The efficient thermal control of buildings is a complex problem. Non controlled perturbations like user behaviour and meteorological conditions can change much with time and make this task difficult. This work proposes considering the building as a 'living system', with an internal activity (user behaviour and internal gains) and placed in an influencing environment (the solar radiation and the external temperature). As a living system, the building must provide thermal comfort conditions to its inhabitants and minimize its energy consumption. Strategies taken from the behavioural thermo regulation of the human body, as well as biomimetic technologies like artificial neural networks and fuzzy logic (technologies that imitate the behaviour of living systems), were used to reach these objectives. An optimal heating controller is proposed. It uses meteorological short term prediction models (for solar radiation and external temperature), as well as a model to describe the thermal evolution of a building. These models are based on artificial neural networks and have the following objectives: Optimize the prediction performance of adaptive meteorological models based on a limited number of local measurements. Precisely describe the thermal behaviour of buildings including the non-linearities introduced by inhabitants, with a self-adaptive model. Limit as much as possible the commissioning procedure necessary to initialize these models. A fuzzy blind controller is also developed. It can manage the sometimes contradictory goals of blinds: Provide optimal visual comfort for the inhabitants; Optimize the thermal exchanges through the window; Satisfy user wishes. To evaluate and compare different heating and blind control strategies, a correlation method is proposed. It evaluates the sensibility to solar and heating gains of a heated and cooled building. The proposed controllers have been tested in two rooms of a non residential building during three years. These rooms have large solar gains (south orientation and a high window fraction) and a massive structure. Many numerical simulations complete the experimental results. When compared to a conventional but efficient heat controller, equipped with internal temperature, external temperature and solar sensors, self adapting heating curve algorithms and optimal start/stop procedures, the biomimetic controller: reduce thermal energy needs (13% for the heating season); optimize the inhabitants' thermal comfort; make easier the commissioning procedure. When the blind controller is added, simulations show an extra 20% of energy savings (when compared to the case with only the biomimetic heating). This results from a reduced global heat transfer coefficient (due to blind's night closing) and an optimization of the solar gains use. This work shows that the heating consumption of buildings depends largely on the way they are controlled. The energy savings obtained with an efficient control is comparable to those resulting of an improvement of the building's envelope. For example, a window area with an automatic blind control can be a heating source (with the solar gains in a winter period) or a cooling source (with passive cooling in a summer period). Appropriate control can use this property to decrease the building's energy consumption.