Combination of the genetic algorithm with the Latin Hypercube sampling process to investigate the robustness of a building optimization process.

The objective of this study is to clarify the influence of the occupant behavior on the robustness of the results of a computer-based building optimization process.

A clear tendency toward a parameter setting or a range of parameter settings is observed for all parameters to be optimized and the results are robust.

For linear parameters, the mean value of the recommendation frequency of the parameter settings is the best indicator for the recommendation process, and the variance is shown to be a good indicator for its robustness.

For non-linear parameter in the search space, the most recommended parameter setting in a sampling run is the preferred indicator rather than the mean value.