A Study On The Mechanism And Model Of Window Opening Behavior Of Infant Families In Hot Summer And Cold Winter Region

Window opening behavior is a universal and effective way to achieve building energy saving,improve thermal comfort and create a healthy environment.Studying its behavior characteristics can provide valuable guidance for energy saving design and the construction of good indoor environment.Scholars in domestic and overseas have carried out a lot of research on different types of buildings（office buildings,residential buildings,hospitals and schools）,among which the window opening behavior of residential personnel is one of the research hotspots in recent years.However,as the typical type of family member structure in China（the infant families）,there is a lack of research on its window opening behavior.The living habits and the demand for indoor environment of infant families have their own characteristics,which may lead to differences in window opening behavior from other types of buildings,so it is necessary to study the window opening behavior of infant families at this stage.This paper selected infants families（including infants aged 0～3）in Hunan in the region of hot summer and cold winter as the object of study.Firstly,the method of literature investigation was used to determine the factors that may affect the window opening behavior of infant families.Four households were randomly recruited to conduct a measurement of window opening behavior in living room and bedroom throughout the year.Based on measured data,the average daily window opening duration and the average daily number of window opening actions of infant families were analyzed.It was found that the average daily window opening duration in infant families reached its maximum in summer,followed by spring and autumn,and was shortest in winter.The average daily window opening duration of infant families during the transition season and the summer were significantly higher than those of the ordinary households in the same climatic zone（average exceeded by 383min/d,334 min/d）,but was significantly lower in winter（average reduced by 159min/d）.In addition,the average number of opening windows in infant families is 0.8 times per day,which is also significantly higher than 0.4 times per day in ordinary families,indicating that infant families have a stronger awareness of window opening control,which may be that the main body of opening windows gives more consideration to the feelings of infants.In the analysis of the effects of non-environmental factors,we mainly explored the effects of weekday or not and different time factors of a day on the window opening behaviors.The study concluded:1)There was no significant difference in window opening probability between working days and non-working days for infant families.2)There were differences in the period when the maximum of window opening frequency appears during different seasons.The window opening frequency reached the maximum at 7:00～8:00 in the morning during the transition season and was advanced and postponed by 1 hour in summer and winter respectively.3)There were differences in the frequency of window opening in living rooms and bedrooms at different times of the day.The window-opening frequency of the living room in the afternoon and evening was larger,and for bedrooms,except in the morning,the window-opening frequency varied less in the rest of the time.Then the present situation of indoor and outdoor environmental factors in different seasons is described,and the influence of single factor is analyzed.The research shows that:1)The window opening probability of outdoor temperature at30℃～36℃was higher than that at 8℃～16℃during the transition season,indicated that infant households were more tolerant to higher outdoor temperatures than to lower outdoor temperatures.2)The critical value of outdoor relative humidity was80%,above which the window opening probability decreased rapidly.3)The lower window opening probability at higher indoor CO₂concentrations was due to that the elevated indoor CO₂concentration was a result of the window state being closed.4)Outdoor wind scale had a weak positive correlation in spring,autumn and winter,and a strong negative correlation in summer.This is because the variation range of wind scale in summer was larger（0～14）.Infant families closed windows when the outdoor wind speed was very high,while the window opening probability was not affected when the wind speed was low in other seasons.Binary logistic regression was used for modeling,and statistical methods such as Kendall tau correlation analysis and collinear diagnosis were used to determine the input variables of the model（including environmental factors,time factors）.It was found that:1)Among the time factors,the morning period was the most significant variable affecting the window opening behavior,followed by the outdoor temperature in the environmental factors.It showed that the window opening behavior of infant families was closely related to time factors,and the influence of time factors cannot be ignored in logistic regression modeling of this type of families in the future.2)The annual prediction accuracy of the model was 79.3%,and the prediction accuracy of model of window closing behavior in summer and window opening behavior in winter was only about 30%.This was due to the amount of data with windows closed in summer and windows open in winter accounts for a small proportion of the entire season,indicated that the model had unstable performance in dealing with unbalanced data.Aiming at this problem,we adopted the random forest algorithm with robust performance when dealing with unbalanced data.In order to optimize the model,we used the combination of learning curve and grid search to adjust the parameters.It was concluded that the hyperparameter n＿estimators（number of decision trees）which had the greatest influence on the random forest models in different seasons was in the range of 16～19.The comprehensive prediction accuracy of the random forest model with optimized parameters was 3.2%higher than that of the unadjusted parameters,and 19.8%higher than the binary logistic regression model.In addition,the model performed very superior in predicting window closing behavior in summer and window opening behavior in winter（prediction accuracy:97.8%,95.6%,respectively）.Which fully confirmed that the random forest model could effectively solve the problem of binary logistic regression model’s poor prediction when dealing with imbalanced data.The research results of this topic can be used to refine the description of the window opening behavior of infant families by energy consumption simulation software and air quality evaluation software.At the same time,it also provides a valuable reference for infant families’members to create a comfortable and healthy indoor environment.