In order to simulate the energy performance of low energy buildings more precisely, human behaviour can be modeled through multi-agent systems and coupled to dynamical building simulation. Exploiting electrical load profiles or other measured data allows for identification of simplified models of buildings or appliances, and, in return, for better sizing and control of the building systems and appliances.
The SUPERBAT project aims to develop new modeling approaches, by taking into account stochastic occupant behaviour in deterministic building energy simulation tools, and using measured data to improve models. The are two main objectives: a better power demand prediction, which in return allows for a better estimation of the annual energy consumption. These models will help us to rethink the building integration into its electrical grid, in order to improve energy efficienty of the global energy system. Furthermore, this work is essential for low energy or positive energy buildings. Low energy buildings have very low power profiles for heating demand, and are therefore very sensitive to internal gains due to occupants and specific electricity uses. The mere presence of occupants, and their activity, can produce enough heat to maintain a correct room temperature, or even induce overheating, which is a source of discomfort. Until now, simulation tools were mainly used for sizing, but there is an increased need in better control, to reach energy sobriety. In this context, simulation tools need a more precise modeling of power demand and global energy behaviour of buildings, for all uses and specific electricity uses in particular.
Occupant behaviour is modeled through agent-based systems. They describe occupants as partially autonomous entities, capable of interacting with their environment (building and its appliances), and with the other occupants, depending on the agent's priorities and particular rules. Building thermal models are based on a classical approach of building simulation: a room is an air node with a homogeneous temperature, and the walls are discretized in order to correctly describe heat exchanges. Energy systems are often modeled in a simplified way with laws which allow for their control. Coupling these two universes is considered through integration in a common tool, or through an interoperability standard for model exchange or co-simulation like the FMI standard (Functional Mock-up Interface). In terms of inverse problematic, several methods for identifying parameters of building dynamic thermal models (gains, time constants) can be used, depending on the amount and quality of available data, and on the objective (energy consumption drift detection, optimal energy management, gain of a refurbishment operation ...).
House energy appliances and thus specific electricity uses have been integrated in the multi-agent simulation tools. Then, a whole building dynamical thermal model has been coupled to the human behaviour simulator by direct co-simulation. This allowed us to integrate occupant actions on heating devices and on building envelope components (windows) in the building energy simulation. Regarding the inverse problem, the structuring parameters to take into account when choosing a model identification method have been identified. These parameters depend on whether control is taken into account, and on the richness and availability of data, as well as the models themselves. The choice depends on the objective of the simulation (energy consumption drift detection, optimal energy management, gain of a refurbishment operation ...). The identification of specific uses of electricity is complementary with building thermal models identification. These works on the load curve have brought a set of methods allowing to identify load states, or to tacle predictive control.
Significant results and ongoing activities include the effective achivement of the coupling between a simulation tool for human behaviour modelling and a dynamic building thermal model. This coupling reveals many perspectives for direct simulation of occupants in buildings, or creation of specific occupancy and uses scenarios much closer to reality.
The works on model identification led to establishing the limits of the various building model identification methods, depending on the richness and availability of data, and the study objective (drift detection, predictive control, non-intrusive characterisation, recommendation of energy efficiency solutions ...)
The project has produced many papers in journals, and national and international congresses on several topics (human behaviour, artificial intelligence, buildings physics, modelling, electric appliances). The works on coupling between dynamic building ene
The fulfilment of the « Grenelle de l’Environnement » is the long-awaited break-up in the field of energy in buildings. Starting from now the energy performance of buildings become a major concern. The will of the « Grenelle de l’Environnement » is to tool up the building sectors in order to design efficiently low energy and positive-energy building. This social and economical break-up is based on a increasing need to modernise the building industry. The energy simulation tools are now especially used only for the design of a new buildings. Impressive progress in the field of energy savings will be accomplished if simulation tools will be used also for the building commissioning. Building simulation programs offer an effective support for the energy management devices. For this kind of applications the simulation tool needs more accurate modelling skills in order to predict the power load of the building including the electricity uses. Questions arise over the ability of the existent simulation tools to predict et manage the power load of the buildings. The power load of the buildings is strongly related to the climate variations, building envelop, HVAC system and occupant behaviours.
Based on physical and sociological competences, the proposed project will try to go beyond the most ardent limitations of the existing dynamic simulation tool for the energy behaviour of the buildings :
- the most part of the simulation tools are designed to predict hourly (more or less) precisely the energy consumption of the building. This approach is inaccurate in order to estimate the power load of the building. Time step above 1 and 10 minutes are needed for all the applications related to the energy management and power load estimation.
- the occupant has a major impact on the energy consumption of the building and especially on the use of electricity uses. The electricity uses are commonly modelled by using standardized occupancy profiles. This profiles are hourly based and conventionally chosen for each thermal zone.
- the simulation tools provide an acceptable (hourly based) modelling of the classical energy uses (heating, ventilation, hot water, cooling) but the modelling is deficient concerning the electricity uses. The electricity use in the recent buildings (low energy and positive-energy building) represents half of the energy consumption.
- the physical modelling of the building are not adjusted or fitted to real time measurements of the power load (or other physical measurements such as internal and external temperatures, solar irradiation, …) in order to capture the caracteristics of the real world.
The developments of the stochastic modelling for the occupancy and energy uses integrated to the dynamic simulation tools and the development of the modelling adjusted to the real life measurements (power load, temperatures, …) will allow to study industrial applications such as the design and energy management of low energy and positive-energy building.
Monsieur Mathieu Schumann (EDF RECHERCHE ET DEVELOPPEMENT) – email@example.com
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
LMT ECOLE NORMALE SUPERIEURE DE CACHAN
GSCOP INSTITUT NATIONAL POLYTECHNIQUE DE GRENOBLE - INPG
UJF - GE2LAB UNIVERSITE GRENOBLE I [Joseph Fourier]
INSA - CETHIL INSTITUT NATIONAL DES SCIENCES APPLIQUEES DE LYON - INSA
EDF R&D / EnerBAT EDF RECHERCHE ET DEVELOPPEMENT
CSTB CENTRE SCIENTIFIQUE ET TECHNIQUE DU BATIMENT
Help of the ANR 905,421 euros
Beginning and duration of the scientific project: - 48 Months