A comparison between Gaussian Process emulation and Genetic Algorithms for optimising energy use of buildings
Eames, ME; Wood, M; Challenor, Peter G.
Date: 1 December 2015
Conference proceedings
Publisher
-
Abstract
Computing speed has increased greatly over recent years. Building designers can now simulate complex building models in a short time. However, even with short simulation times, building optimisation routines can still take too long for some applications. In this paper, we compare how well genetic algorithms (GAs) and Gaussian process ...
Computing speed has increased greatly over recent years. Building designers can now simulate complex building models in a short time. However, even with short simulation times, building optimisation routines can still take too long for some applications. In this paper, we compare how well genetic algorithms (GAs) and Gaussian process emulation with sequential optimisation (GPESO) optimise a building to minimise the energy use. The GA approach performs a GA routine on an EnergyPlus model and the GPESO technique creates a Gaussian Process emulator (GPE) also based on the EnergyPlus model. The GPESO uses an expected improvement algorithm to sequentially improve the GPE. The results show that the GPESO technique outperforms the GA in terms of minimising the number of simulations required and the solution obtained.
Physics and Astronomy
Faculty of Environment, Science and Economy
Item views 0
Full item downloads 0