Machine learning for human comfort evaluation of HVAC system
Singapore is a world-renowned developed tropical country and its outdoor temperature is always high due to its geography location near to the equator. For this reason, Heating, Ventilation, Air-Conditioning (HVAC) systems are widely used in indoor ventilation and temperature regulation to ensure the...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Theses and Dissertations |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/76034 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Singapore is a world-renowned developed tropical country and its outdoor temperature is always high due to its geography location near to the equator. For this reason, Heating, Ventilation, Air-Conditioning (HVAC) systems are widely used in indoor ventilation and temperature regulation to ensure the indoor comfort of the staff and the citizens.
In the indoor environment, because of the individual differences of each person, the comfort feeling is a different experience for everyone, and the work of HVAC is particularly important. The discomfort caused by defects in the environment will consume people's energy and make people feel unwell, which will greatly reduce people's work efficiency, and it will also make people have negative emotions.
Therefore, predicting a comfortable experience based on various factors of the environment and people is a very meaningful thing. By predicting comfort, the bad experience indoor can be avoided as much as possible.
In this dissertation, human comfort PMV indexes are explored and predicted by human gender, height, weight, BMI, metabolic rate, body surface area, stress levels, clothing comfort, and air temperature in the environment. This will be achieved by training these parameters into the constructed neural network model and changing the training function, the number of hidden neurons, the learning rate, and the method of changing the activation function. And measured by the mean square error and the accuracy of the test. The best neural network model will be achieved with minimal MSE and maximum test accuracy. So the best performance can be achieved by using 5 hidden neurons and Levenberg-Marquardt train function, and the learning rate of 0.01. |
|---|