Applying Global And Local SA In Identification Of Variables Importance With The Use Of Multi-Objective Optimization

Citation

Wang M, Wright JA, Brownlee A & Buswell R (2014) Applying Global And Local SA In Identification Of Variables Importance With The Use Of Multi-Objective Optimization. In: Malki-Epsthein L, Spataru C, Halburd L & Mumovic D (eds.) Proceedings of the Building Simulation and Optimization Conference 2014. Building Simulation and Optimization 2014, London, UK, 23.06.2014-24.06.2014. London: The Bartlett, UCL Faculty of the Built Environment. http://www.bso14.org/BSO14_Papers/BSO14_Paper_096.pdf

Abstract
Methods for global and local Sensitivity analysis are designed to identify and rank variables importance for each design objective and constraint. This paper investigates the application of local sensitivity analysis to a set of Pareto optimum solutions resulting from the multi-objective minimization of energy use and capital cost, with occupant thermal comfort acting as a constraint. It is concluded that the local sensitivities vary along the trade-off and that these sensitivities are different to the global sensitivities. Different sensitivity behaviour is also observed both along the Pareto trade-off and between variables.

People

Dr Sandy Brownlee

Senior Lecturer in Computing Science, Computing Science and Mathematics - Division