Identification of the optimal glass type for buildings

AGC Glass Europe has a tool named 'Building Energy Efficiency Calculator' (BEE) that permits to determine the impact of the glass on a building in term of thermal insulation, natural light and solar protection. The goal of this master thesis is to develop an optimization algorithm in order to determine the optimal glass for a specific building. First, an analysis of the glass domain and a study of a database of glasses were done to create a convex hull containing the entire glazing catalog. We have developed derivative-free black-box optimization algorithms as there is no analytic formulation of the objective function. The objective function value is a result from a simulation generated by EnergyPlus, a full-year energy simulation for buildings. The methods are deterministic algorithm such as Nelder-Mead, evolution strategies as Covariance Matrix Adaptation (CMA-ES), Particle Swarm Optimization (PSO), Differential Evolution (DE) and surrogate model as Regression model, Radial Basis Function Network (RBF), Kriging model. An improvement of algorithms has been made with the projection on the database, in particular on the convex hull, by quadratic programming. The study has two kinds of problem : either the case of one glass for the building or the case of multi-glasses. The user of this tool can add some constraints to the problem : with constraints related to variables, with aesthetic constraints as the color of the glass and the price constraint. This tool will allow to give easier advises for employees for a type of glass for a building. Before, their only tool is to determine the glazing according to their experience and to verify, through a simulation 'EnergyPlus', the energetic impact of the building. At present, with the optimization tool, they will directly be able to ask for a specific problem which glass is optimal.