UTHM Institutional Repository

Development of recurrent neural network modelling for indoor temperature and indoor air quality (IAQ) prediction: a reflection of outdoor parameters

Ismail, Lokman Hakim and Ghazali, Rozaida and Khamidi, Mohd Faris and Abdul Wahab, Izudinshah and Abdul Razak, Azman and Ghazali, Suraya (2013) Development of recurrent neural network modelling for indoor temperature and indoor air quality (IAQ) prediction: a reflection of outdoor parameters. Other thesis, Universiti Tun Hussein Onn Malaysia.


Download (4MB)


Nowadays, people are demanding on their quality lifestyle. For a quality lifestyle, healthy, safe and comfortable indoor environments are needed. Therefore, this research investigates the environment condition within two types of buildings, natural ventilated and air-conditioned to gain better understanding of the indoor environment condition of the selected buildings. The main purpose of this study is to develop predictive model to forecast indoor environmental parameters by using Artificial Neural Network (ANN) technique. Hourly time-series data of three indicators including: air temperature, relative humidity and air velocity have been used to forecast the indoor environmental parameters. The data collected fiom the field measurements conducted at four traditional Malay houses have been used in training the network. Meanwhile, another two data sets fiom the field measurements conducted at a restaurant which applied'the concept of vernacular architecture and a library which was built with green building concept have been used to validate the model developed, The performance of the developed model is evaluated using R sqwed (R~a)n d was messed through a measure of Mean Square Error (MSE). The accuracy,. , .o: f the model is measured using the Mean Absolute Percentage Error (MAPE). The study has successfully developed ANN model to forecast indoor temperature, indoor humidity and indoor velocity. From the results, it was found that the best ANN to forecast indoor temperature is 3-37-1, 3-12-1, 3-30-1 and 3-17-1. As for the indoor humidity forecasting, the best ANN is 3-40-1, 3-1 1-1, 3-31-1 and 3-36-1. Lastly, the best ANN to forecast indoor velocity is 3-23-1, 3-30-1, 3-35-1 and 3-18-1. In conclusion, the evaluation performance of the model through the MAPE value shows that except for air velocity, the ANN model can forecast temperature and humidity within natural ventilated building. As for the airconditioned building, it is 'not accurate' due to the different building characteristics.

Item Type: Thesis (Other)
Subjects: T Technology > TD Environmental technology. Sanitary engineering > TD881-890 Air pollution and its control
Divisions: Faculty of Civil and Environmental Engineering > Department of Structural and Material Engineering
Depositing User: Normajihan Abd. Rahman
Date Deposited: 28 Apr 2016 06:35
Last Modified: 28 Apr 2016 06:35
URI: http://eprints.uthm.edu.my/id/eprint/7962
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item


Downloads per month over past year