Action Filename Description Size Access License Resource Version
Show more files...


The building sector alone accounts for around half of the energy consumed in Switzerland and most other developed countries, with associated adverse environmental consequences, and there is a great potential for savings in this sector. For this reason, the development of efficient solutions for predicting and optimising the energy and environmental performance of buildings is clear. Dynamic building thermal simulation programs are increasingly used for this purpose. However, some key processes are still not taken into account by these tools, leading to potentially significant errors. Most noteworthy is the influence of buildings' occupants, whose actions such as the use of windows and shading devices have an important impact on the indoor environment and the overall energy performance of a building. Furthermore, occupants' environmental comfort is the central underlying concept influencing actions on building controls; but the intrinsic interaction between these notions is not well known. This thesis develops adequate models for the prediction of occupants' actions that have an impact on building performance and further proposes an innovative global formulation of the link between environmental comfort, human adaptive actions in the built environment and their feedback in terms of satisfaction and acceptability. Furthermore, detailed integration procedures of these methods into building and urban simulation tools are described. Based on detailed statistical analyses of eight years of continuous measurements, a model for the prediction of actions on windows performed by office occupants is proposed. It is formulated as an occupancy-dependent Markov chain extended to a continuous-time process for opening durations. The explanatory variables have been carefully selected on the basis of statistical relevance, which are indoor and outdoor temperature, the occurrence of rain, and occupant presence and absence durations. The choice of the specific form of the model is justified by cross-validation and its superior predictive accuracy is determined by comparison with model variants and previously published work. A similar procedure was carried out for the inference of a model to predict actions on shading devices. Its formulation is also based on rigorously selected predictors used as inputs to an occupancy-dependent Markov chain expressing action probabilities. The model has also been extended to predict the choice of shaded fraction. Once again simulations of model variants support the choice of the final model. Using results of a long-term survey of building occupants, we evaluate the accuracy of currently accepted models for thermal comfort prediction and identify clear weaknesses. We go on to propose a probabilistic formulation for the distribution of thermal sensation and for the occurrence of the state of thermal comfort and extend this to visual comfort. The result is a simple and accurate definition of comfort probability and its variations amongst individuals. We have also analysed variables which influence occupants' comfort temperature. This has enabled us to assign weights to the key variables influencing comfort temperature: adaptation, acclimatisation and individuality. We also consider the feedback of actions on comfort and numerically estimate "adaptive increments to comfort temperature". This results in a proposed formulation for a new adaptive model for thermal comfort, for general application in buildings with variable degrees of adaptation available to occupants. The link between thermal and visual comfort with actions on windows and shading devices is also studied and formulated as a single unified concept linked by human action inertia whose properties are discussed. Finally, new modelling approaches have been developed for the prediction of adaptations of personal characteristics such as clothing and metabolic activity, an assessment of the very limited degree of interaction between thermal, olfactory and visual comfort and finally an analysis of factors influencing perceived productivity in office environments, in which hot conditions are shown to cause a decrease of the order of 10% compared to relatively cooler conditions.