Comparison between two genetic algorithms minimizing carbon footprint of energy and materials in a residential building

Abstract

The emergence of building performance optimization is recognized as a way to achieve sustainable building designs. In this paper, the problem consists in minimizing simultaneously the emissions of greenhouse gases (GHG) related to building energy consumption and those related to building materials. This multi-objective optimization problem involves variables with different hierarchical levels, i.e. variables that can become obsolete depending on the value of the other variables. To solve it, NSGA-II is compared with an algorithm designed specifically to deal with hierarchical variables, namely sNSGA.Evaluation metrics such as convergence, diversity and hypervolume show that both algorithms handle hierarchical variables, but the analysis of the Pareto front confirms that in the present case, NSGA-II is better to identify optimal solutions than sNSGA. All the optimal solutions are made of buildings with wooden envelopes and relied either on heat pumps or on electrical heaters for proving heating.

Effectiveness of standalone simulation-based optimization software in optimizing the life cycle cost of residential buildings

“…Fonseca et al [25] proposed a methodology to evaluate building energy and comfort performance while considering life cycle cost. Gagnon et al [26] developed an integrated system to compare an NSGA II optimization and a hierarchical optimization technique performance in minimizing the carbon footprint of building energy consumption and materials. The results of the study indicated that the NSGA II optimization methodology is better in identifying the optimal designs.…”

BIMBOT-(Artificial intelligence applied to BIM design)

“…Likewise, the most common evolutionary search techniques have demonstrated their limitations in certain design problems within the field of architecture (Gagnon, Gosselin, Park, Stratbücker and Decker, 2019) (Wortmann, Waibel, Nannicini, Evins, Schroepfer and Carmeliet, 2017). Therefore, more sophisticated heuristic optimisation techniques are bound to become more viable alternatives in the future (LaTorre, Muelas and Peña, 2015).…”

“…Un aspecto importante a considerar es que los mecanismos de PCG que usan técnicas de búsqueda heurística (por ejemplo algoritmos evolutivos) tienden a requerir un número de evaluaciones de soluciones tentativos que puede ser medianamente alto, para ello se sugiere usar técnicas de surrogados que agilicen el proceso de evaluación descartando soluciones potencialmente poco interesantes antes de evaluarlas (Karavolos, Liapis and Yannakakis, 2009: 1). Asimismo, las técnicas de búsqueda evolutiva más habituales han demostrado sus limitaciones en determinados problemas de diseño dentro del campo de la arquitectura (Gagnon, Gosselin, Park, Stratbücker and Decker, 2019) (Wortmann, Waibel, Nannicini, Evins, Schroepfer and Carmeliet, 2017). Es por ello que técnicas más sofisticadas de optimización heurística están llamadas a ser alternativas más viables en el futuro (LaTorre, Muelas and Peña, 2015).…”

Optimization of building fenestration and shading for climate-based daylight performance using the coupled genetic algorithm and simulated annealing optimization methods

“…There have been different modifications applied in the GA to resolve the shortcomings in the GA. Examples of these modifications include, Micro-GA, which uses a smaller population size and uses reinitialization, 36 and different variations of the Non-dominated Sorting Genetic Algorithm (NSGA), including NSGA and NSGA-II, which applies non-dominated sorting strategy and niching, 37 structured NSGA (sNSGA), designed to deal with hierarchical variables, 38 and Elitist Non-dominated Sorting Evolution Strategy, a two-phase optimization using the GA (PR-GA), and the epsilon dominance multi-objective evolutionary algorithm. 39 While there have been improvements in specific aspects of the GA using the modifications, there still needs further investigation in this area.…”

“…In a comparative study of GA and micro-GA by Gagne and Andersen, 17 the tendency of the micro-GA to trap in local optimum and the inconsistency in results of the GA are mentioned as shortcomings of the optimization methods whereas the latter can be overcome with multiple runs. In a comparative analysis of the NSGA-II and sNSGA by Gagnon et al., 38 dual mutation rate, decreasing during the process and allowing to mate high-level variables are introduced as the potential modifications to sNSGA to improve its performance.…”