Assessing the suitability of genetic algorithms for application in building envelope optimization

In order to reduce building energy consumption most effectively, heating and cooling loads due to the building envelope must be addressed early in the design process. Minimizing energy use is certainly one goal, but a more valuable approach would factor in economic considerations, such as the first cost of building components as well as energy costs (i.e. life-cycle cost). Traditionally, this type of analysis has been performed with parametric runs using a building energy simulation tool such as DOE-2 or EnergyPlus. The problem is better served, however, by using optimization, which can capture interactions that parametric analysis might miss, and is also more efficient than performing a full enumeration of the search space.;A simulation/optimization tool has been developed (using Matlab and Perl) that couples a genetic algorithm to a building energy simulation engine (DOE-2). Genetic algorithms (GAs) are global optimization methods that use the evolutionary concept of natural selection in order to converge on a solution. They have been used successfully in many optimization applications. For validation and comparison purposes, the GA was compared to the sequential search technique used in BEopt, a building energy optimization tool developed at the National Renewable Energy Laboratory. Different population sizes and mutation rates were investigated for the GA. For relatively large search spaces, it was found that the GA could identify the minimum cost point to within 0.4% difference from the optimum, using around 60% of the simulations required by sequential search. A population size of 32 and mutation rate of 1% generally worked well.;Different building shapes were investigated as part of the envelope optimization, including rectangle, L, T, cross, U, H, and trapezoid. The rectangle and trapezoid consistently had the best performance (lowest life-cycle cost) across five different climates, and also exhibited the least variability from best to worst within the shape.;The simulation/optimization tool can also be used with constraints in order to find the optimum building (lowest energy costs) for a given construction budget, or to find a building with a certain target energy savings relative to a reference building.