Modeling indoor air carbon dioxide concentration using artificial neural network
Many studies have been conducted on estimating the number of occupants in a building to set the right ventilation rate in order to maintain standard indoor air quality. However, few have focused on predicting carbon dioxide itself based on the room’s available parameters, such as temperature and humidity. This study was aimed at predicting indoor air carbon dioxide concentration in a room using a multilayer perceptron neural network with relative humidity and temperature as inputs. The neural network is a popular data-driven method to provide geometry-independent prediction algorithms. In this study, the neural network was trained in three different ways with the complete, partial, and zero real carbon dioxide concentrations available in the learning process. The sensitivity and specificity analyses were conducted on the output. The most accurate model, based on the calculated mean-square-error method, was five-steps-ahead prediction model with less than 17 PPM difference on average to actual CO2 concentration in the room. Results were also promising for the open-loop model. Carbon dioxide predictions can be used in maintaining indoor air quality by improving ventilation control in buildings.
KeywordsCarbon dioxide Data-driven Indoor air quality Neural network Ventilation
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
- Arens E, Zhang H, Hoyt T, Kaam S, Bauman F, Zhai Y, Paliaga G, Stein J, Seidl R, Tully B, Rimmer J, Toftum J (2015) Effects of diffuser airflow minima on occupant comfort, air mixing, and building energy use (RP-1515). Sci Technol Built Environ 21:1075–1090. https://doi.org/10.1080/23744731.2015.1060104 CrossRefGoogle Scholar
- ASHRAE (2007) Standard 62-2007 (2007). Ventilation for acceptable indoor air quality. American Society of Heating, Refrigerating and Air-Conditioning Engineers Inc., AtlantaGoogle Scholar
- Ebadat A, Bottegal G, Varagnolo D, Wahlberg B, Johansson KH (2013). Estimation of building occupancy levels through environmental signals deconvolution. In: Proceedings of the 5th ACM workshop on embedded systems for energy-efficient buildings. ACM, pp 1–8Google Scholar
- Ebadat A, Bottegal G, Varagnolo D, Wahlberg B, Hjalmarsson H, Johansson KH (2015) Blind identification strategies for room occupancy estimation. In: Control conference (ECC), 2015 European. IEEE, pp 1315–1320Google Scholar
- Haykin SS (2009) Neural networks and learning machines. Pearson Education, Upper Saddle RiverGoogle Scholar
- Skön J, Johansson M, Raatikainen M, Leiviskä K, Kolehmainen M (2012) Modelling indoor air carbon dioxide (CO2) concentration using neural network. Presented at the International Journal of Environmental, Chemical, Ecological, Geological and Geophysical Engineering, World Academy of Science, Engineering and Technology, pp 37–41Google Scholar
- Zhong K, Kang Y (2009) Applicability of air-to-air heat recovery ventilators in China. Appl Therm Eng 29:830–840. https://doi.org/10.1016/j.applthermaleng.2008.04.003 CrossRefGoogle Scholar