Final Report Summary - TRIBUTE (Take the energy bill back to the promised building performance)
Executive Summary:
The TRIBUTE project ended by end of October after 4 years of research on closing the gap between simulation and reality. 16 partners were contributing to the results of the project, with an overall budget of 10 M€, and 6.7 M€ funding from European commission in the frame of FP7 framework.
Buildings represent more than 40% of EU energy consumption and 36% of EU CO2 emissions. To reach the 2020 20% energy efficiency target, the EU energy efficiency directive sets binding measures to be implemented by EU members by the end of 2016. A new package of measures has recently been proposed to extend the energy efficiency target to 40% for 2030, including proposed revision of European Building Performance Directive. Building Energy Performance Simulation (BEPS) tools are widely used for modern mid-size to large buildings. When used across the building commissioning and operation phases, BEPS models can be powerful tools to assist building operators and facility managers in creating virtual sensors, performing sensitivity analyses, assessing building energy performance, detecting anomalies and identifying potential retrofit solutions.
European Tribute project FP7 has been exploring various BES applications to enable building energy managers to understand BES, and to reduce the gap between predicted and real performances. The pre-requisites for the deployment of analytics are a commissioned building and a calibrated model. The analytics proposed entail a large number of real data and related simulations. To facilitate use by building facility managers, not only does all data management have to be automatic (real data collection and storage, simulation of settings, processing, and real versus simulated mapping); the execution of analytics and display of results must be automatic as well.
TRIBUTE has provided a building monitoring system for bridging the gap between the design-stage simulation model and the physical building.
The approach relies on a calibrated simulation model of the building that would typically be developed in the design phase. Business margins are small in this industry, and the cost of developing a tailored model for the operation phase would often be prohibitive. From a general simulation perspective, such a design model is severely over-parameterized. These Parameters correspond to physical measures of individual building-parts, and therefore the total number of parameters for an individual room can be in the order of hundreds. To manage this over-parameterized model for system identification is a challenge.
The developed system targets the expert operator. It is envisioned that each such operator can manage a large number of buildings from a remote location. Separate user interfaces may in subsequent work also be developed for the day-to-day building maintenance personnel, tenants, and other stakeholders such as owners.
Furthermore, the present work is strictly aiming at the development and presentation of a model that is in real-time kept in tune with the physical building. Further use of this model for Fault Detection and Diagnosis or predictive control is not presently encompassed.
One of the most common building analytics is to use a model built on historical energy performance and compare this model prediction on a new period with the real measurement. After a retrofit, or major change, have been applied to the building, this comparison allows to evaluate the result of the modification, as suggested for example in the International Performance Measure & Validation Protocol (IPMVP). When used in building normal operation (no change assumed), this comparison allows to detect abnormal behaviour in a process named performance follow-up.
This document reports the evaluation of using a continuously calibrated building simulation model. The IDA-ICE model is compared to simpler linear, or multi-linear model typically used for performance follow-up. A simple GUI have been developed to quickly compare the various models on different settings. Note that, at this stage, this GUI is more a research tool than something useful for facility manager.
The software was tested on the Vaucanson office building in La Rochelle. As expected, the building simulation get better results than the other simpler models, both on the learning and testing horizon. This reflects the fact that the simulation model considers a larger number of phenomena thus allowing better identification of real faults.
With error reduced by more than 20% compared to advanced thermal signature, which is itself 10% lower than pure linear model, the simulation can allow to reduce greatly the number of day that will be considered as faulty compare to the learning period.
Of course, simulation based follow-up is much costlier to set than more typical methods, so it will probably never be developed for this single usage. But, if we have the model available for multiple usage (fault detection, advanced control, retrofit option analysis), it will bring a high value at limited cost for this type of analytic.
Project Context and Objectives:
Today, Building Energy Performance Simulation (BEPS) analysis tends to show a large discrepancy with real energy performance. Most cases are due to gross mistakes rather than fundamental inadequacy of available technology and methods. The reasons are manifold. Highly simplified calculation methods are used far beyond their domain of validity. Assumed boundary conditions such as occupant behaviour are not in accordance with actual usage; gross malfunctions in control and HVAC systems are left undetected in the commissioning process, while thermal bridges and distribution system losses are left without attention. Moreover, metered and sub-metered data are not used efficiently in calculation tools and engineering based simulation models during the Measurement and Verification (M&V) phase.
TRIBUTE aims at minimizing the gap between computed and measured energy performances through the improvement of the predictive capability of a commercial BEPS. To achieve this goal, a set of intermediate objectives were set:
• Extend the use of simulation to the commissioning and operation stages of a building
• Allow efficient on-line identification of building key parameters
• Deploy a generous monitoring infrastructure on which the simulation could be trained and benchmarked
• Reduce the number of sensors to allow long term / affordable monitoring of the buildings
• Development of significantly faster and more precise retrofit decision tool that will be offered to European regions and individuals
• Improvement of energy flow management systems taking into account occupancy behaviour and featuring a continuous building health monitoring approach.
The methodology and tools will be evaluated in the context of different buildings types and locations. Within the TRIBUTE project we propose to reduce the gap between planned and actual energy performance down to 5 % (+- 2 %).
TRIBUTE proposes to demonstrate an automated, model-based whole-building performance monitoring system at sites in partnership with Communauté d’agglomération de La Rochelle and Città di Torino.
In these existing buildings, measurement and verification techniques will be developed and deployed.
The system will continuously acquire measurements of HVAC-lighting, plug loads, occupancy usage from existing control system augmented by additional sensors to capture other building parameters measurements. The system will connect to Building Health Monitoring and Energy Flow Management applications (adapted from a commercial energy monitoring system).
The system will enable identification and quantification of building performance deviations, identification of the conditions under which performance deviations occurs, provide a means to compare Energy Conservation Measures (ECMs) and finally a means to validate improved performance once Energy Saving Measures (retrofit or other) have been taken. Advanced analytics methods will help evaluating those deviations.
Project Results:
All results of the project are summarized in the project publications, deliverables and in the TRIBUTE webpage (http://www.tribute-fp7.eu/).
1. Building design and life cycle and the role of simulation
In current practice, simulation is used almost exclusively in the design stage. In the Nordic countries a fairly large percentage of projects are simulated, while in continental Europe and the US, this share is lower, perhaps 10-20%. It is extremely rare in current practice that the design stage simulation model is used after the design stage. In occasional projects, manual parameter tuning is done in the building commissioning stage to gain understanding in the cause of, primarily, excessive energy use. The developed 3D Tracker enables a systematic use of the design simulation model also during the building operation.
The different modes of simulation usage are described below:
Design optimization. This is the traditional usage of a simulation model in the building process and we will not digress on this further here.
Manual calibration. The commissioning of a new building is often critical, since the hand-over between the contractor and the buyer normally takes place here and at the same time all building systems should be put to operation and quality controlled. There is a need to predict building long-term performance within a limited time (often much less than a full year of operation).
The manual calibration serves the purpose of tuning the simulation model to match collected measurements. In the process, important lessons are learned in terms of the performance of the building, the accuracy of the modelling as well as of measurement errors. There is little prospect of automating this process.
Once the building seems to be operating roughly in the intended way, the simulation model can be connected to the building for Automatic calibration as well as Real time monitoring.
In the Automatic calibration mode, a copy of the real-time simulator image is taken at regular intervals. It is separately simulated, faster than real-time, and parameter identification methods are applied to minimize the error between monitored and simulated signals. A separate class of normalized parameters is introduced for this mode, called key parameter modifiers (KPMs). These are further explained in Section 4. The temporal progress of KPMs represent a key diagnostic mechanism, providing indicators for building long-term health.
In the real-time monitoring mode, the simulator runs in real-time and monitored signals from the real building are fed to the simulator for continuous state estimation. The signals required in order to keep the simulator on the measured trajectory are called control adjustment signals (CAS). These provide short term error diagnostics and are further explained in Section 3.
Once the building has been handed over and system function tests have been passed, we enter into what is here called the continuous commissioning phase. Here, the challenge is to continuously achieve optimal building operation, while being on the alert for sudden as well as gradually aggravated errors. In traditional practice, errors often go undetected for long periods, causing unnecessary energy waste as well as sub-standard occupant comfort (a financial risk for the owner and a frequent cause for tenant lease termination.)
The real-time mode provides the opportunity for real-time virtual sensing, one of the most attractive advantages of the proposed system. Although many signals are measured in a modern building, there are still important measurements that are close to impossible to provide with a reasonable number of installed sensors. There are many examples of meaningful virtual sensing in a building context, but here we will just present two: operative temperature control and fair tenant billing.
Operative temperature is the temperature of a space that is actually felt by an occupant, a mixture of air and long-wave radiation temperatures. Operative temperatures are difficult to measure directly. If local loop controllers maintain operative temperature rather than the traditional air temperature, significant comfort gains can be expected, since operative and air temperatures often differ by several degrees. Better precision in comfort control, leads to less wasted heating and cooling energy.
For multi-family residential buildings, a central heating system is normally installed that serves the whole building. To measure system heat emission in each apartment is difficult, error-prone and costly. (In spite of this, it is often a legal requirement, since it has been shown to save on the order of 20% of heating energy.) In addition to the difficulty of measurement, flats that have a disproportionately large external surface area, e.g. on the corner of the topmost floor, are unfairly penalized, since the building opaque envelope is a common necessity, the cost of which should be distributed among tenants in proportion to floor (useful) area rather than envelope area. A further complication is that significant flows of energy take place between flats, due to conduction. It is not practical to measure these flows directly and this is never done in practice. However, for an individual tenant, these flows can be a significant contribution to the overall heating energy and they should be accounted for in a fair cost distribution system.
When a permanently state-adjusted simulation model is available, all of the surrounding surface energy flows as well as the intended system heat emissions can be virtually measured. This enables truly fair tenant heat billing with a minimum of sensors, without resorting to complex, expensive and still unfair direct heat flux measurements.
The benefit of automatic calibration in the continuous commissioning phase is primarily to track gradual degradation of building components, such as heat exchanger fouling, filter saturation, valve degradation etc. Monitoring the time evolution of the KPMs will also provide information about systematic misuse, such as excessive window ventilation and simultaneous heating and cooling of the same space.
Optimal control is another potential benefit of the 3D Tracker. Similarly, to the off-line identification of parameters, optimal setpoint trajectories can be computed with an off-line copy of the simulator image. A common application for this is to control heating and cooling with respect to forecasted weather and occupancy patterns.
Finally, manual control optimization can be carried out with an off-line copy of the real-time simulator image. Many energy conservation measures involve changes of control setpoints, the actual consequences of which can be difficult to predict. Experimentation on the real building is of course possible, but may lead to unfavourable consequences for building occupants. Performing such experiments instead on the tuned simulator, radical experiments can be made without risk of impairing system operation.
One of the most common building analytics is to use a model built on previous energy performance and compare this model prediction on a new period with the real measurement. After a retrofit, or major change, have been applied to the building, this comparison allows to evaluate the result of the modification, as suggested for example in the International Performance Measure & Validation Protocol (IPMVP). When used in building normal operation (no change assumed), this comparison allows to detect abnormal behaviour (performance follow-up).
The quality of the predicted building performance, and so the accuracy of the comparison, depend on the quality of data but even more on the type of model. Typical model give the building daily consumption as a (multi)linear function of one input (typically outdoor temperature), using other variable named drivers to cluster the data in group of day (weekday or weekend, cooling or heating degree days, activity metric...). In this report, our purpose is to use the continuously calibrated building simulation model in place of simpler models. Manual or automatic calibration in building simulation, use of multiple real data as simulation input, theoretically allows a much better prediction than typical linear or multi-linear model.
The simulation model is compared to linear model (function of heating degree day) and the so-called thermal signature which is a multi-linear model with outdoor temperature and type of day as main input. For the last one, two levels of model identification, namely standard or advanced, have been assessed.
1.1. Central Control System
The system model also receives settings and signals from the Central control system (BMS) of the building. All signals and settings that affect the operation of the building that are known to the BMS should be made available to the system model also settings that are under the authority of occupants. (Occupant settings that are unknown to the BMS are treated by the Occupancy model). The control setting signals sent to the system model must normally be mapped and transformed by a Central control signal transformer in order to match the signals that are available in the system model to cater to similar functions.
1.2. State Controller
A State controller receives all sensed signals from the Real building as well as equivalent computed signals from the system model. A Sensor transformation module will filter, map and transform real sensor signals in order to match the definitions of equivalent measures that are obtained from the system model.
Special among sensed signals are Local meter signals, i.e. electricity use measures that reflect local usage for mains plugs and lights and that can be directly attributable to a given zone in the system model.
The purpose of the State controller is to compute required control action that, when allowed to act on the system model, will minimize the difference between sensed signals of the Real building and the system model.
A control action, called a Control adjustment signal, that will lessen the difference between a real measured and the corresponding computed signal is sent to the system model. This signal will interfere with the local control loop that controls the relevant signal and attempt to increase the output of the controlled device, e.g. mass flow through the radiator in a heating situation.
When direct measures of lighting and/or plug load power are available for a zone (Local meter signals), these are fed through the State controller to the Occupancy model (which in turn feeds them to the system model, overriding any occupant plug load and lighting operation models that may be active in the Occupancy model).
A special module of the State controller is the Occupancy detector. Its purpose is to detect possible occupancy. The method of doing this depends on available sensor signals. Lacking direct occupancy detection signals, the occupancy detector will be sensitive to other traces of occupancy, such as raised levels of temperature, CO2 and humidity, or other events that would implicate an occupant being present such as a light or plug load switching event, a probable window opening etc.
1.3. Occupancy Model
The Occupancy model tries to mimic the behaviour of real zone occupants, in terms of (1) actual presence (Occupancy), (2) any control action that the occupant is authorized to take but that is unknown by the BMS (e.g. change self-acting radiator valve settings, draw blinds, open window etc.), (3) signals to mimic real occupancy plug load and lighting action, and (4) signals for domestic hot water usage.
Occupant control action is modelled based on occupant presence and Computed comfort signals that are read directly from the system model.
When the Occupancy detector module of the State controller is unable to advice on the level of zone occupancy (e.g. if there are no sensors in place with relevance to occupancy), “-1” is sent in terms of Detected occupancy, in which case the internal stochastic presence prediction models of the Occupancy model will estimate zone occupancy.
Any stochastic signal generator in the Occupancy model must have the property of reproducing the exact same signal in two repeated evaluations with the same input, otherwise the optimization process for physical KMPs will be seriously impaired.
Various types of black and white box models can be tried for the Occupancy model. The training of these models is separated from the tuning process of the physical KMPs of the system model. However, having some parameters of the Occupancy model visible and given a “physical” interpretation could be a great advantage to assist the operation of the building by providing diagnostic information about occupant habits, preferences and possible wasteful behaviour (such as regulating temperature by window opening rather than radiator settings in the heating season).
1.4. Load Estimator
In the common situation when Local meter signals are unavailable or incomplete, the Load estimator module attempts to reconcile system model plug loads and lighting power with the physical readings of central electricity meters. This is done by sending a load adjustment requests to the relevant Occupancy models. The occupancy models act on this request by increasing or decreasing the plug loads and lighting that is under its authority.
2. Data Collection
The City of Turin and the Urban Community of La Rochelle are Tribute project partners, and public building portfolio managers. They represent potential operation-phase analytics users, and provide the project with real pilot sites. The paper focuses on one of these sites, the ‘Vaucanson’ building in La Rochelle, France. It is a 3750 m² two-floor office building, erected in 1986 and retrofitted in 2012.
The pilot site of ‘Vaucanson’, in La Rochelle, France, has been used to test the performance follow-up algorithm. All the results presented in this document are from this building.
The main data used by the algorithms are the external temperature and the global building power. Then, the data are processed to obtain a daily value (average external temperature and integral of the building power).
In Vaucanson, the building power is the global electricity consumption as there is no other type of energy used. In other building, like the Torino library, the user must sum the different type of energy consumed by the building (here electricity and gas). The fact that the gas meter was implemented to late in the Torino library explains why we do not apply this analytic to this building.
The Vaucanson building features a small data centre that consumes about 25% of the total energy, and a complex heat production system including a heat pump mounted on ground river pipes and heat recovery from the data centre.
Building models were developed using the IDA Indoor Climate and Energy (IDA-ICE) tool developed by EQUA Simulation AB, a TRIBUTE partner. IDA-ICE is a whole-building simulation tool, based on dynamic multi-zone computations to provide results on thermal indoor climate and energy consumption.
The Vaucanson model has 23 thermal zones (for about 100 rooms). The simplified model takes less computation time and matches available zone ambient sensors. It can run a whole year in 1.5 hours, and log some 100 time-series for real data mapping.
Massive data collection is implemented on all TRIBUTE pilot sites. The Vaucanson data includes:
• Electrical and thermal energy (40 sensors).
• BMS data (124 sensors).
• Ambient sensors for temperature (110 sensors), humidity (75 sensors) and CO2 (21 sensors).
• Weather station (9 sensors).
Data from the different sources are collected via gateways and pushed to StruxureWare™ Energy Operation. This flexible software-as-a-service (SaaS) tool offers scalable functionality, from simple, out-of-the-box reporting to in-depth energy analytics.
Energy data
A Com’X 200 gateway stores and pushes ordered, pre-formatted energy data time-series by FTP in xml files. The gateway requests data from the sub-meters every 10 minutes using Modbus protocol:
• 2 COP monitor systems acquire specific heat pump data.
• Electrical sub-metering includes the main meter, 3 plug load meters, 3 lighting meters, and specific meters for AHU coil, server UPS and cooling system, external lighting and PV production.
• Thermal energy meters are installed on heat pumps, ground river pipes and the primary hot and cold water collectors.
BMS variables
A computer with Niagara AX software acquires the relevant BMS data from two Honeywell Controllers. Data are sampled every 10 minutes and emailed twice a day to the Energy Operation FTP relay server. The server converts the raw CSV files to Energy Operation-readable CSV format.
Zone comfort data
Five Raspberry Pi3 ZigBee Green Power to Ethernet network gateways collect data from Vaucanson’s 87 zone comfort sensors. Time-stamped sample files are sent to the FTP relay server. The temperature, humidity and light sensor samples are sent at less than 2 min intervals. Above the 600+ ppm threshold, samples from the twenty-one CO2 sensors are sent every 2.5 min, and at 20 min intervals when the CO2 ppm is below 600. This feature, combined with the ultra-low power Zigbee Green Power protocol, gives the battery-powered sensor a service life of more than 10 years.
Several other data are used for the simulation model. This include a few energy meters (plug power, lighting, server room consumption) that are automatically inserted as input in the simulation. Also, some key BMS data are used to set the control. This process is manually done, but in future implementation, it will be interesting to have this process more automated. All data are collected with a 10-min time step.
3. Principle
To assess the different algorithms, two time periods are identified: a learning period and a validation period. Learning period is linked to model creation and validation period is linked to model testing.
The learning period is the period when the model is learnt and created, while the validation period is the period when the previous model is tested.
Start date and end date from both periods can be chosen. According to the data available on this building, learning period has been chosen from the 1st of August 2015 to the 31st of December 2015. Validation period is from 1st of January 2016 to 31st of July 2016. In both periods, summer and winter conditions are present.
4. Performance follow-up methods description
Below are the performance follow-up methods which are described in this document:
• Linear regression ax+b
• Thermal Signature - Standard
• Thermal Signature - Advanced
• Simulation model - Real data
4.1. HDD calculation
All three simple performance follow-up methods (linear regression and standard/advanced thermal signature) are using HDD calculation based on daily average external temperature. HDD can be calculated by different formula that are described in Annex. For the current experimentation, we will use the ‘basicMean’ approach.
Also, for each of the HDD calculation method, a heating reference temperature and a cooling reference temperature are needed. It will determine when HDD and CDD will be counted. Explanation will be given in the following part on how these values are defined.
4.2. Linear regression
This method is the simplest to implement but also the one with the less interesting results. It calculates a linear regression from the learning period values, and extend it for validation period on validation values.
For this method, HDD were calculated with ‘basicMean’, and heatingRefTemp was arbitrary chosen:
• heatingRefTemp = 15°C
• coolingRefTemp is not used in this method, since the regression is only made on HDD
For the present dataset (Vaucanson building), it has been stated that influence of heatingRefTemp value is negligible on the linear regression result (see result section below).
Note that this model only considers the heating season. Each point outside this case (e.g. with HDD<0) should either be removed from the analysis, or compare with the model value at HDD=0.
4.3. Thermal Signature
This is the most typical method used for building performance follow-up. Several variants exist, and two of them are evaluated here: the standard and the advanced. They are in fact coming from the same algorithm but with different chosen options.
Schneider has developed a generic calculation brick that can compute thermal signature for many different cases and several method variants. We integrate the Matlab code version of this brick in our analytic framework.
Thermal signature uses the outdoor temperature (transform into HDD) as the main input. It also uses additional drivers to create cluster of day and identify a linear model for each cluster. The most commonly used clusters are the heating/cooling/no thermal control seasons and type of day. The first type of cluster might be identifying based on HVAC control signal, our automatically defined based on HDD/CDD threshold.
In Vaucanson there is no real cooling, so we will only consider two thermal control type. However, to evaluate potential non-constant performance when heating is off, we choose to keep the heating and cooling seasons (instead of heating and no thermal control). In that case, the two temperature references (heatingRefTemp and coolingRefTemp) will be considered, but they will be forced to have the same value.
We will also use the type of day to divide the day in WeekDay and WeekEnd, as there is clearly a different behavior of the building in week end and holidays.
The result presented in the next pages, are obtained first without the type of day driver, then taking this driver into account. All results are presented with two level of method, the standard and advanced one.
Here is the list of options which are the same in both methods:
• No maximal nor minimal external temperature have been defined in the methods (no days are removed because too hot or too cold).
• For each day, daily temperature is calculated by averaging every temperature of the day.
• Chosen calculation for degree days is set to ‘basicMean’. The heatingRefTemp is automatically set by the algorithm and is equal to coolingRefTemp.
Here is the list of the options which are different in the methods:
Standard:
• Robustness mode of the linear regression: best fit found with Least Absolute Deviation regression (Norm 1).
• Outliers are not removed.
• ‘fourParameters’ mode chosen: two linear model with temperature reference and slope. In fact, as we also force heating and cooling reference temperature to be the same, it is a three parameters model. In case where the type of day driver is considered, the number of parameter will be six.
Advanced:
• Robustness mode of the linear regression: a first step uses best fit found with Least Absolute Deviation regression (Norm 1) as in the standard mode. Based on this first regression, outliers are detected and removed using a ‘nbSigma’ scalable parameter (see below). Then, a new regression is made on inputs without the outliers using the Ordinary Least Square regression (Norm 2).
• ‘nbSigma’ parameter set by default: it is a positive scalar value that defines the number of standard deviation used to remove outliers. This value determines the probability of false detection of outliers based on normally distributed data. By default, the value corresponds to 99.7% of false detection probability.
• ‘bestFit’ mode chosen: the algorithm looks for the mode which gives the best result (CVRMSE, see results section below) between the available modes:
o twoParameters: linear regression for HDD
o threeParametersLeading: linear regression for HDD and second line with no slope
o threeParametersLagging line with no slope and linear regression for CDD
o fourParameters: linear regression and slope for both HDD and CDD
o fiveParameters: linear regression and slope for both HDD and CDD and line with no slope in between.
• The advanced algorithm will create all the model and select the one with the best result. In the studied case the bestFit chose the ‘fourParameters’ mode. Note that, as previously mentioned, one parameter is sometime removed as we force heating and cooling reference to be the same. In case where the type of day driver is considered, all the parameter numbers are multiplied by two.
To improve the model quality, we introduce the “Type of Day” driver. Simulation model
In Tribute, we assume that we can use a calibrated building model that is couple with real data. More specifically:
• A model is first calibrated, using a mix of manual and automatic methods.
• The model is then connected to the analytic framework that can be used to push into the model some parameters like the real weather, loads, occupancy or BMS set points.
In Vaucason case, we are pushing in the simulation the real value for external temperature and humidity, internal loads power, lighting power, servers power, inverter power and supply air temperature set point in AHU. This real data input helps the model to be as close as possible to site energy consumption. Indeed, we observe, on the learning period, a reduction of more than 22% of the CVRMSE.
Ideally, more data should be pushed in the model, especially other BMS set points. This has not been done systematically in Tribute, for time reason (implementing this link in the simulation tool and in the automatic data process took some time) and because doing that is not always possible without breaking the energy balance within the simulation tool.
5. Results
5.1. Results indicators
For this performance follow-up algorithm, the CVRMSE (coefficient of variation of root mean square error) metric has been chosen to compare model performance. This metric has already been described in Tribute deliverable D5.2.
5.2. Results on pilot site: Vaucanson building
his section present the results found on the Vaucanson building. In the algorithm, a choice can be made to run every method at the same time. This was done to produce the graphs below, which display result in function of the default HDD value (the one compute with the reference selected in the GUI. On the first graph learning period data can be seen (blue), while linear regression is red and thermal signature (standard and advanced) are in yellow and purple, respectively.
Thermal signatures are both using HDD and CDD, which explains why the curve are composed by two slopes instead of one. We also see the “two models” corresponding to the two “Type of Day”. For the advanced thermal signature, the weekend model seems to be correct. On the contrary, the standard thermal signature for weekend behave curiously (decreasing when HDD rise). This error is probably due to the fact that the weekend data set has too few points with some outliers (not removed in standard method) that led to incorrect model identification.
Simulation data are in green. We clearly see how closer to the real data the simulation is compared to the (multi)linear models. The simulation error is 22% lower than the advanced thermal signature one, which is itself 10% better than a linear model.
Comments:
• Every method give calibration acceptable results according to the standards.
• The ‘Simulation model’ line is not to be taken into account here.
• Vales under the table are indications of input parameters. The heatingRefTemp value is only relevant for linear regression. HDD learn and HDD val are number of heating degree days in each period. Computation method has been explained above.
The improved accuracy of simulation impacts the number of days with an error above the specified threshold (last column of the table), where the simulation is the method with the least days with an important error.
Comments:
• Whatever the threshold value is, the simulation give better results than the others methods.
• For a given threshold (if the user specify that he want to be informed when the error is above a given value), the improvement is greater for intermediate value of the threshold, which are also the most interesting ones for performance follow-up (too small threshold will flag to many days, whils too large one will detect only a few abnormal days).
• We can also consider that the user will define the mean number of day in which he wants to receive an alarm. In that case, he will have to set the error threshold to a much higher value for linear model than for simulation.
6. Conclusion
Accuracy of the various methods analyzed are quite different, and it is directly related to the time spent to build the model. Linear regression is very simple to make, but results are less accurate than simulation, and less robust on new set of data. Calibrated simulation model is very long to make, but it provides accurate results with strong robustness on new data sets.
It must be noted that the methods were compared only on one site. Even if the general trend will probably be confirmed by other site, the absolute value obtained in the comparison are very much link with this specific site. Indeed, the high server consumption and constant pump consumption make the Vaucanson building data set quite “flat” and “noisy” compared to other buildings. If will be nice to apply the same kind of comparison on other buildings, although none with the right data were available in the Tribute project.
Finally, the whole purpose of this analytic is to help the facility manager follow-up his building performance, with less day flagged, less false detection and better analysis capacity. This was not fully tested in the scope of Tribute.
Potential Impact:
1. Introduction to new business models’ opportunities
The solutions developed by some members of the TRIBUTE consortium have a real added value on the market:
• which can modify the added value chain of the different actors of the market;
• which can modify the routes to market of these offers;
• and by consequences which can be considered as new business models for these offers.
Two main new opportunities as business models were identified. The first one is a major need of property developers in construction market: How can he be sure that the delivered building will consume less than a definite value? The solutions, tools and methodology developed in TRIBUTE allow to guarantee the maximum value of the consumption in operation.
The second one is one of the main need of property managers in operation markets: how to decide which instrumentation plan is necessary to be invested to be able to identify energy conservation measures which are profitable by reducing the energy consumption?
Some other solutions of the tribute consortium are very valuable on the market but don’t change the added value chain and so will be commercialised as usual.
2. New business model in construction
The dream of new construction market is to know before, how much they will consume after. The new business model guaranteed energy result for new construction thanks to progresses realised during TRIBUTE research programme is done for that.
2.1. Principle
The principle is to engage the builder/ contractor to assure the quality of the parameters influencing the whole energy consumption, which are under his responsibility whatever are the parameters which are out of its responsibility.
Example the weather is out of its responsibility, the choice of the power of the coffee machine is out of its responsibility. The ability to cut the power of the coffee machine when nobody is in the coffee room is under his responsibility, the fact that an occupant opens a window because it’s too hot in the office is under its responsibility, the decision of the site manager to close the site between Christmas and new year is not under his responsibility.
So The principle of the contract is to create a model based on common and shared hypothesis of the exogeneous variables (out of his responsibility), with these hypothesis to give an energy objectives for his construction, and to compare, what he built ; So the protocol to assess this contract is using all the different aspects studied in TRIBUTE PROJECT;
2.2. Protocol of assessment
This protocol has been applied between Schneider Electric Real Estate as the owner of a new 11000m² office and research building in Grenoble, and GA the main contractor which has been chosen to build it according a building licence. This building is a part of a project called GreenOValley; in this project, the owner assigned an energy ambition for these buildings which are resumed in the power point created for the press conference of the “first stone event”. These objectives are given in chapter 6 as an annexe.
This protocol has been invented, and created between Schneider Electric Innovation Center (represented by the writer of this Tribute deliverable) and ARTELIA, design institute chosen by the main contractor to study and design the technical parts. As it is an experimentation the financial part of the contract are great enough to be motivated, but small enough to avoid that one of the partner is disadvantaged if the protocol and the engagement were not successful.
Better than explaining all the points of this protocol, he is given here in chapter 7 as an annexe. The reader must take care that the original contract between two French companies is written in French, and here is a “internet automatic” translation which can be sometimes not perfect.
2.3. Result and exploitation of Result
The initial timing of the real estate project was perfectly in accordance with tribute project (building delivery end of 2016, and we should have had 6 months of energy consumption to verify if we are far or close from the objective.
Unfortunately, one of the essential part of the project has been delayed: the energy monitoring part, so we haven’t been able to begin the comparison between measurement and simulation.
Thus, the results are not presented here. During the closing ceremony of the project we’ve presented our intentions:
If the experimentation is successful, our (Artelia, GA and Schneider-electric) intention is to write and publish some papers in scientific publications and/ or in the world of real estate medias. In addition, the results will be presented to the French ministry in charge of the constructive regulation.
If the experimentation is not successful, the analysis of the reasons will be done and the conclusions will help to drive the second building of GreenOValley project which will be delivered end of 2019.
3. New business model in operation
A large range of buildings have “in home facility managers” or “in home property managers” in charge and/or in responsibility of the right operation of the building. The standard of energy management has been done for them to help to drive energy improvement.
They are in front of a lot of suppliers which offer “energy performance contract”, “audit and diagnosis services”.
But they haven’t whether the budget to pay for that, or the envy to let other people to do. The “save it yourself“ offer and business model is done for them
3.1. Principle
The principle of the Save It Yourself, offer and BM is to sell or rent a set of sensors easy to install to be able to monitor a part of the installation that the “in home energy Manager wants to control.
Main advantage is if the instrumentation and monitoring doesn’t’ show great saving potential, they can be displaced to find the next source of savings.
The second part of the business is to supply data collection, data monitoring and analytics under different forms (software as a service, pay per use, or usage sales) to help to identify the ECM.
4. Conclusions on business models
These two new business models have been rendered possible on the market thanks to TRIBUTE key findings.
For the first one it’s too early to know, if the assessment protocol is the right one. We are in the creation process in evaluation on a pilot site. Nevertheless, this need is a huge demand on the market, and as soon as some actors will be able to do, keep and success to the pledge of site energy consumption all end uses, the real estate market will ask for that.
For the second one, it is launched and commercialised since a few weeks. It’s too early to know if the business model of tools for “save It Yourself” will be a great success or not, but it is a new way to analyse the market of energy services face to the “great international actors” which investigate this market with a whole, complete, set of solutions and services to do it by themselves. David or Goliath ?
Give the fish of energy savings or teach how to catch it, that is the question.
The EPC contract is for the fish, the ISO50001 management standards is for the lesson. And the gulplug offer imagined, partially thanks to TRIBUTE project is the net.
We’ll see.
5. Dissemination activities
The TRIBUTE website (http://www.TRIBUTE-fp7.eu) has been operational from the end of January 2014. The webpage is considered as a successful tool for raising awareness of the project and its activities. In total the website had more than 17.000 users with their peak around the time of the TRIBUTE workshop (June 2016), TRIBUTE training (June 2017) and TRIBUTE final conference (September 2017). The website has been maintained during the whole course of the project.
During the project, several dissemination materials were developed including the project leaflet, project roll-up as well as TRIBUTE video. The TRIBUTE video was released in the beginning summarizing the project basic scope and objectives, presenting the project test sitesTRIBUTE also contributed to several editions of EeB project review. The aim of the publication is to provide a dissemination route for EeB PPP projects; capture and showcase the impact of EeB PPP projects; cluster projects by topics to demonstrate continuity of activity; demonstrate the wave-based activity facilitated by the EeB PPP; constitute an effective set of case studies that will encourage further companies to join the association and participate on future EeB PPP calls.
TRIBUTE press release was prepared in the beginning and in the end of the project and shared through partners’ networks. Objective of this press release was to announce the end of the project, to highlight the main achievements and perspectives for future exploitation. In the relation with the TRIBUTE final event, press conference was organized and numerous French newspapers having the highest reading rates, presented articles dedicated to the TRIBUTE project’s scopes and targets.
Clustering activities were planned as a necessary part of the project dissemination. TRIBUTE project participated in several editions of Workshop on the Impact of EeB PPP that was jointly organized by the European Commission with the support of the E2BA. The objectives of the workshops were to address innovation and exploitation issues in running projects and explore potential for cross-project clustering. Several joint workshops were also organised in cooperation with the PERFORMER and ENERGY IN TIME projects. The main goal of these meetings was to discuss and understand the positioning of the projects towards each other (since they have been all funded from the same call topic), to identify common actions and share ideas on possible cooperation. Key common areas of interest were identified, action list proposed and further cooperation between the projects had been foreseen. Common areas of interest were identified as data acquisition, fault detection and dissemination. TRIBUTE had also very strong synergies AMBASSADOR project.
Publication of TRIBUTE results to relevant scientific and industrial periodicals, journals and key conferences in Europe was assured during the whole project lifetime. The project participated with a project booth in conferences such as Industrial Technologies 2014 (April 2014, Athens) or Re-Industrialisation of Europe 2016 (October 2016, Bratislava). Project results were also presented at several editions of Sustainable Places conference (2014, 2015, 2016), where also joint clustering workshops were organised with related projects in EeB domain. Furthermore, the project was represented at ECTP-E2BA Conference 2014 (Brussels, June 2014), Energy Smart Building (Torino, July 2014), L’Energia del Territorio (Torino, March 2015), TRIBUTE presentation during EQUA users group in Switzerland (March 2015), IBPSA 2015 (Hyderabad, December 2015), France-Sweden Workshop for Sustainable Cities (La Rochelle, May 2015), PROBIS EU PROJECT Event (Turin, July 2015), Climamed conference (Juan les Pins, September 2015), TURIN ENERGY MANAGEMENT EVENT (Turin, September 2015), AICARR BOLOGNA (Bologna, October 2015), SBE16 (Turin, February 2016), IRES Conference 2016 (Düsseldorf, March 2016), IntelliSys - 25th SAI Intelligent Systems Conference (London, September 2016), CESB 2016 (Prague, June 2016), IBPSA 2017 (San Francisco, August 2017), 12th IEA Heat Pump Conference (Rotterdam, April 2017), PETRA 2017 (Island of Rhodes, 2017).
Peer reviewed articles were prepared for Building and Environment journal, Applied Thermal Engineering journal and Applied Energy journal. In addition, peer reviewed articles were prepared for SAI Intelligent Systems Conference (IntelliSys) 2016, Building simulation 2015 conference, 10th International Renewable Energy Storage conference 2016, 2nd ACM International Conference on Embedded Systems for Energy-Efficient Built Environments – BuildSys 2015, 12th IEA heat pump conference 2017, as well as 10th International Conference on PErvasive Technologies Related to Assistive Environments - PETRA 2017 and are available in their proceedings. AMBASSADOR project was also presented in the VTT Research Highlights publication.
TRIBUTE results were also disseminated through organization of TRIBUTE events, especially TRIBUTE dedicated workshop for energy managers that took place in the frame of Sustainable places 2016 conference in Biarritz, France in June 2016. TRIBUTE training for energy managers that took place in June 2017 in Grenoble, France in the premises of the Schneider Electric was dedicated to topics such as overall energy context, optimization strategies for buildings and closing the gap between the reality and simulation.
At the end of the project the TRIBUTE final conference took place (La Rochelle, France, September 2017). During the event the project was revealed to the public and the results were made well known to the audience. This was a very important action for the recognition of the project to the public and for giving the importance of project’s targets regarding the sustainable energy management in building level. The representatives from the whole project’s consortium were giving all the necessary information and gladly answering all the questions given from the public. The conference attracted a wide range of stakeholders from the whole value chain, such as representatives from the industrial sector and the public authorities, engineers from different field as well as students and private investors. Additionally, national media, having high broadcasted rates, covered the final conference of the TRIBUTE project and make it well known to the national audience.
6. Exploitation activities
TRIBUTE exploitation activities were designed in order to evaluate the collective impact potential of the consortium by evaluating the market potential and to determine product opportunities in relation to the customer/product requirements throughout the course of the project. The activities covered the methodology developed by the consortium to complete a market description of buildings, to identify the different actors of the added value chains and their behaviours and expectations in the field of value of TRIBUTE. The exploitation plan was focused on different business cases depending on the organisation of actors and their territory of responsibility. For the behavioural typology of actors in exploitation market, the analogy is done selecting situations where the different types belong to the same organisation, with a common management or not. According to this definition, we meet during our survey 7 main business cases.
The main actors to be targeted for energy optimisations at building level or at district level are the Energy Manager persona actors.
The easiest situation is when they belong to the same organisation as the owner actors’ persona (business cases number 1 and 2). In the same decision process can be balanced investment costs and energy bill reductions and all the business models can be easily developed.
The second situation is when the energy manager belongs to a tenant organisation with a long term contract (business cases number 4 and 5). In that case the balance between investments costs and energy bill reduction can be argued if and only if the payback is shorter than the duration of the lease.
A special situation is when the site is managed by an external ESCO in charge of FM and EM through an Energy Performance Contract (EPC) as shown in business case number 3. The target becomes the ESCO itself, if the added value brought can serve, optimise or simplify the execution of the contract.
The last positive situation is described by business case number 6 where the EM belongs to a short term lease, but in this case only business models without customers’ investments (service provider or energy optimiser) can be developed actively on the market.
In all these situations, the FM can be a brake, when he is not in the same organisation than the EM, as solutions of energy optimisation can modify the conditions of execution of his own objectives.
In all these situations, the occupant can be a positive or negative influence maker, depending on the effect on the quality of building services, or on his perception of this effect.
Finally, the risky situation is when all the stakeholders have different objectives (business cases number 7). This can be met when they belong to the same organisation, but a structured one, with a segmentation of different services, departments or divisions, or when they belong to completely different organisations. In this situation, it will be very difficult whatever the business model is to promote energy optimisations.
List of Websites:
Webpage
• http://ambassador-fp7.eu/
Project coordinator
• Mr. Martin Sénéclauze (CSEM Centre Suisse d'Electronique et de Microtechnique SA - Recherche et Developpemnent)
• E-mail: martin.seneclauze@csem.ch
• Tel: +41 32 720 5340
Technical coordinator
• Mr. Henri Obara (Schneider Electric Industries SAS)
• E-mail: henri2.obara@fr.schneider-electric.com
• Tel: +33 (0) 476 57 65 12
Project manager / Dissemination manager
• Dr. Václav Smítka (Amires s.r.o.)
• E-mail: smitka@amires.eu
• Tel: +420 732 304 379
Exploitation manager
• Mr. Olivier Cottet (Schneider Electric Industries SAS)
• E-mail: olivier.cottet@schneider-electric.com
• Tel: +33 (0) 476 39 11 48
The TRIBUTE project ended by end of October after 4 years of research on closing the gap between simulation and reality. 16 partners were contributing to the results of the project, with an overall budget of 10 M€, and 6.7 M€ funding from European commission in the frame of FP7 framework.
Buildings represent more than 40% of EU energy consumption and 36% of EU CO2 emissions. To reach the 2020 20% energy efficiency target, the EU energy efficiency directive sets binding measures to be implemented by EU members by the end of 2016. A new package of measures has recently been proposed to extend the energy efficiency target to 40% for 2030, including proposed revision of European Building Performance Directive. Building Energy Performance Simulation (BEPS) tools are widely used for modern mid-size to large buildings. When used across the building commissioning and operation phases, BEPS models can be powerful tools to assist building operators and facility managers in creating virtual sensors, performing sensitivity analyses, assessing building energy performance, detecting anomalies and identifying potential retrofit solutions.
European Tribute project FP7 has been exploring various BES applications to enable building energy managers to understand BES, and to reduce the gap between predicted and real performances. The pre-requisites for the deployment of analytics are a commissioned building and a calibrated model. The analytics proposed entail a large number of real data and related simulations. To facilitate use by building facility managers, not only does all data management have to be automatic (real data collection and storage, simulation of settings, processing, and real versus simulated mapping); the execution of analytics and display of results must be automatic as well.
TRIBUTE has provided a building monitoring system for bridging the gap between the design-stage simulation model and the physical building.
The approach relies on a calibrated simulation model of the building that would typically be developed in the design phase. Business margins are small in this industry, and the cost of developing a tailored model for the operation phase would often be prohibitive. From a general simulation perspective, such a design model is severely over-parameterized. These Parameters correspond to physical measures of individual building-parts, and therefore the total number of parameters for an individual room can be in the order of hundreds. To manage this over-parameterized model for system identification is a challenge.
The developed system targets the expert operator. It is envisioned that each such operator can manage a large number of buildings from a remote location. Separate user interfaces may in subsequent work also be developed for the day-to-day building maintenance personnel, tenants, and other stakeholders such as owners.
Furthermore, the present work is strictly aiming at the development and presentation of a model that is in real-time kept in tune with the physical building. Further use of this model for Fault Detection and Diagnosis or predictive control is not presently encompassed.
One of the most common building analytics is to use a model built on historical energy performance and compare this model prediction on a new period with the real measurement. After a retrofit, or major change, have been applied to the building, this comparison allows to evaluate the result of the modification, as suggested for example in the International Performance Measure & Validation Protocol (IPMVP). When used in building normal operation (no change assumed), this comparison allows to detect abnormal behaviour in a process named performance follow-up.
This document reports the evaluation of using a continuously calibrated building simulation model. The IDA-ICE model is compared to simpler linear, or multi-linear model typically used for performance follow-up. A simple GUI have been developed to quickly compare the various models on different settings. Note that, at this stage, this GUI is more a research tool than something useful for facility manager.
The software was tested on the Vaucanson office building in La Rochelle. As expected, the building simulation get better results than the other simpler models, both on the learning and testing horizon. This reflects the fact that the simulation model considers a larger number of phenomena thus allowing better identification of real faults.
With error reduced by more than 20% compared to advanced thermal signature, which is itself 10% lower than pure linear model, the simulation can allow to reduce greatly the number of day that will be considered as faulty compare to the learning period.
Of course, simulation based follow-up is much costlier to set than more typical methods, so it will probably never be developed for this single usage. But, if we have the model available for multiple usage (fault detection, advanced control, retrofit option analysis), it will bring a high value at limited cost for this type of analytic.
Project Context and Objectives:
Today, Building Energy Performance Simulation (BEPS) analysis tends to show a large discrepancy with real energy performance. Most cases are due to gross mistakes rather than fundamental inadequacy of available technology and methods. The reasons are manifold. Highly simplified calculation methods are used far beyond their domain of validity. Assumed boundary conditions such as occupant behaviour are not in accordance with actual usage; gross malfunctions in control and HVAC systems are left undetected in the commissioning process, while thermal bridges and distribution system losses are left without attention. Moreover, metered and sub-metered data are not used efficiently in calculation tools and engineering based simulation models during the Measurement and Verification (M&V) phase.
TRIBUTE aims at minimizing the gap between computed and measured energy performances through the improvement of the predictive capability of a commercial BEPS. To achieve this goal, a set of intermediate objectives were set:
• Extend the use of simulation to the commissioning and operation stages of a building
• Allow efficient on-line identification of building key parameters
• Deploy a generous monitoring infrastructure on which the simulation could be trained and benchmarked
• Reduce the number of sensors to allow long term / affordable monitoring of the buildings
• Development of significantly faster and more precise retrofit decision tool that will be offered to European regions and individuals
• Improvement of energy flow management systems taking into account occupancy behaviour and featuring a continuous building health monitoring approach.
The methodology and tools will be evaluated in the context of different buildings types and locations. Within the TRIBUTE project we propose to reduce the gap between planned and actual energy performance down to 5 % (+- 2 %).
TRIBUTE proposes to demonstrate an automated, model-based whole-building performance monitoring system at sites in partnership with Communauté d’agglomération de La Rochelle and Città di Torino.
In these existing buildings, measurement and verification techniques will be developed and deployed.
The system will continuously acquire measurements of HVAC-lighting, plug loads, occupancy usage from existing control system augmented by additional sensors to capture other building parameters measurements. The system will connect to Building Health Monitoring and Energy Flow Management applications (adapted from a commercial energy monitoring system).
The system will enable identification and quantification of building performance deviations, identification of the conditions under which performance deviations occurs, provide a means to compare Energy Conservation Measures (ECMs) and finally a means to validate improved performance once Energy Saving Measures (retrofit or other) have been taken. Advanced analytics methods will help evaluating those deviations.
Project Results:
All results of the project are summarized in the project publications, deliverables and in the TRIBUTE webpage (http://www.tribute-fp7.eu/).
1. Building design and life cycle and the role of simulation
In current practice, simulation is used almost exclusively in the design stage. In the Nordic countries a fairly large percentage of projects are simulated, while in continental Europe and the US, this share is lower, perhaps 10-20%. It is extremely rare in current practice that the design stage simulation model is used after the design stage. In occasional projects, manual parameter tuning is done in the building commissioning stage to gain understanding in the cause of, primarily, excessive energy use. The developed 3D Tracker enables a systematic use of the design simulation model also during the building operation.
The different modes of simulation usage are described below:
Design optimization. This is the traditional usage of a simulation model in the building process and we will not digress on this further here.
Manual calibration. The commissioning of a new building is often critical, since the hand-over between the contractor and the buyer normally takes place here and at the same time all building systems should be put to operation and quality controlled. There is a need to predict building long-term performance within a limited time (often much less than a full year of operation).
The manual calibration serves the purpose of tuning the simulation model to match collected measurements. In the process, important lessons are learned in terms of the performance of the building, the accuracy of the modelling as well as of measurement errors. There is little prospect of automating this process.
Once the building seems to be operating roughly in the intended way, the simulation model can be connected to the building for Automatic calibration as well as Real time monitoring.
In the Automatic calibration mode, a copy of the real-time simulator image is taken at regular intervals. It is separately simulated, faster than real-time, and parameter identification methods are applied to minimize the error between monitored and simulated signals. A separate class of normalized parameters is introduced for this mode, called key parameter modifiers (KPMs). These are further explained in Section 4. The temporal progress of KPMs represent a key diagnostic mechanism, providing indicators for building long-term health.
In the real-time monitoring mode, the simulator runs in real-time and monitored signals from the real building are fed to the simulator for continuous state estimation. The signals required in order to keep the simulator on the measured trajectory are called control adjustment signals (CAS). These provide short term error diagnostics and are further explained in Section 3.
Once the building has been handed over and system function tests have been passed, we enter into what is here called the continuous commissioning phase. Here, the challenge is to continuously achieve optimal building operation, while being on the alert for sudden as well as gradually aggravated errors. In traditional practice, errors often go undetected for long periods, causing unnecessary energy waste as well as sub-standard occupant comfort (a financial risk for the owner and a frequent cause for tenant lease termination.)
The real-time mode provides the opportunity for real-time virtual sensing, one of the most attractive advantages of the proposed system. Although many signals are measured in a modern building, there are still important measurements that are close to impossible to provide with a reasonable number of installed sensors. There are many examples of meaningful virtual sensing in a building context, but here we will just present two: operative temperature control and fair tenant billing.
Operative temperature is the temperature of a space that is actually felt by an occupant, a mixture of air and long-wave radiation temperatures. Operative temperatures are difficult to measure directly. If local loop controllers maintain operative temperature rather than the traditional air temperature, significant comfort gains can be expected, since operative and air temperatures often differ by several degrees. Better precision in comfort control, leads to less wasted heating and cooling energy.
For multi-family residential buildings, a central heating system is normally installed that serves the whole building. To measure system heat emission in each apartment is difficult, error-prone and costly. (In spite of this, it is often a legal requirement, since it has been shown to save on the order of 20% of heating energy.) In addition to the difficulty of measurement, flats that have a disproportionately large external surface area, e.g. on the corner of the topmost floor, are unfairly penalized, since the building opaque envelope is a common necessity, the cost of which should be distributed among tenants in proportion to floor (useful) area rather than envelope area. A further complication is that significant flows of energy take place between flats, due to conduction. It is not practical to measure these flows directly and this is never done in practice. However, for an individual tenant, these flows can be a significant contribution to the overall heating energy and they should be accounted for in a fair cost distribution system.
When a permanently state-adjusted simulation model is available, all of the surrounding surface energy flows as well as the intended system heat emissions can be virtually measured. This enables truly fair tenant heat billing with a minimum of sensors, without resorting to complex, expensive and still unfair direct heat flux measurements.
The benefit of automatic calibration in the continuous commissioning phase is primarily to track gradual degradation of building components, such as heat exchanger fouling, filter saturation, valve degradation etc. Monitoring the time evolution of the KPMs will also provide information about systematic misuse, such as excessive window ventilation and simultaneous heating and cooling of the same space.
Optimal control is another potential benefit of the 3D Tracker. Similarly, to the off-line identification of parameters, optimal setpoint trajectories can be computed with an off-line copy of the simulator image. A common application for this is to control heating and cooling with respect to forecasted weather and occupancy patterns.
Finally, manual control optimization can be carried out with an off-line copy of the real-time simulator image. Many energy conservation measures involve changes of control setpoints, the actual consequences of which can be difficult to predict. Experimentation on the real building is of course possible, but may lead to unfavourable consequences for building occupants. Performing such experiments instead on the tuned simulator, radical experiments can be made without risk of impairing system operation.
One of the most common building analytics is to use a model built on previous energy performance and compare this model prediction on a new period with the real measurement. After a retrofit, or major change, have been applied to the building, this comparison allows to evaluate the result of the modification, as suggested for example in the International Performance Measure & Validation Protocol (IPMVP). When used in building normal operation (no change assumed), this comparison allows to detect abnormal behaviour (performance follow-up).
The quality of the predicted building performance, and so the accuracy of the comparison, depend on the quality of data but even more on the type of model. Typical model give the building daily consumption as a (multi)linear function of one input (typically outdoor temperature), using other variable named drivers to cluster the data in group of day (weekday or weekend, cooling or heating degree days, activity metric...). In this report, our purpose is to use the continuously calibrated building simulation model in place of simpler models. Manual or automatic calibration in building simulation, use of multiple real data as simulation input, theoretically allows a much better prediction than typical linear or multi-linear model.
The simulation model is compared to linear model (function of heating degree day) and the so-called thermal signature which is a multi-linear model with outdoor temperature and type of day as main input. For the last one, two levels of model identification, namely standard or advanced, have been assessed.
1.1. Central Control System
The system model also receives settings and signals from the Central control system (BMS) of the building. All signals and settings that affect the operation of the building that are known to the BMS should be made available to the system model also settings that are under the authority of occupants. (Occupant settings that are unknown to the BMS are treated by the Occupancy model). The control setting signals sent to the system model must normally be mapped and transformed by a Central control signal transformer in order to match the signals that are available in the system model to cater to similar functions.
1.2. State Controller
A State controller receives all sensed signals from the Real building as well as equivalent computed signals from the system model. A Sensor transformation module will filter, map and transform real sensor signals in order to match the definitions of equivalent measures that are obtained from the system model.
Special among sensed signals are Local meter signals, i.e. electricity use measures that reflect local usage for mains plugs and lights and that can be directly attributable to a given zone in the system model.
The purpose of the State controller is to compute required control action that, when allowed to act on the system model, will minimize the difference between sensed signals of the Real building and the system model.
A control action, called a Control adjustment signal, that will lessen the difference between a real measured and the corresponding computed signal is sent to the system model. This signal will interfere with the local control loop that controls the relevant signal and attempt to increase the output of the controlled device, e.g. mass flow through the radiator in a heating situation.
When direct measures of lighting and/or plug load power are available for a zone (Local meter signals), these are fed through the State controller to the Occupancy model (which in turn feeds them to the system model, overriding any occupant plug load and lighting operation models that may be active in the Occupancy model).
A special module of the State controller is the Occupancy detector. Its purpose is to detect possible occupancy. The method of doing this depends on available sensor signals. Lacking direct occupancy detection signals, the occupancy detector will be sensitive to other traces of occupancy, such as raised levels of temperature, CO2 and humidity, or other events that would implicate an occupant being present such as a light or plug load switching event, a probable window opening etc.
1.3. Occupancy Model
The Occupancy model tries to mimic the behaviour of real zone occupants, in terms of (1) actual presence (Occupancy), (2) any control action that the occupant is authorized to take but that is unknown by the BMS (e.g. change self-acting radiator valve settings, draw blinds, open window etc.), (3) signals to mimic real occupancy plug load and lighting action, and (4) signals for domestic hot water usage.
Occupant control action is modelled based on occupant presence and Computed comfort signals that are read directly from the system model.
When the Occupancy detector module of the State controller is unable to advice on the level of zone occupancy (e.g. if there are no sensors in place with relevance to occupancy), “-1” is sent in terms of Detected occupancy, in which case the internal stochastic presence prediction models of the Occupancy model will estimate zone occupancy.
Any stochastic signal generator in the Occupancy model must have the property of reproducing the exact same signal in two repeated evaluations with the same input, otherwise the optimization process for physical KMPs will be seriously impaired.
Various types of black and white box models can be tried for the Occupancy model. The training of these models is separated from the tuning process of the physical KMPs of the system model. However, having some parameters of the Occupancy model visible and given a “physical” interpretation could be a great advantage to assist the operation of the building by providing diagnostic information about occupant habits, preferences and possible wasteful behaviour (such as regulating temperature by window opening rather than radiator settings in the heating season).
1.4. Load Estimator
In the common situation when Local meter signals are unavailable or incomplete, the Load estimator module attempts to reconcile system model plug loads and lighting power with the physical readings of central electricity meters. This is done by sending a load adjustment requests to the relevant Occupancy models. The occupancy models act on this request by increasing or decreasing the plug loads and lighting that is under its authority.
2. Data Collection
The City of Turin and the Urban Community of La Rochelle are Tribute project partners, and public building portfolio managers. They represent potential operation-phase analytics users, and provide the project with real pilot sites. The paper focuses on one of these sites, the ‘Vaucanson’ building in La Rochelle, France. It is a 3750 m² two-floor office building, erected in 1986 and retrofitted in 2012.
The pilot site of ‘Vaucanson’, in La Rochelle, France, has been used to test the performance follow-up algorithm. All the results presented in this document are from this building.
The main data used by the algorithms are the external temperature and the global building power. Then, the data are processed to obtain a daily value (average external temperature and integral of the building power).
In Vaucanson, the building power is the global electricity consumption as there is no other type of energy used. In other building, like the Torino library, the user must sum the different type of energy consumed by the building (here electricity and gas). The fact that the gas meter was implemented to late in the Torino library explains why we do not apply this analytic to this building.
The Vaucanson building features a small data centre that consumes about 25% of the total energy, and a complex heat production system including a heat pump mounted on ground river pipes and heat recovery from the data centre.
Building models were developed using the IDA Indoor Climate and Energy (IDA-ICE) tool developed by EQUA Simulation AB, a TRIBUTE partner. IDA-ICE is a whole-building simulation tool, based on dynamic multi-zone computations to provide results on thermal indoor climate and energy consumption.
The Vaucanson model has 23 thermal zones (for about 100 rooms). The simplified model takes less computation time and matches available zone ambient sensors. It can run a whole year in 1.5 hours, and log some 100 time-series for real data mapping.
Massive data collection is implemented on all TRIBUTE pilot sites. The Vaucanson data includes:
• Electrical and thermal energy (40 sensors).
• BMS data (124 sensors).
• Ambient sensors for temperature (110 sensors), humidity (75 sensors) and CO2 (21 sensors).
• Weather station (9 sensors).
Data from the different sources are collected via gateways and pushed to StruxureWare™ Energy Operation. This flexible software-as-a-service (SaaS) tool offers scalable functionality, from simple, out-of-the-box reporting to in-depth energy analytics.
Energy data
A Com’X 200 gateway stores and pushes ordered, pre-formatted energy data time-series by FTP in xml files. The gateway requests data from the sub-meters every 10 minutes using Modbus protocol:
• 2 COP monitor systems acquire specific heat pump data.
• Electrical sub-metering includes the main meter, 3 plug load meters, 3 lighting meters, and specific meters for AHU coil, server UPS and cooling system, external lighting and PV production.
• Thermal energy meters are installed on heat pumps, ground river pipes and the primary hot and cold water collectors.
BMS variables
A computer with Niagara AX software acquires the relevant BMS data from two Honeywell Controllers. Data are sampled every 10 minutes and emailed twice a day to the Energy Operation FTP relay server. The server converts the raw CSV files to Energy Operation-readable CSV format.
Zone comfort data
Five Raspberry Pi3 ZigBee Green Power to Ethernet network gateways collect data from Vaucanson’s 87 zone comfort sensors. Time-stamped sample files are sent to the FTP relay server. The temperature, humidity and light sensor samples are sent at less than 2 min intervals. Above the 600+ ppm threshold, samples from the twenty-one CO2 sensors are sent every 2.5 min, and at 20 min intervals when the CO2 ppm is below 600. This feature, combined with the ultra-low power Zigbee Green Power protocol, gives the battery-powered sensor a service life of more than 10 years.
Several other data are used for the simulation model. This include a few energy meters (plug power, lighting, server room consumption) that are automatically inserted as input in the simulation. Also, some key BMS data are used to set the control. This process is manually done, but in future implementation, it will be interesting to have this process more automated. All data are collected with a 10-min time step.
3. Principle
To assess the different algorithms, two time periods are identified: a learning period and a validation period. Learning period is linked to model creation and validation period is linked to model testing.
The learning period is the period when the model is learnt and created, while the validation period is the period when the previous model is tested.
Start date and end date from both periods can be chosen. According to the data available on this building, learning period has been chosen from the 1st of August 2015 to the 31st of December 2015. Validation period is from 1st of January 2016 to 31st of July 2016. In both periods, summer and winter conditions are present.
4. Performance follow-up methods description
Below are the performance follow-up methods which are described in this document:
• Linear regression ax+b
• Thermal Signature - Standard
• Thermal Signature - Advanced
• Simulation model - Real data
4.1. HDD calculation
All three simple performance follow-up methods (linear regression and standard/advanced thermal signature) are using HDD calculation based on daily average external temperature. HDD can be calculated by different formula that are described in Annex. For the current experimentation, we will use the ‘basicMean’ approach.
Also, for each of the HDD calculation method, a heating reference temperature and a cooling reference temperature are needed. It will determine when HDD and CDD will be counted. Explanation will be given in the following part on how these values are defined.
4.2. Linear regression
This method is the simplest to implement but also the one with the less interesting results. It calculates a linear regression from the learning period values, and extend it for validation period on validation values.
For this method, HDD were calculated with ‘basicMean’, and heatingRefTemp was arbitrary chosen:
• heatingRefTemp = 15°C
• coolingRefTemp is not used in this method, since the regression is only made on HDD
For the present dataset (Vaucanson building), it has been stated that influence of heatingRefTemp value is negligible on the linear regression result (see result section below).
Note that this model only considers the heating season. Each point outside this case (e.g. with HDD<0) should either be removed from the analysis, or compare with the model value at HDD=0.
4.3. Thermal Signature
This is the most typical method used for building performance follow-up. Several variants exist, and two of them are evaluated here: the standard and the advanced. They are in fact coming from the same algorithm but with different chosen options.
Schneider has developed a generic calculation brick that can compute thermal signature for many different cases and several method variants. We integrate the Matlab code version of this brick in our analytic framework.
Thermal signature uses the outdoor temperature (transform into HDD) as the main input. It also uses additional drivers to create cluster of day and identify a linear model for each cluster. The most commonly used clusters are the heating/cooling/no thermal control seasons and type of day. The first type of cluster might be identifying based on HVAC control signal, our automatically defined based on HDD/CDD threshold.
In Vaucanson there is no real cooling, so we will only consider two thermal control type. However, to evaluate potential non-constant performance when heating is off, we choose to keep the heating and cooling seasons (instead of heating and no thermal control). In that case, the two temperature references (heatingRefTemp and coolingRefTemp) will be considered, but they will be forced to have the same value.
We will also use the type of day to divide the day in WeekDay and WeekEnd, as there is clearly a different behavior of the building in week end and holidays.
The result presented in the next pages, are obtained first without the type of day driver, then taking this driver into account. All results are presented with two level of method, the standard and advanced one.
Here is the list of options which are the same in both methods:
• No maximal nor minimal external temperature have been defined in the methods (no days are removed because too hot or too cold).
• For each day, daily temperature is calculated by averaging every temperature of the day.
• Chosen calculation for degree days is set to ‘basicMean’. The heatingRefTemp is automatically set by the algorithm and is equal to coolingRefTemp.
Here is the list of the options which are different in the methods:
Standard:
• Robustness mode of the linear regression: best fit found with Least Absolute Deviation regression (Norm 1).
• Outliers are not removed.
• ‘fourParameters’ mode chosen: two linear model with temperature reference and slope. In fact, as we also force heating and cooling reference temperature to be the same, it is a three parameters model. In case where the type of day driver is considered, the number of parameter will be six.
Advanced:
• Robustness mode of the linear regression: a first step uses best fit found with Least Absolute Deviation regression (Norm 1) as in the standard mode. Based on this first regression, outliers are detected and removed using a ‘nbSigma’ scalable parameter (see below). Then, a new regression is made on inputs without the outliers using the Ordinary Least Square regression (Norm 2).
• ‘nbSigma’ parameter set by default: it is a positive scalar value that defines the number of standard deviation used to remove outliers. This value determines the probability of false detection of outliers based on normally distributed data. By default, the value corresponds to 99.7% of false detection probability.
• ‘bestFit’ mode chosen: the algorithm looks for the mode which gives the best result (CVRMSE, see results section below) between the available modes:
o twoParameters: linear regression for HDD
o threeParametersLeading: linear regression for HDD and second line with no slope
o threeParametersLagging line with no slope and linear regression for CDD
o fourParameters: linear regression and slope for both HDD and CDD
o fiveParameters: linear regression and slope for both HDD and CDD and line with no slope in between.
• The advanced algorithm will create all the model and select the one with the best result. In the studied case the bestFit chose the ‘fourParameters’ mode. Note that, as previously mentioned, one parameter is sometime removed as we force heating and cooling reference to be the same. In case where the type of day driver is considered, all the parameter numbers are multiplied by two.
To improve the model quality, we introduce the “Type of Day” driver. Simulation model
In Tribute, we assume that we can use a calibrated building model that is couple with real data. More specifically:
• A model is first calibrated, using a mix of manual and automatic methods.
• The model is then connected to the analytic framework that can be used to push into the model some parameters like the real weather, loads, occupancy or BMS set points.
In Vaucason case, we are pushing in the simulation the real value for external temperature and humidity, internal loads power, lighting power, servers power, inverter power and supply air temperature set point in AHU. This real data input helps the model to be as close as possible to site energy consumption. Indeed, we observe, on the learning period, a reduction of more than 22% of the CVRMSE.
Ideally, more data should be pushed in the model, especially other BMS set points. This has not been done systematically in Tribute, for time reason (implementing this link in the simulation tool and in the automatic data process took some time) and because doing that is not always possible without breaking the energy balance within the simulation tool.
5. Results
5.1. Results indicators
For this performance follow-up algorithm, the CVRMSE (coefficient of variation of root mean square error) metric has been chosen to compare model performance. This metric has already been described in Tribute deliverable D5.2.
5.2. Results on pilot site: Vaucanson building
his section present the results found on the Vaucanson building. In the algorithm, a choice can be made to run every method at the same time. This was done to produce the graphs below, which display result in function of the default HDD value (the one compute with the reference selected in the GUI. On the first graph learning period data can be seen (blue), while linear regression is red and thermal signature (standard and advanced) are in yellow and purple, respectively.
Thermal signatures are both using HDD and CDD, which explains why the curve are composed by two slopes instead of one. We also see the “two models” corresponding to the two “Type of Day”. For the advanced thermal signature, the weekend model seems to be correct. On the contrary, the standard thermal signature for weekend behave curiously (decreasing when HDD rise). This error is probably due to the fact that the weekend data set has too few points with some outliers (not removed in standard method) that led to incorrect model identification.
Simulation data are in green. We clearly see how closer to the real data the simulation is compared to the (multi)linear models. The simulation error is 22% lower than the advanced thermal signature one, which is itself 10% better than a linear model.
Comments:
• Every method give calibration acceptable results according to the standards.
• The ‘Simulation model’ line is not to be taken into account here.
• Vales under the table are indications of input parameters. The heatingRefTemp value is only relevant for linear regression. HDD learn and HDD val are number of heating degree days in each period. Computation method has been explained above.
The improved accuracy of simulation impacts the number of days with an error above the specified threshold (last column of the table), where the simulation is the method with the least days with an important error.
Comments:
• Whatever the threshold value is, the simulation give better results than the others methods.
• For a given threshold (if the user specify that he want to be informed when the error is above a given value), the improvement is greater for intermediate value of the threshold, which are also the most interesting ones for performance follow-up (too small threshold will flag to many days, whils too large one will detect only a few abnormal days).
• We can also consider that the user will define the mean number of day in which he wants to receive an alarm. In that case, he will have to set the error threshold to a much higher value for linear model than for simulation.
6. Conclusion
Accuracy of the various methods analyzed are quite different, and it is directly related to the time spent to build the model. Linear regression is very simple to make, but results are less accurate than simulation, and less robust on new set of data. Calibrated simulation model is very long to make, but it provides accurate results with strong robustness on new data sets.
It must be noted that the methods were compared only on one site. Even if the general trend will probably be confirmed by other site, the absolute value obtained in the comparison are very much link with this specific site. Indeed, the high server consumption and constant pump consumption make the Vaucanson building data set quite “flat” and “noisy” compared to other buildings. If will be nice to apply the same kind of comparison on other buildings, although none with the right data were available in the Tribute project.
Finally, the whole purpose of this analytic is to help the facility manager follow-up his building performance, with less day flagged, less false detection and better analysis capacity. This was not fully tested in the scope of Tribute.
Potential Impact:
1. Introduction to new business models’ opportunities
The solutions developed by some members of the TRIBUTE consortium have a real added value on the market:
• which can modify the added value chain of the different actors of the market;
• which can modify the routes to market of these offers;
• and by consequences which can be considered as new business models for these offers.
Two main new opportunities as business models were identified. The first one is a major need of property developers in construction market: How can he be sure that the delivered building will consume less than a definite value? The solutions, tools and methodology developed in TRIBUTE allow to guarantee the maximum value of the consumption in operation.
The second one is one of the main need of property managers in operation markets: how to decide which instrumentation plan is necessary to be invested to be able to identify energy conservation measures which are profitable by reducing the energy consumption?
Some other solutions of the tribute consortium are very valuable on the market but don’t change the added value chain and so will be commercialised as usual.
2. New business model in construction
The dream of new construction market is to know before, how much they will consume after. The new business model guaranteed energy result for new construction thanks to progresses realised during TRIBUTE research programme is done for that.
2.1. Principle
The principle is to engage the builder/ contractor to assure the quality of the parameters influencing the whole energy consumption, which are under his responsibility whatever are the parameters which are out of its responsibility.
Example the weather is out of its responsibility, the choice of the power of the coffee machine is out of its responsibility. The ability to cut the power of the coffee machine when nobody is in the coffee room is under his responsibility, the fact that an occupant opens a window because it’s too hot in the office is under its responsibility, the decision of the site manager to close the site between Christmas and new year is not under his responsibility.
So The principle of the contract is to create a model based on common and shared hypothesis of the exogeneous variables (out of his responsibility), with these hypothesis to give an energy objectives for his construction, and to compare, what he built ; So the protocol to assess this contract is using all the different aspects studied in TRIBUTE PROJECT;
2.2. Protocol of assessment
This protocol has been applied between Schneider Electric Real Estate as the owner of a new 11000m² office and research building in Grenoble, and GA the main contractor which has been chosen to build it according a building licence. This building is a part of a project called GreenOValley; in this project, the owner assigned an energy ambition for these buildings which are resumed in the power point created for the press conference of the “first stone event”. These objectives are given in chapter 6 as an annexe.
This protocol has been invented, and created between Schneider Electric Innovation Center (represented by the writer of this Tribute deliverable) and ARTELIA, design institute chosen by the main contractor to study and design the technical parts. As it is an experimentation the financial part of the contract are great enough to be motivated, but small enough to avoid that one of the partner is disadvantaged if the protocol and the engagement were not successful.
Better than explaining all the points of this protocol, he is given here in chapter 7 as an annexe. The reader must take care that the original contract between two French companies is written in French, and here is a “internet automatic” translation which can be sometimes not perfect.
2.3. Result and exploitation of Result
The initial timing of the real estate project was perfectly in accordance with tribute project (building delivery end of 2016, and we should have had 6 months of energy consumption to verify if we are far or close from the objective.
Unfortunately, one of the essential part of the project has been delayed: the energy monitoring part, so we haven’t been able to begin the comparison between measurement and simulation.
Thus, the results are not presented here. During the closing ceremony of the project we’ve presented our intentions:
If the experimentation is successful, our (Artelia, GA and Schneider-electric) intention is to write and publish some papers in scientific publications and/ or in the world of real estate medias. In addition, the results will be presented to the French ministry in charge of the constructive regulation.
If the experimentation is not successful, the analysis of the reasons will be done and the conclusions will help to drive the second building of GreenOValley project which will be delivered end of 2019.
3. New business model in operation
A large range of buildings have “in home facility managers” or “in home property managers” in charge and/or in responsibility of the right operation of the building. The standard of energy management has been done for them to help to drive energy improvement.
They are in front of a lot of suppliers which offer “energy performance contract”, “audit and diagnosis services”.
But they haven’t whether the budget to pay for that, or the envy to let other people to do. The “save it yourself“ offer and business model is done for them
3.1. Principle
The principle of the Save It Yourself, offer and BM is to sell or rent a set of sensors easy to install to be able to monitor a part of the installation that the “in home energy Manager wants to control.
Main advantage is if the instrumentation and monitoring doesn’t’ show great saving potential, they can be displaced to find the next source of savings.
The second part of the business is to supply data collection, data monitoring and analytics under different forms (software as a service, pay per use, or usage sales) to help to identify the ECM.
4. Conclusions on business models
These two new business models have been rendered possible on the market thanks to TRIBUTE key findings.
For the first one it’s too early to know, if the assessment protocol is the right one. We are in the creation process in evaluation on a pilot site. Nevertheless, this need is a huge demand on the market, and as soon as some actors will be able to do, keep and success to the pledge of site energy consumption all end uses, the real estate market will ask for that.
For the second one, it is launched and commercialised since a few weeks. It’s too early to know if the business model of tools for “save It Yourself” will be a great success or not, but it is a new way to analyse the market of energy services face to the “great international actors” which investigate this market with a whole, complete, set of solutions and services to do it by themselves. David or Goliath ?
Give the fish of energy savings or teach how to catch it, that is the question.
The EPC contract is for the fish, the ISO50001 management standards is for the lesson. And the gulplug offer imagined, partially thanks to TRIBUTE project is the net.
We’ll see.
5. Dissemination activities
The TRIBUTE website (http://www.TRIBUTE-fp7.eu) has been operational from the end of January 2014. The webpage is considered as a successful tool for raising awareness of the project and its activities. In total the website had more than 17.000 users with their peak around the time of the TRIBUTE workshop (June 2016), TRIBUTE training (June 2017) and TRIBUTE final conference (September 2017). The website has been maintained during the whole course of the project.
During the project, several dissemination materials were developed including the project leaflet, project roll-up as well as TRIBUTE video. The TRIBUTE video was released in the beginning summarizing the project basic scope and objectives, presenting the project test sitesTRIBUTE also contributed to several editions of EeB project review. The aim of the publication is to provide a dissemination route for EeB PPP projects; capture and showcase the impact of EeB PPP projects; cluster projects by topics to demonstrate continuity of activity; demonstrate the wave-based activity facilitated by the EeB PPP; constitute an effective set of case studies that will encourage further companies to join the association and participate on future EeB PPP calls.
TRIBUTE press release was prepared in the beginning and in the end of the project and shared through partners’ networks. Objective of this press release was to announce the end of the project, to highlight the main achievements and perspectives for future exploitation. In the relation with the TRIBUTE final event, press conference was organized and numerous French newspapers having the highest reading rates, presented articles dedicated to the TRIBUTE project’s scopes and targets.
Clustering activities were planned as a necessary part of the project dissemination. TRIBUTE project participated in several editions of Workshop on the Impact of EeB PPP that was jointly organized by the European Commission with the support of the E2BA. The objectives of the workshops were to address innovation and exploitation issues in running projects and explore potential for cross-project clustering. Several joint workshops were also organised in cooperation with the PERFORMER and ENERGY IN TIME projects. The main goal of these meetings was to discuss and understand the positioning of the projects towards each other (since they have been all funded from the same call topic), to identify common actions and share ideas on possible cooperation. Key common areas of interest were identified, action list proposed and further cooperation between the projects had been foreseen. Common areas of interest were identified as data acquisition, fault detection and dissemination. TRIBUTE had also very strong synergies AMBASSADOR project.
Publication of TRIBUTE results to relevant scientific and industrial periodicals, journals and key conferences in Europe was assured during the whole project lifetime. The project participated with a project booth in conferences such as Industrial Technologies 2014 (April 2014, Athens) or Re-Industrialisation of Europe 2016 (October 2016, Bratislava). Project results were also presented at several editions of Sustainable Places conference (2014, 2015, 2016), where also joint clustering workshops were organised with related projects in EeB domain. Furthermore, the project was represented at ECTP-E2BA Conference 2014 (Brussels, June 2014), Energy Smart Building (Torino, July 2014), L’Energia del Territorio (Torino, March 2015), TRIBUTE presentation during EQUA users group in Switzerland (March 2015), IBPSA 2015 (Hyderabad, December 2015), France-Sweden Workshop for Sustainable Cities (La Rochelle, May 2015), PROBIS EU PROJECT Event (Turin, July 2015), Climamed conference (Juan les Pins, September 2015), TURIN ENERGY MANAGEMENT EVENT (Turin, September 2015), AICARR BOLOGNA (Bologna, October 2015), SBE16 (Turin, February 2016), IRES Conference 2016 (Düsseldorf, March 2016), IntelliSys - 25th SAI Intelligent Systems Conference (London, September 2016), CESB 2016 (Prague, June 2016), IBPSA 2017 (San Francisco, August 2017), 12th IEA Heat Pump Conference (Rotterdam, April 2017), PETRA 2017 (Island of Rhodes, 2017).
Peer reviewed articles were prepared for Building and Environment journal, Applied Thermal Engineering journal and Applied Energy journal. In addition, peer reviewed articles were prepared for SAI Intelligent Systems Conference (IntelliSys) 2016, Building simulation 2015 conference, 10th International Renewable Energy Storage conference 2016, 2nd ACM International Conference on Embedded Systems for Energy-Efficient Built Environments – BuildSys 2015, 12th IEA heat pump conference 2017, as well as 10th International Conference on PErvasive Technologies Related to Assistive Environments - PETRA 2017 and are available in their proceedings. AMBASSADOR project was also presented in the VTT Research Highlights publication.
TRIBUTE results were also disseminated through organization of TRIBUTE events, especially TRIBUTE dedicated workshop for energy managers that took place in the frame of Sustainable places 2016 conference in Biarritz, France in June 2016. TRIBUTE training for energy managers that took place in June 2017 in Grenoble, France in the premises of the Schneider Electric was dedicated to topics such as overall energy context, optimization strategies for buildings and closing the gap between the reality and simulation.
At the end of the project the TRIBUTE final conference took place (La Rochelle, France, September 2017). During the event the project was revealed to the public and the results were made well known to the audience. This was a very important action for the recognition of the project to the public and for giving the importance of project’s targets regarding the sustainable energy management in building level. The representatives from the whole project’s consortium were giving all the necessary information and gladly answering all the questions given from the public. The conference attracted a wide range of stakeholders from the whole value chain, such as representatives from the industrial sector and the public authorities, engineers from different field as well as students and private investors. Additionally, national media, having high broadcasted rates, covered the final conference of the TRIBUTE project and make it well known to the national audience.
6. Exploitation activities
TRIBUTE exploitation activities were designed in order to evaluate the collective impact potential of the consortium by evaluating the market potential and to determine product opportunities in relation to the customer/product requirements throughout the course of the project. The activities covered the methodology developed by the consortium to complete a market description of buildings, to identify the different actors of the added value chains and their behaviours and expectations in the field of value of TRIBUTE. The exploitation plan was focused on different business cases depending on the organisation of actors and their territory of responsibility. For the behavioural typology of actors in exploitation market, the analogy is done selecting situations where the different types belong to the same organisation, with a common management or not. According to this definition, we meet during our survey 7 main business cases.
The main actors to be targeted for energy optimisations at building level or at district level are the Energy Manager persona actors.
The easiest situation is when they belong to the same organisation as the owner actors’ persona (business cases number 1 and 2). In the same decision process can be balanced investment costs and energy bill reductions and all the business models can be easily developed.
The second situation is when the energy manager belongs to a tenant organisation with a long term contract (business cases number 4 and 5). In that case the balance between investments costs and energy bill reduction can be argued if and only if the payback is shorter than the duration of the lease.
A special situation is when the site is managed by an external ESCO in charge of FM and EM through an Energy Performance Contract (EPC) as shown in business case number 3. The target becomes the ESCO itself, if the added value brought can serve, optimise or simplify the execution of the contract.
The last positive situation is described by business case number 6 where the EM belongs to a short term lease, but in this case only business models without customers’ investments (service provider or energy optimiser) can be developed actively on the market.
In all these situations, the FM can be a brake, when he is not in the same organisation than the EM, as solutions of energy optimisation can modify the conditions of execution of his own objectives.
In all these situations, the occupant can be a positive or negative influence maker, depending on the effect on the quality of building services, or on his perception of this effect.
Finally, the risky situation is when all the stakeholders have different objectives (business cases number 7). This can be met when they belong to the same organisation, but a structured one, with a segmentation of different services, departments or divisions, or when they belong to completely different organisations. In this situation, it will be very difficult whatever the business model is to promote energy optimisations.
List of Websites:
Webpage
• http://ambassador-fp7.eu/
Project coordinator
• Mr. Martin Sénéclauze (CSEM Centre Suisse d'Electronique et de Microtechnique SA - Recherche et Developpemnent)
• E-mail: martin.seneclauze@csem.ch
• Tel: +41 32 720 5340
Technical coordinator
• Mr. Henri Obara (Schneider Electric Industries SAS)
• E-mail: henri2.obara@fr.schneider-electric.com
• Tel: +33 (0) 476 57 65 12
Project manager / Dissemination manager
• Dr. Václav Smítka (Amires s.r.o.)
• E-mail: smitka@amires.eu
• Tel: +420 732 304 379
Exploitation manager
• Mr. Olivier Cottet (Schneider Electric Industries SAS)
• E-mail: olivier.cottet@schneider-electric.com
• Tel: +33 (0) 476 39 11 48