“Développement de solutions d'information et d'optimisation de la qualité de l'air intérieur renseignée en temps réel”

par Thi-hao Nguyen

Projet de thèse en Sciences de l'ingénieur

Sous la direction de Evelyne Gehin et de Anda Ionescu.

Thèses en préparation à Paris Est , dans le cadre de École doctorale Sciences, Ingénierie et Environnement (Champs-sur-Marne, Seine-et-Marne ; 2015-....) , en partenariat avec CERTES - Centre d'Etude et de Recherche en Thermique, Environnement et Systèmes (laboratoire) depuis le 01-10-2018 .


  • Résumé

    Le monde du bâtiment vient de commencer sa révolution numérique, dans lequel les matériaux et équipements deviennent communicants par l'intégration de métadonnées dès la conception. Par ailleurs, les capteurs et objets connectés se multiplient et permettent le développement de services dédiés à l'amélioration du confort des occupants et l'optimisation de leurs usages. Le traitement de ces données collectées en continu en est encore à ses débuts et reste essentiellement tourné vers une exploitation monodimensionnelle des données. Pour autant, être capable de fournir un service d'information sur le niveau de confort et de qualité de l'air à un instant donné dans une pièce définie, nécessite une approche à la fois multidimensionnelle et agrégée des données fournies par des capteurs, en tenant compte de leur évolution temporelle. Parfois, les indicateurs mesurés, comme le formaldéhyde (dû aux sources intérieures au bâtiment) et les particules (provenant des sources extérieures et intérieures au bâtiment), fournissent une information contradictoire en termes de qualité de l'air intérieur. L'enjeu est donc de fournir une information cohérente en termes de qualité de l'air d'un point de vue de l'occupant de la pièce et des solutions d'optimisation d'action à mettre en place à l'échelle de l'ensemble des occupants du bâtiment.

  • Titre traduit

    Development of information and optimisation solutions for real-time monitoring of indoor air quality


  • Résumé

    Scientific Issues: The digital revolution is reshaping the building environment, where communication between materials and devices becomes possible with specific metadata integrated at the design stage of the building. Furthermore, lots of sensors and connected objects are available and with them the development of services dedicated to comfort improvement of occupants and their well-being. Meanwhile, the processing of these continuously collected data is still at its beginning and usually limited to simple data treatment. However, to be able to provide reliable information on the comfort level and the air quality at a given moment in a given room, requires a multidimensional approach of the data provided by the sensors altogether, by taking into account their temporal evolution. Sometimes, the monitored indices such as formaldehyde (mostly generated by the sources inside the building) and the particulate matter (originating from sources which can be located both inside or outside the building) produce the risk of a contradictory information concerning the air quality. Thus, the challenge is to supply coherent information on air quality from the room occupant point of view and to propose optimal solutions to be implemented at the scale of the whole building occupants. Objective: The objective is to develop a novel approach of data processing combining both the classical and robust methods of data mining with the mathematical methods specific to time series analysis and to signal processing, in order to explain the fluctuations of the underlying processes and their causes, and thus, be able to better anticipate their consequences. These methods should permit 1) to provide pertinent information in terms of air quality exposure and 2) to optimise the action to implement in order to reduce the occupants' exposure. These solutions of information processing and action optimisation have to be translated in algorithms implemented in web-services. This approach will be developed using real- time data monitored in a tertiary environment during several years, with a high time resolution (one or several minutes). Approach: The approach is based on inverse modelling based on the recordings of different parameters characterising indoor and outdoor air quality, climatic indoor and outdoor recordings and specific uses: human presence (occupation) and state of opening elements (opened or closed windows or doors), in order to extract the pertinent information of the underlying process(es). To do so, several tracks can be explored, namely the Blind Source Separation Methods (BSS) based on matrix factorization under constraints and using, in addition to air quality data, some exogenous variables having a potential impact on it. The obtained results will be analysed with regard to the data structure characteristics. According to the methods' robustness, their generalisation capacity will be evaluated. The analysis of typologies obtained by different classification methods can be an alternative to highlight the clustering effects corresponding to the most representative situations. The most appropriate methods and parameters will be than selected in order to automate the procedure of information analysis and recommendation of the most suitable action. Expected Results: The solutions to be developed will be based essentially on the information processing concerning Indoor Air Quality. Nevertheless, they can represent a basis for the development in other final purposes, such as the comfort. The results of this thesis will allow 1) to adapt these methods for web-services dedicated to health and comfort, 2) to develop optimisation methods necessary to process contradictory information on air quality or comfort and 3) to integrate the Indoor Air Quality dimension in the concept of Smart Building.