Multi-criteria decision support methodology for energy renovation process of residential buildings through a systemic approach

We spend most of our lives in buildings. These were built mostly before the oil shocks of 1974 and 1979 and offer both poor energy performances and improvable comfort conditions (thermal, acoustic, natural lighting). In France, with 32 million units and an annual turnover of less than 1% of the existing building stock, the energy renovation of housing becomes a necessity. This, as much for political ends (energy dependence), economic reasons (revival of construction jobs, real-estate valorisation), social reasons (well-being of occupants), as environmental reasons (reduction of greenhouse gas emissions). This thesis aims to provide knowledge and a methodology to contribute to the decision support for prescribing efficient energy renovation scenarios of houses built during the 1945- 1974 period. Three issues are highlighted in this research work: 1) the integration of an holistic approach of renovation process (systemic and multi-criteria) to avoid not foreseen collateral effects due to bad choices; 2) help formalize the preferences of decision-takers (homeowners) in a format interpretable by multi-criteria analysis tools; 3) integration of uncertainties related to the characterization of existing buildings in the process of scenarios generation and decision support. Through a systemic description of buildings and a multi-criteria performance assessment of formerly identified renovation actions, we propose an innovative methodology, consisted of 6 modular and interchangeable technical sub-models, which aims to automate the generation, assessment, optimization and performance-based ranking of renovation scenarios. The heart of the methodology is based on the formalization of renovation knowledge from construction specialists in two of the six sub-models. The first one is an influence matrix that we use to translate most common renovation goals (equivalent to the wishes expressed by homeowners) into a profile of relative weights and a profile of targeted-levels of performance on indicators modelled. The second one is a probabilistic inference tool (using the technology of bayesian networks) to both optimize assemblies of renovation actions (programming by successive constraints) and achieve multi-criteria evaluation of these assemblies (by the use of aggregation functions of local performances). A sixth and final sub-model uses the ELECTRE outranking methods to sort and classify, by preference order, renovation scenarios previously generated. At last, our methodology provides the ability to let users test their own energy renovation scenarios in order to analyse their multi-criteria performances and compatibility with the characteristics of the existing capitalized during the technical diagnosis of their building. The methodology proposed is intended to be educational and transposable into a functional computer system prototype. A first version was developed and used to apply our decision process to a real case of individual house to renovate. First results obtained are consistent and allow validating the approach. However, keep in mind that like any model using expert knowledge, robustness and validity limit of its scope of application depend on the quality of knowledge work capitalized. (author)