Performance of Neural Networks vs. Radial Basis Functions When Forming a Metamodel for Residential Buildings

Average temperatures worldwide are expected to continue to rise. At the same time, major cities in developing countries are becoming increasingly populated and polluted. Governments are tasked with the problem of overheating and air quality in residential buildings. This paper presents the development of a model, which is able to estimate the occupant exposure to extreme temperatures and high air pollution within domestic buildings. Building physics simulations were performed using the EnergyPlus building physics software. An accurate metamodel is then formed by randomly sampling building input parameters and training on the outputs of EnergyPlus simulations. Metamodels are used to vastly reduce the amount of computation time required when performing optimisation and sensitivity analyses. Neural Networks (NNs) have been compared to a Radial Basis Function (RBF) algorithm when forming a metamodel. These techniques were implemented using the PyBrain and scikit-learn python libraries, respectively. NNs are shown to perform around 15% better than RBFs when estimating overheating and air pollution metrics modelled by EnergyPlus.

Authors:

• Philip Symonds
• Jon Taylor
• Zaid Chalabi
• Michael Davies

Keywords:

• Neural Networks
• Radial Basis Functions
• Metamodelling
• Python machine learning libraries.

References:

[1] D. J. Rowlands et al. Broad range of 2050 warming from an
observationally constrained large climate model ensemble. Nature
Geoscience, 5:256–260, 03/2012 2012.
[2] S. Hajat, S. Vardoulakis, C. Heaviside, and B. Eggen. Climate change
effects on human health: projections of temperature-related mortality for
the uk during the 2020s, 2050s and 2080s. Journal of Epidemiology and
Community Health, 2014.
[3] World Health Organisation. Ambient (outdoor) air pollution in cities
database 2014.
[4] A. Mavrogianni, P. Wilkinson, M. Davies, P. Biddulph, and
E. Oikonomou. Building characteristics as determinants of propensity
to high indoor summer temperatures in London dwellings. Building and
Environment, (55):117–30, 2012.
[5] J. Taylor, A. Mavrogianni, M. Davies, P. Das, C. Shrubsole, and
P. Biddulph. Understanding and mitigating overheating and indoor
PM2.5 risks using coupled temperature and indoor air quality
models. Building Services Engineering Research and Technology,
(0143624414566474), 2015. [6] A. Mavrogianni, M. Davies, J. Taylor, Z. Chalabi, P. Biddulph, and
E. Oikonomou. The impact of occupancy patterns, occupant-controlled
ventilation and shading on indoor overheating risk in domestic
environments. Building and Environment, (78):183198, 2013.
[7] S. Porritt and P. Cropper. Heat wave adaptations for UK dwellings and
introducing a retrofit toolkit. International Journal of Disaster Resilience
in the Built Environment, (4:3):269–286, 2010.
[8] R. Gupta and M. Gregg. Preventing the overheating of English suburban
homes in a warming climate. Building Research & Information,
(41):281–300, 2013.
[9] J. Taylor, M. Davies, A. Mavrogianni, Z. Chalabi, P. Biddulph, and
E. Oikonomou. The relative importance of input weather data for indoor
overheating risk assessment in dwellings. Building and Environment,
(76):81–91, 2014.
[10] E. Oikonomou, M. Davies, A. Mavrogianni, P. Biddulph, P. Wilkinson,
and M. Kolokotroni. Modelling the relative importance of the urban heat
island and the thermal quality of dwellings for overheating in London.
Building and Environment, (57):223–38, 2012.
[11] US-DoE. EnergyPlus V8. 2013.
[12] L. Van Gelder, P. Das, H. Janssen, and S. Roels. Comparative study of
metamodelling techniques in building energy simulation: Guidelines for
practitioners. Simulation Modelling Practice and Theory, (49):245–57,
2014.
[13] R. E. Edwards. Predicting future hourly residential electrical
consumption: A machine learning case study. Energy Buildings, 2012.
[14] B. Eisenhower, Z. ONeill, S. Narayanan, V. A. Fonoberov, and I. Mezi.
A methodology for meta-model based optimization in building energy
models. Energy and Buildings, (47):292–301, 2012.
[15] B. Tang. Orthogonal Array-Based Latin Hypercubes. Journal of the
American Statistical Association, 2012.
[16] Indian Society of Heating Refrigerating and Air Conditioning Engineers.
New delhi weather file.
[17] A. J. McMichael et al. International study of temperature, heat and urban
mortality: the ’isothurm’ project. International Journal of Epidemiology,
37(5):1121–1131, 2008.
[18] W. S. McCulloch and W. Pitts. Neurocomputing: Foundations of
research. chapter A Logical Calculus of the Ideas Immanent in Nervous
Activity, pages 15–27. MIT Press, Cambridge, MA, USA, 1988.
[19] T. Schaul et al. PyBrain. Journal of Machine Learning Research, 2010.
[20] D. E. Rumelhart, G. E. Hinton, and R. J. Williams. Neurocomputing:
Foundations of research. chapter Learning Representations by
Back-propagating Errors, pages 696–699. MIT Press, Cambridge, MA,
USA, 1988.
[21] C. Igel and M. H¨usken. Empirical evaluation of the improved rprop
learning algorithms. Neurocomputing, 50:105–123, 2003.
[22] D. S. Broomhead and D. Lowe. Multivariable Functional Interpolation
and Adaptive Networks. Complex Systems 2, pages 321–355, 1988.
[23] F. Pedregosa et al. Scikit-learn: Machine learning in Python. Journal
of Machine Learning Research, 12:2825–2830, 2011.
[24] M. C. Peel, B. L. Finlayson, and T. A. McMahon. Updated world map
of the kppen-geiger climate classification. Hydrology and Earth System
Sciences, 11(5):1633–1644, 2007.