The aim of this thesis is to investigate the direction of innovation in the building energy efficiency sector by combining empirical and semantic analysis of patents, experimental economics and prospective modeling. First, the statistical use of Patstat, a patent database, makes it possible to highlight the flow of knowledge in energy efficiency technologies. The intensity of innovation is studied in view of the competitive structure of the market in order to assess the propensity of the most important players to maintain a status quo in the technological offer. The second part specifies the flow of knowledge by differentiating the technological characteristics thanks to the latent semantic analysis of patents. Thus, the regularity of trends in know-how offers is an essential indication of the "demand pull" or "technology push" feature of the sector. In a third part, the analysis of the "demand pull" feature is strengthen by the balance of the characteristics offered with the preferences of households in a choice experiment survey. Finally, in a last part, household preferences are introduced into the MENFIS prospective model in order to redefine well-being with non-financial attributes, and to improve the simulations of directed innovation or subsidy policies as MaPrimeRénov ' .