Elaboration of the decision space for an optimization of building retrofit

Multiobjective optimization is widely used in building physics but it has to face construction and regulation constraints to elaborate feasible solutions. This paper investigate how to integrate constraints in genetic optimization carried out with NSGA2 algorithm: it studies the implementation of constraints in the decision space and its impact on convergence speed and diversity in the optimization. The study was carried out on a 8 appartments building with three objective function to optimize: energy demand, comfort of the tenants and economic cost. As a result, convergence speed was improved and expert knowledge was included inside the decision space in a comprehensive way.