Research On Prediction Of Thermal Comfort Model And Interface Design

As the primary mission of buildings is to provide comfortable environment for people to occupy, the research of thermal comfort in buildings has attracted lots of attention and been well developed over the last century. Especially within the current global situation of urbanization, climate change and energy shortage, new requirements and challenges have been proposed to buildings that whether they can still meet the satisfaction of people about indoor environment in a sustainable way. This study explored how thermal sensation is influenced by physical factors of environments and subjective factors of individuals in both thermal comfort models and artificial neural network(ANN), and provide references for indoor environment design.Contents of this study consist of:(1) Measure environmental parameters in a naturally ventilated office building and collect thermal sensation of people by questionnaire to analyze the relationship between actual sensation and theoretical models;(2) Employ secondary data from highly quality-controlled database established by other researchers to see whether the hypotheses can be better explained after classifying data into different climate zones;(3) Analyze prediction of classic thermal comfort model and use ANN to reanalyze the data which thermal comfort model does not have perfect precision.(4) Design friendly interface of thermal comfort according to relevant standard.Key findings could be:(1) Theoretical models did not accurately predict individual thermal sensations, especially when sample size is small;(2) PMV model overestimates hot sensation and underestimates cold sensation and the discrepancy varies in different climate zones;(3) General prediction of ACS model shows a good coincidence to national and international standards, but discrepancy still remains in several climate zones;(4) Compared with classical thermal comfort model, ANN shows a better prediction in specific scenarios and a broad distribution of data may positively contribute to training ANN logic;(5) Interface designed in Matlab matches current standards tightly which can friendly shows the thermal state of people and provide references for understanding indoor environments.