, which states that the convective heat transfer Q_ RS;r is computed for any pair of heat exchanging systems from: their temperature difference, the size Ar of the shared surface, and the heat transfer coefficient hr of the surface. Assuming that each room’s floor is used completely for heating, the convective heat transfer computes to

Q_ RS;r ðkÞ ¼ hr Ar ðT W; r ðkÞ  T r ðkÞÞ

ð6Þ

692

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

with a mean water temperature T W;r ðkÞ ¼ 1=2ðT in; FH ðkÞ þT out; FH ðkÞÞ. This is an idealized model that neglects many effects of a floor heating system [44] and the computed heat transfer Q_ RS;r ðkÞ may not be equal to the actual heat transfer Q_ r ðkÞ. Nonetheless, it should provide an estimation of the relative heating coefficient C r ðkÞ. The relative heating coefficient is defined for the case that the valves in all rooms are open, such that v r ðkÞ ¼ 1 for all rooms. Then, Eq. (4) would only hold if the relative heating coefficient is proportional to the computed heat transfer, i.e.

C RS;r ðkÞ  Q_ RS;r ðkÞ:

ð7Þ

The heat transfer coefficient hr of an underfloor heating system depends on its configuration, the pipe radii, materials and floor materials. It may be provided by the system integrator or can be computed or measured for an underfloor heating system using standards such as EN 1264 [41]. But this requires already detailed information that is not always available. However, if it is assumed that all rooms in the building have the same heating system configuration with identical heat transfer coefficient hr , then hr can be factored out in Eq. (3) and eliminated in Eq. (4), such that it also can be removed from C RS;r ðkÞ without loss of generality as C RS;r only defines a relative coefficient in comparison to other rooms def

C RS;r ðkÞ ¼ hr Ar ðT W; r ðkÞ  T r ðkÞÞ; or; if hr is identical for all rooms;

ð8Þ

def

C RS;r ðkÞ ¼ Ar ðT W; r ðkÞ  T r ðkÞÞ: Beside the room size Ar , other indicators also may be used to define (more precisely) the active shared surface size such as the valve number, the heating system size or the flow rate as discussed in the next subsections. 3.4.2. Relative heating coefficient based on the valve number Using the room size to define the surface used for heat transfer assumes that the underfloor heating system uses relatively the same amount of surface in each room. But, an underfloor heating system does not always use all the surface of a room’s floor and if radiators are used instead of an underfloor heating system then their size has not necessarily a direct relation to the room size. The number of valves VNr per room might then provide a better estimation. Further, it is usually easily countable for existing buildings. Using this coefficient assumes that the heating system size and flow rate is about the same for each valve. def

C VN;r ðkÞ ¼ VN r ðT W; r ðkÞ  T r ðkÞÞ:

ð9Þ

3.4.3. Relative heating coefficient based on the heating system size The actual size SAr of the heating system should provide a more precise estimation than the valve number or room size. The size of the heating system now can be variously measured: (1) by the active surface size of the heating system like a radiator; (2) by the contained water volume in liter (assuming identical pipe radii); or (3) the tube length in meter (same assumption). def

C S;r ðkÞ ¼ SAr ðT W; r ðkÞ  T r ðkÞÞ

the pipe system and the room. It is computed from the inlet and return temperature in the pipes as well as the air temperature of a room. Based upon the EN 1264, the heat consumption of an underfloor heating system can then be estimated to

T in; r ðkÞ  T out; r ðkÞ : Q_ r ðkÞ ¼ Ar U r T LMTD ðkÞ ¼ Ar U r T in; r ðkÞT r ðkÞ ln T out; r ðkÞT r ðkÞ

We assumed that the inlet and return temperature of the room’s are equal to the temperatures measured centrally ðT in; r ðkÞ  T in; FH ðkÞ; T out; r ðkÞ  T out; FH ðkÞÞ. The overall heat transfer coefficient U r is defined by the system properties of the heating pipes and the floor structure for various systems in the EN 1264 [41]. If the building contains the same heating system configuration in all rooms, the overall heat transfer coefficient U r can be neglected in the computation of the relative heating coefficient as in Eq. (8), such that def

C D;r ðkÞ ¼ Ar U r

T in;

FH ðkÞT out; FH ðkÞ T FH ðkÞT r ðkÞ

lnT in;

;

out; FH ðkÞT r ðkÞ

or; if U r is identical for all rooms; def

C D;r ðkÞ ¼ Ar

T in;

FH ðkÞT out; FH ðkÞ T in; FH ðkÞT r ðkÞ

lnT

ð12Þ

:

out; FH ðkÞT r ðkÞ

3.4.5. Relative heating coefficient based on the commissioned flow rate The previous relative heating coefficients assumed that the heating system was only based on the size of the heating system and neglected the differences in the water flow of the pipes and, therefore, also the provided heat. It results from the fact that the flows in each room are usually not monitored and simulations are usually too complex [44]. However, the water flow rates of an underfloor heating system are usually measured once during commissioning (e.g. [45]). If this information is still available then it can be used to estimate rooms’ real flow rates and heat consumption quite precisely. Basically, a room’s heat consumption can be computed from its _ r as inlet and outlet temperature T in; r , and T out; r and its flow rate m done for the central system in Eq. (2)

_ r ðkÞcw ðT in; r ðkÞ  T out; r ðkÞÞ: Q_ r ðkÞ ¼ m

ð13Þ

An issue is that the flow rate and inlet and outlet temperatures for each room are unknown. The inlet and outlet temperatures vary for each room and decrease logarithmically in the direction of flow depending on the flow rate and the heat losses to the room. The flow rate further depends on many influences like pump pressure, pipe architecture and which valves are open. Previously, all these effects were neglected assuming equal flow rate and water temperature for all rooms. However, if the room’s flow rates can be estimated then a more precise model is possible.

ð10Þ

3.4.4. Relative heating coefficient based on the design size The EN 1264 [41] defines some models to estimate the heating performance of an underfloor heating system which are used to design the system’s capacity. The physical behavior of an underfloor heating system is not exactly as simple as assumed by Eq. (6) due to logarithmic temperature gradients in pipes. The EN 1264 defines a more complex model commonly used for heat exchangers. It is based upon the log mean temperature difference T LMTD between

ð11Þ

Fig. 2. Schematic of the underfloor heating system.

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

Fig. 2 shows the schematic of the assumed parallel underfloor heating system. It contains a central system with a pump from which the heat pipes of all rooms split in parallel. As the pipe system contains an incompressible medium and neither leaks nor stores with differential behavior, the mass flow in the heating system needs to be consistent such that

_ FH ðkÞ ¼ m

X

_ r ðkÞ: m

ð14Þ

r2R

PFH ðkÞ PFH ðkÞ R ðkÞ _ FH ðkÞ FH : ¼ v r ðkÞm ¼ v r ðkÞ Rr ðkÞ RH;r RH;r

ð15Þ

_ C;r is assumed to be measured for the case The reference flow rate m that the pump provides a fixed pressure PC;FH , the valve of the room measured is fully open ðv r ¼ 1; RV;r ¼ 1Þ and all other valves are closed. Then the hydraulic resistance RH;r of each room is computed as

RH;r ¼

Finally, the room’s heating coefficient is defined equal to the room’s heat consumption estimated in Eq. (13) and estimates to

  Ar U r def _ r ðkÞcw ðT in; FH ðkÞ  T r ðkÞÞ 1  ecw m_ r ðkÞ : C F;r ðkÞ ¼ m

ð20Þ

4. Case study for the Environmental Research Institute 4.1. Comparison of the relative heating coefficients

_ r ðkÞ varies at each time step depending on Each room’s flow rate m the open valves and the resulting pressure. Assuming that a refer_ C;r is known for each room from design or commisence flow rate m sioning (estimations, simulations or measurements) and that it is measured for ideal cases with the room’s valve fully open and all other valves are open or closed, it then allows the estimation of a room’s flow rate during operation. Therefore, the parallel system in Fig. 2 is analyzed using the electric circuit analogy to Hagen–Poiseuille’s law of fluid flow in a cylindrical pipe. The flow rate is computed in the analogy from the pressure provided by the central pump PFH ðkÞ and the hydraulic _ FH ðkÞ ¼ PFH ðkÞ=RFH ðkÞ. resistance RFH ðkÞ of the whole system to m The individual hydraulic resistance Rr ðkÞ of each room depends on two aspects: (1) the hydraulic resistance RH;r of the room’s pipe system which is assumed to be time-independent and constant; (2) this hydraulic resistance is increased with the valve opening v r ðkÞ linear proportional to the flow rate such that Rr ðkÞ ¼ RH;r =v r ðkÞ. As the system is parallel, the same pressure P FH ðkÞ applies to each room and its flow rate is computed as

_ r ðkÞ ¼ m

693

PC;FH : _ C;r m

The introduced approach was applied to the ERI building and measurements from 3 years were evaluated, including the heat meter HP at the heating pump, the heat meter FH of the underfloor heating and temperature sensors in 58 rooms. The control values were computed from these temperature readings with Eq. (5) using the fixed set-points defined in the BMS which is 20 °C for most rooms. Based on the measurements, the underfloor heating has an absolute heat flow of 95.3 MWh for the 3 years. The geothermal heat pump provides 66% of the heat (63.5 MWh) and the remaining 33% are backed-up by the gas boiler. Fig. 3 illustrates the mean weekly and monthly heat flows of the heat pump meter HP and underfloor heat meter FH. It is visible that both systems are active mainly in the early morning hours of the cold month. Using the algorithm at the end of Section 3.2, the heat flow of the individual rooms was estimated. The different introduced relative heating coefficients C r were computed for comparison. For a better comparison, the relative heat consumption QRr ¼ RQ_ r =Ar per square meter was computed that is defined as the sum of the room’s heat consumption Q_ r over the investigated three years divided by the room’s size Ar . Table 1 compares the results of the different relative heating coefficients C RS;r , C VN;r , C S;r , and C D;r towards the commissioned flow rate coefficient C F;r assuming that this is probably the best estimation as it uses the most complex model. The comparison uses different quality measures to evaluate the different aspects that are relevant for analyzing building’s heat consumption:

ð16Þ

As the room resistance is assumed to be time-invariant, the hydraulic resistance of the building in operation can then be computed for the given parallel system to

X v r ðkÞ X 1 1 ðkÞ ¼ ¼ RFH ðkÞ r2R Rr RH;r r2R

ð17Þ

such that each room’s flow rate can be estimated via Eq. (15). The next unknown variables in Eq. (13) are the inlet and outlet temperatures of each room. From assumption A1, it follows that the central system and the pipes leading to or from the room have no heat loss such that the inlet temperature T in; r ðkÞ equals the central inlet temperature T in; FH ðkÞ. In contrast, the central return temperature of the system T out; FH ðkÞ is the weighted mean of the return temperature per room T out; r ðkÞ weighted by its mass flow such that

T in; FH ðkÞ ¼ T in; r ðkÞ and T out; FH ðkÞ ¼

X 1 _ ðkÞT out; r ðkÞ: m _ FH ðkÞ r2R r m ð18Þ

The return temperature of a room is estimated using the model of the underfloor heating system standardized in the EN 1264 [41]. Setting Eqs. (13) and (11) equal allows solving the equation to the return temperature, which results in r Ur c Am _ ðkÞ

T out; r ðkÞ ¼ T r ðkÞ þ ðT in; FH ðkÞ  T r ðkÞÞe

w r

:

ð19Þ

(1) The mean over all samples of the Euclidian error of the heat consumption per sample (15 min). It provides a measurement of the estimation quality of the heat consumption Q_ r ðkÞ and should be as low as possible. (2) The mean error of the estimated relative heat consumption QRr per room provides a value of how much the estimated relative heat consumption differs in mean per room. (3) To identify rooms in a building with abnormal high heat consumption, the correct ranking of the rooms sorted by their relative heat consumption is probably more important than its specific value. The mean error of the ranking of the rooms provides a measurement how much both rankings differ. (4) The top 20% of rooms with the highest energy consumption are particularly interesting for energy analysis. Therefore, it is also analyzed how many of these rooms, that were identified using C F;r , are also listed in the top 20% returned from the other coefficients. The rank of room 1.09, which is identified as biggest consumer using C F;r is also provided for the other coefficients for comparison. The first conclusion drawn from the comparison is that all coefficients have comparable errors to the commissioned flow rate coefficient C F;r . The mean error of the relative heat consumption is for all other coefficients ðC RS;r ; C VN;r ; C S;r ; C D;r Þ about 5 kWh/m2. This is about 20% of the average room’s consumption. This rather large mean error results from the large model differences from the commissioned flow rate coefficient. However, the ranking of

694

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

Fig. 3. Weekly and monthly mean values of the underfloor heating meter FH and heating pump meter HP.

Table 1 Comparison of the relative heat coefficients to the flow rate.

Valve number ðC VN;r Þ Room size ðC RS;r Þ System size ðC S;r Þ Design size ðC D;r Þ Flow rate ðC F;r Þ

Eucl. err. Q_ r ðkÞ in kWh/smp

Abs. err. QRr in kWh/year

Abs. err. rankðQRr Þ in ranks

Top 20% overlap (%)

Rank of room 1.09

9.12E02 4.15E02 3.76E02 5.02E02 0

5.25 5.38 5.16 5.01 0

4.34 4.20 3.83 3.61 0

73 73 73 73 100

1 2 4 2 1

the rooms differ in mean by about 4 ranks in comparison to the C F;r ranking, which is acceptable. 73% (8 of 11) of the rooms ranked in the top 20% do overlap with C F;r . Only the rough coefficient C VN;r using the valve number estimated room 1.09 correctly as the biggest consumer. But, the coefficient also has the highest Euclidian error in estimating the room’s heat consumption per time step. Fig. 4 shows in detail the estimated relative heat consumptions QRr for the rooms in the top 20%. It is visible that the valve number coefficient has also the strongest variance in comparison to the commissioned flow rate coefficient (4.3E02 vs. ca. 2.3E02 for the other coefficients). The design size coefficient C D;r has the lowest mean error in the relative heat consumption and rank, which makes it the second best coefficient after the commissioned flow rate coefficient. Nonetheless, the differences between the four simple coefficients C RS;r , C VN;r , C S;r , and C D;r are small and will depend mostly on the quality of the available information. It is also possible to combine several relative heating coefficients using a weighted sum of the normalized coefficients. The coefficients best suited for a combination are the valve number with one of the room size, system size, or design size coefficients. Reason is that the commissioned flow rate coefficient computes from the flow rate and the floor heating area and the combination will join both aspects. For example, the weighted sum of the room size and the number of valves computes to

C comb;r ðkÞ ¼ k

C RS;r ðKÞ C VN;r ðkÞ þ ð1  kÞ : C RS;FH ðkÞ C VN;FH ðkÞ

ð21Þ

The parameter k is the weighting factor, expressing the trustworthiness of the room size influence in comparison to the valve number influence. The evaluation of this combination for the ERI case study (k = 0.5) resulted in a slightly improvement of the room size and valve number coefficient with comparable quality results to the system size coefficient. The system size coefficient and the designs size coefficient could not be improved by a combination with the valve number. However, in practice there remains the problem how to choose the weighting factor k, resulting in confusion of the energy consultant who wants to apply automatically the methods of this paper without understanding all details. 4.2. Analysis of the rooms’ heat consumption Fig. 5 shows for all rooms the estimated heat consumption based on the commissioned flow rate coefficient. The rooms were sorted by their heat consumption to illustrate the resulting Pareto distribution often measured in real buildings due to a few rooms using up the majority of the heat [4]. In the case of the ERI, the 11 rooms shown in dark red/gray are the top 20% consumers and use about 50% of the heat. Fig. 8 illustrates on the left side the

Fig. 4. Comparison of the relative room heat consumption of the top 20% rooms for relative heat coefficients.

695

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

< 20% > 20 % > 40%

109 LG28 LG01 LG30 G06 G30 G02 LG23 LG27 LG29 LG37 G01 127 LG25 128 LG36 LG24 G29 LG26 LG04 G27 G10 G11 LG33 LG05 G03 125 101 104 102 130 G26 107 108 103 LG03 LG35 G25 G08 131 126 122 G05 G07 124 105 133 G04 106 G22 121 G24 123 LG21 G09 G23 LG07 G28

Valve Open in % of the samples

Fig. 5. Comparison of the room heat consumption for all rooms: dark red/gray – top 20% (20% of rooms with the highest heat consumption); orange/gray – top 50% (rooms with more heat consumption than average); light green/gray – lower 50% (less heat consumption than average). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

room number

Fig. 6. Valve opening for all rooms in percent of the samples, i.e. 10% for a room mean that the valves were open in 1 out of 10 samples or in 10% of the time: dark red/gray – valves are open in more than 40% samples; orange/gray – open in more than 20% samples; light green/gray – open in less than 20% samples. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 7. Comparison of the mean PMV thermal comfort value for all rooms and actions drawn in comparison to Fig. 5: dark blue/gray – rooms that are at least cool in more than 66% of the time; blue/gray – rooms that are at least cool in 33% of time; light green/gray – rooms with a neutral thermal comfort in mean. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

location of the rooms in the building using the same color scheme. The 58 rooms with a heating system have in mean a relative heat flow of 24.6 kWh/m2 per year. Generally speaking, the rooms facing south have a lower heat flow, while the rooms on the lower ground floor demand a higher heat flow. This behavior is understandable, as the sun supports heating on the south facing side and on the ground floor heat is emitted to the ground while on the 1st and 2nd floor heating is supported by the rooms below. However, surprisingly the room with highest relative heat consumption of 115 kWh/m2 is room 1.09 on the first floor, which is nearly five times of all room’s average heat consumption. It is the

director’s office and has a higher set point of 21 °C. But, the mean temperature in the room is 19.5 °C and the valves are open in 66% of the samples (see Fig. 6). The reason for this very high heat consumption is probably that the room is free floating over the building entrance area and that it has two exterior walls. Thus, it has a very high surface to emit heat to the outside and it needs to be investigated if insulation can be added. The rooms LG28, LG01, and LG30 on lower ground floor are the other rooms with a high relative heat flow of 102, 90, and 72 kWh/ m2, respectively. The rooms LG28, LG29, and LG30 have exterior doors that are probably the reason for the heat loss and should

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

Ground Floor

LG01

LG02

LG03

LG04

LG05

LG06

LG07 LG08 LG09

LG35 LG13

LG33

LG16

LG17

LG19

LG37

LG21

LG36

LG23

LG24

G01

G02

LG25

G03

LG26

G04

G14

G06

G05

G32

LG27

LG28

G07

G08 G09

G33

G15

G22

G17 G23

LG29

LG30

N

G10

G34 G18 G25

G24

G29

G19 G20 G36 G26

G27

G28

G30

was created from four rooms that possess also radiant temperature and humidity sensors [37,39]. The estimated comfort level permits to identify potential rooms where the temperature set point can be reduced. About 43 of the 71 rooms have a reduced comfort level of either ‘slightly cool’ (1.5 < PMV < 0.5) or ‘cool’ (2.5 < PMV < 1.5) in more than 33% of the samples. The rooms LG03, LG28, LG29, and LG30 have the lowest comfort level with cool comfort in more than 25% of the cases. Reducing their set-points would degrade the comfort further and is not recommendable. Instead, it is the office 1.27 with a mean temperature of 21.8 °C that has the highest potential to optimize its heat consumption of 39 kWh/m2 by reducing its set point from 21 to 20 °C as it provides one of the best comfort levels in the building, while also having a high heat loss. 5. Discussion Table 2 summarizes the required information, the presumptions, and the concluding notes for the introduced and compared relative heating coefficients. In general, the precision of the approaches increases with the information available and used as demonstrated for the example. Thus, it is usually better to use the commissioned flow rate coefficient than the valve number

Lower Ground Floor

Lower Ground Floor

be better isolated. The only reason that LG29 has a lower relative heat flow of 48 kWh/m2 is that it has a very low set point of 16 °C. Also LG30 has a reduced set point of 18 °C, which gives the reason that it consumes 29% less heat than LG28. Its set point of 20 °C should be reduced to save energy as its valves are open in 77% of the samples, the highest value in the building. But, this would reduce the thermal comfort of the people working in the laboratory. The thermal comfort is an important aspect in considering actions to reduce heat losses, as it represents the experience of the human users of the rooms. Reducing the set-points is usually a quick, simple, and cheap approach to reduce a room’s heat usage. But, it also may drastically reduce user’s thermal comfort for a reasonable impact on the heat consumption. Improving a room’s insulation on the other hand is a sustainable solution that is costly, but, influences the user comfort often positively in the long term. This renders the thermal comfort important for decision making about improvements of a building’s heat consumption. Fig. 7 shows the mean thermal comfort for the rooms, and Fig. 8 illustrates the estimated comfort levels on the right side. The thermal comfort Predictive Mean Vote (PMV) as defined in the ISO 7730 [46] was predicted from the room temperature and the outside conditions based on a data mining classification model that

Ground Floor

696

LG01

LG02

LG13

LG23

G01

LG16

LG24

G02

LG25

G03

G14

1.12

1.05

1.03 1.04 1.31 1.15 1.16 1.17

1.13

1.06

1.07

1.09

1.32 1.18

1.20

1.33

1.34 1.24

1.22 1.23

1.21

No heating

Lower 50%

1.25

Top 50 %

1.26

1.27

First Floor

First Floor

1.02 1.30

1.28

Top 20 %

1.01

LG26

G04

G15 G23

1.02

LG04

LG05

LG17

G07

LG29

G08 G09

LG30

N

G10

G19 G20 G25

1.03

LG07 LG08 LG09

LG21

LG28

G18 G24

LG06

LG19

LG27

G06

G17

G22

1.01

LG03

G26

1.05

G27

G28

1.06

G30

1.07

1.09

1.04

1.12

1.15

1.13

1.16 1.17

1.21

Neutral

1.18

1.25

1.23

>33%Cool

1.20

1.33

1.24

1.22

> 66%Cool

1.26

1.27

1.28

Used for modelling

Fig. 8. (Left) Results for the heat consumption at the ERI (see Fig. 5 for color scheme). (Right) User comfort for the ERI (see Fig. 7 for color scheme). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Table 2 Summary of the relative heat coefficients.

a b

Coefficient

Required knowledgea

Presumptionb

Notes

Valve number ðC VN;r Þ

– Number of valves VN r

Heating system size and flow rate is about the same for each valve

Room size ðC RS;r Þ

– Room area or floor heating area Ar – (Heat transfer coefficient hr if it varies)

System size ðC S;r Þ

– Water volume, tube length or radiator surface SAr

Floor is completely used by a floor heating system or heating system is dimensioned proportional to room size Flow rate is about the same for each system

Good for simple assumption if not much is known about the system Use design size instead coefficient ðC D;r Þ

Design size ðC D;r Þ

– Room area or floor heating area Ar – (Overall heat transfer coefficient U r if it varies)

Flow rate ðC F;r Þ

_ C;r – Room area or floor heating – Commissioned flow rate m area Ar – Overall heat transfer coefficient U r – Specific heat capacity of the heating medium cw

Floor is completely used by a floor heating system or heating system is dimensioned proportional to room size Floor is completely used by a floor heating system

_ FH , room temperature T r , and central inlet, outlet temperature T in; FH ; T out; FH . Required knowledge additional to: central heat flow m Additional to A1–A3.

Good if system size is not proportional to the room size Good if system size is proportional to the room size Best if the commissioned flow rate is known

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

coefficient. But, if only the valve number is known for a system, then it will still provide a basic estimation of the heat distribution in the building. The design size coefficient should be preferred over to the room size coefficient as it uses the same information, but considers also the logarithmic temperature gradient and is, therefore, more precise. If the heating system size is not directly related to the room size, then it is usually better to use the system size coefficient. Apart from the specific presumptions defined for each coefficient in Table 2, all coefficients presume that the assumptions A1–A3 are met. This will not be valid for each real world example. For example, A1 assumes that heat is only absorbed by rooms and no heat is lost in the pipes leading to or from them. This allows setting the inlet temperature T in; r equal for all rooms, which is a simplification of the real world. For example, the heating pump is located in LG05 in the ERI such that the room furthest away is 1.21 on the first floor which lies ca. 6 m above and ca. 45 m away from the heating pump. Of course there is a temperature difference between the inlet temperature of room 1.21 and room LG07 just next to the heating pump. It is possible to model such differences by: (1) considering them as part of the relative heating coefficient, i.e. increasing the system size coefficient C S;r by the additional pipe length assuming that the temperature in the rooms passed is about the same. (2) If this is not the case, the model also can be easily extended by individual losses to the rooms passed. Both extensions simply modify the model of the relative heating coefficient with a more precise one, which illustrates the flexibility of the approach. The assumptions A2 and A3 remove the dynamic behavior of heat exchange, which also is strong simplification. But, the aim of the approach is to create a virtual sensor model, that assigns heat flow in a building to the causes, which are open control valves of rooms, and not to model how exactly the heat is lost. Thus, the approach provides not a physical model of the heat distribution in a building, but simple statistical estimations of this cause-and-effect principle. These assumptions are only a limitation of the currently defined relative heating coefficients, and not a limitation of the general approach. More precise coefficients can be defined, based on simulation results or measurements of the system. Particularly, if BIM will be broadly available for buildings in the future, then this process can be automated and the virtual sensor model can be automatically generated. The benefit is in the end, that still only simple temperature sensors are needed in the rooms to understand and assign heat flow in a building.

6. Conclusion The introduced virtual sensors approach for estimation of a room’s heat consumption provides a simple and scalable way to estimate the heating energy of rooms from simple temperature readings and a central heat meter. The approach can be applied to any building’s heating system whose rooms are individually controlled, as long as a representative relative heating coefficient C r can be defined. The various relative heating coefficients also exemplify the flexibility of this approach. Due to the simple relational equation (4) even simple available information such as room size or valve number can be utilized to estimate a room’s heat consumption. The comparison showed that simple coefficients provide reasonable estimations to identify rooms with high heat consumption, but cannot replace more complex models such as the commissioned flow rate coefficient. Using this coefficient and remodeling the control behavior minimizes the required sensor equipment for building performance optimization. As wireless temperature sensors can, nowadays, be cheaply and easily integrated in any building [10], the approach

697

has practical value for performance analysis of existing buildings. The approach can be used both for offline data analysis and online as a virtual sensor to measure rooms’ heat flow. In that case no additional heat flow meters need to be installed in the rooms. Instead, the virtual sensor is realized as software in the centralized BMS or on embedded devices that are responsible for several rooms. Combined with virtual sensors for thermal comfort [39], the temperature sensors serve multiple purposes from controlling the room temperature, to monitoring of the temperature, comfort, and heat flow. The investigated case study of the ERI building showed that the approach permits to identify energy consumption problems in a building. This, in combination with an approach to estimate the thermal comfort in the rooms [37,39], further complements the basis for decision making to improve the building’s energy consumption. Both approaches mainly utilize simple temperature sensors and, therefore, provide simple ways to evaluate buildings system efficiency and rooms’ consumption while reducing the building monitoring cost. 7. Future directions and challenges Tools to evaluate and control energy consumption of buildings will become increasingly important because of limited energy resources and climate change. For their broad application it is needed that they can be used with reliable benefit and low or even no human effort. Their tasks range from detecting construction failures and energy leakages, suggesting and implementing energy optimizations and helping to improve occupants’ comfort. These tools should become standard in every building. Up to now, analyzing buildings’ energy consumption is often a manual expert task based on graphical plots instead of automated computational algorithms. It usually neglects the benefits of energy simulations due to their complexity for non-experts. Easy creation of buildings’ behavioral and simulation models is a precondition for improved acceptance. The growing spread of building information models (BIMs) reduces the effort for simulations significantly. Also, monitoring is usually too expensive. Smarter approaches like virtual sensors will help to make monitoring affordable for all building owners. The approach shown in this paper should be enhanced to become a holistic instrument that includes other heating types, cooling, electricity consumption, lighting, and hot water preparation and integrates into simple, autonomous building’s energy management systems. Such systems are a challenge for the future, particularly as the usage of renewable-energy sources, combined with energy storages (PCM, cars, etc.) as well as a localized dynamic smart energy market will be the driver to much more dynamic in buildings’ energy optimization and control. Acknowledgements Work in the Strategic Research Cluster ‘ITOBO’ is funded by Grant 06-SRC-I1091 from Science Foundation Ireland (SFI) with additional contributions from five industry partners. Joern Ploennigs thanks the Humboldt-Foundation and the German BMBF for supporting his research in Ireland. The authors thank Luke Allan and Ena Tobin, Civil Engineering, UCC for their contribution to this research. References [1] European Commission, Action Plan for Energy Efficiency: Realising the Potential, Communication from the Commission, October 2006. [2] L. Itard, F. Meijer, E. Vrins, H. Hoiting, Building Renovation and Modernisation in Europe: State of the Art Review, January 2008.

698

J. Ploennigs et al. / Advanced Engineering Informatics 25 (2011) 688–698

[3] Mark Levine et al., Residential and commercial buildings, in: Climate Change 2007: Mitigation. Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, 2007, pp. 387–446. [4] T. Ueno, F. Sano, O. Saeki, K. Tsuji, Effectiveness of an energy-consumption information system on energy savings in residential houses based on monitored data, Applied Energy (83) (2006) 166–183. [5] G. Augenbroe, C.-S. Park, Quantification methods of technical building performance, Building Research and Information 33 (2) (2005) 159–172. [6] G. Kats, L. Alevantis, A. Berman, E. Mills, J. Perlman, The costs and financial benefits of green buildings, California’s Sustainable Building Task Force, USA, 2003. [7] F. Wolfgang, Assessing Building Performance, Butterworth–Heinemann, 2005. [8] D. Claridge, J. Haberl, M. Liu, J. Houcek, A. Athar, Can you achieve 150 percent predicted retrofit savings: is it time for recommissioning? in: ACEEE 1994 Summer Study on Energy Efficiency in Buildings, Washington, DC, USA, 1994, pp. 73–88. [9] W. Kastner, G. Neugschwandtner, S. Soucek, H.M. Newman, Communication systems for building automation and control, Proceedings of the IEEE 93 (6) (2005) 1178–1203. [10] K. Menzel, D. Pesch, B. O’Flynn, M. Keane, C. O’Mathuna, Towards a wireless sensor platform for energy efficient building operation, in: Twelfth International Conference on Computing in Civil and Building Engineering, vol. 13, Beijing, China, 2008, pp. 381–386. [11] M. Kintner-Meyer, M. Brambley, T. Bauman, N. Carlon, Wireless sensors: technology and cost-savings for commercial buildings, in: ACEEE Summer Study on Energy Efficiency in Buildings, 2002. [12] L. Jagemar, D. Olsson, The EPBD and Continuous Commissioning, 2007. [13] A. Schumann, M. Burillo, N. Wilson, Predicting the desired thermal comfort conditions for shared offices, in: ICCCBE – 13th International Conference on Computing in Civil and Building Engineering, Nottingham, GB, 2010. [14] M.J. Willis, C. Di Massimo, G.A. Montague, M.T. Tham, A.J. Morris, Artificial neural networks in process engineering, in: IEEE Proceedings on Control Theory and Applications, vol. 138, no. 3, 1991, pp. 256–266. [15] A. Dementjev, H.-D. Ribbecke, K. Kabitzsch, Improvement of prediction performance for data-driven virtual sensors, in: ETFA – 16th IEEE International Conference on Emerging Technologies and Factory Automation, Bilbao, ES, 2010. [16] D. Chwieduk, Towards sustainable-energy buildings, Applied Energy 76 (1–3) (2003) 211–217. [17] C.A. Roulet, B. Anderson, CEN standards for implementing the European directive on energy performance of buildings, in: PLEA – 23rd International Conference on Passive and Low Energy Architecture, Geneva, Switzerland, 2006. [18] A.B. Ngowi, Creating competitive advantage by using environment-friendly building processes, Building and Environment 36 (3) (2001) 291–298. [19] D. Chiras, The Homeowner’s Guide to Renewable Energy: Achieving Energy Independence through Solar, Wind, Biomass and Hydropower, New Society Publishers, 2006. [20] ITOBO, 2007 (online). Available from: . [21] M. Keane, Hidden depths, Building Service Journal (2005) 23–28. Available from: . [22] A. Ahmed, J. Ploennigs, K. Menzel, B. Cahill, Multi-dimensional building performance data management for continuous commissioning, Advanced Engineering Informatics 24 (4) (2010) 466–475. [23] T. Schmid, D. Mehdi, G. Abba, Modeling & control of a thermal system, in: Third IEEE Conference on Control Applications, Glasgow, GB, 1994, pp. 1795–1800. [24] J.M. House, T.F. Smith, Optimal control of building and HVAC systems, in: American Control Conference, Seattle, WS, USA, 1995, pp. 4326–4330. [25] M.-L. Chiang, L.-C. Fu, Adaptive and robust control for nonlinear HVAC system, in: International Conference on Systems, Man, and Cybernetics, Taipei, Taiwan, 2006, pp. 4982–4987.

[26] Arno Schlueter, Frank Thesseling, Building information model based energy/ exergy performance assessment in early design stages, Automation in Construction 18 (2) (2008) 153–163. [27] A.M. Khudhair, M.M. Farid, A review on energy conservation in building applications with thermal storage by latent heat using phase change materials, Energy Conversion and Management 45 (2) (2004) 263–275. [28] C. Struck, P.J.C.J. de Wilde, C.J. Hopfe, J.L.M. Hensen, An investigation of the option space in conceptual building design for advanced building simulation, Advanced Engineering Informatics 23 (2) (2009) 386–395. [29] Drury B. Crawley, Jon W. Hand, Michaël Kummert, Brent T. Griffith, Contrasting the capabilities of building energy performance simulation programs, Building and Environment 43 (4) (2008) 661–673. [30] J. Park, M. Moon, S. Hwang, K. Yeom, CASS: a context-aware simulation system for smart home, in: SERA – Fifth International Conference on Software Engineering Research, Management & Applications, 2007, pp. 461–467. [31] F. Felgner, Design of Virtual Airflow Sensors for Thermal Comfort Control, Shaker Verlag, 2009. [32] V. Bazjanac, Building energy performance simulation as part of interoperable software environments, Building and Environment 39 (8) (2004). [33] Salman Azhar, D. Olsen, W.A. Carlton, I. Ahmad, Building information modeling for sustainable design and LEED rating analysis, Automation in Construction 20 (2) (2011) 217–224. [34] G.V.R. Holness, Building information modeling gaining momentum, ASHRAE Journal 50 (6) (2008) 28–40. [35] R. Howard, B.-C. Björk, Building information modelling – experts’ views on standardisation and industry deployment, Advanced Engineering Informatics 22 (2) (2008) 271–280. [36] W. Shen et al., Systems integration and collaboration in architecture, engineering, construction, and facilities management: a review, Advanced Engineering Informatics 24 (2) (2010) 196–207. [37] A. Ahmed, J. Ploennigs, Y. Gao, K. Menzel, Analyse building performance data for energy-efficient building operation, in: CIB W78 – 26th International Conference on Managing IT in Construction, Istanbul, Turkey, 2009. [38] A. Ahmed, N.E. Korresc, J. Ploennigs, K. Menzel, Mining building performance data for energy-efficient operation, Advance Engineering Informatics 25 (2) (2011) 341–352. [39] J. Ploennigs, B. Hensel, K. Kabitzsch, Wireless, collaborative virtual sensors for thermal comfort, in: BuildSys – Second ACM Workshop On Embedded Sensing Systems For Energy-Efficiency in Buildings and Surroundings, Zurich, Switzerland, 2010, pp. 79–84. [40] J. Ploennigs, A. Ahmed, P. Stack, K. Menzel, Model-based virtual sensors for room heat metering and energy consumption analysis in buildings with renewable energy sources, in: ICCCBE – 13th International Conference on Computing in Civil and Building Engineering, Nottingham, UK, 2010, p. 231ff. [41] EN 1264, Water based surface embedded heating and cooling systems, 2009. [42] I. Jaffal, C. Inarda, C. Ghiaus, Fast method to predict building heating demand based on the design of experiments, Energy and Buildings 41 (6) (2009) 669– 677. [43] V. Vasyutynskyy, J. Ploennigs, K. Kabitzsch, Passive monitoring of control loops in building automation, in: FET 2005 – Sixth IFAC International Conference on Fieldbus Systems and Their Applications, Mexico, 2005, pp. 263–269. [44] P. Weitzmann, S. Svendsen, Comparison of performance of simulation models for floor heating, in: Seventh Symposium on Building Physics in the Nordic Countries, 2005, pp. 372–379. [45] T&G Somers, Water Balancing Commissioning Report for Underfloor Heating in the ERI Building, 2005. [46] ISO 7730, Ergonomics of the thermal environment – analytical determination and interpretation of thermal comfort using calculation of the PMV and PPD indices and local thermal comfort criteria, 2005.