Multi objective optimization of detailed building models with Typical Short Sequences considering sequential and adaptive methods.

Multi-objective optimization of detailed building models may lead to high computational time expenses. The work presented in this paper presents an alternative of classical approaches based on the usage of reduced sequences in a sequential and an adaptive approach. The sequential approach runs directly the reduced sequence obtained from a typical day selection algorithm which employs a k- medoids clustering algorithm in a multi-objective optimization study. It was applied on a solar combisystem connected to a building model and showed its capability of reproducing the Pareto front 25 times faster with relative errors not exceeding 5% when considering several parametric configurations of the model and focusing on a limited number of selection criteria during the day selection process. On the other hand, the adaptive approach for optimization named OptiTypSS achieves day selection during the optimization process by combining an iterative day selection algorithm (TypSS) with a genetic algorithm of optimization (NSGA-II). The method succeeded in obtaining results with errors inferior to 3% being therefore as efficient as a classical metamodel adaptive approach yet with relatively higher computational time. Further improvements including parallel simulation during the day selection process can be done to enhance the performance. • Optimization of a solar combi system model using short simulation sequences in a sequential and an adaptive approach. • In the sequential approach, the optimization algorithm converged 25 times faster with relative errors of a maximum of 5%. • In the adaptive approach, the Pareto front superposed the reference one recording relative errors inferior to 2.5%. • Results obtained by the adaptive approach were as efficient as using metamodels with relatively higher computational time. [ABSTRACT FROM AUTHOR]