Recognition Research Of Occupancy Information Based On Indoor Environmental Parameters

Occupancy information has an important impact on building energy consumption,and mastering the status and behavior of occupants can provide guidance and help for building energy simulation research and the operation of HVAC systems.However,the variability of the number of occupants and the uncertainty of their behaviors bring many challenges to the recognition of occupancy information.At present,the domestic research on this aspect mainly focuses on occupants behaviors,while the research on the recognition of the number and status of occupants is less.Moreover,the commonly used occupancy information recognition methods may invade occupants' privacy and it is necessary to explore a suitable method.Therefore,this paper takes the graduate student workshop of a university in Harbin as the experimental object to carry out the work.Based on the online monitoring of indoor environmental parameters,this paper uses and mines the occupancy information contained in environmental data and proposes a method of using time series analysis and data mining model to identify the status,number and typical behaviors of occupants.Firstly,environmental data?indoor temperature,relative humidity,illuminance and CO₂ concentration?were collected for 9 weeks by using environmental parameters equipment.Meanwhile,occupancy information was recorded synchronously by the author.Through the data obtained,it is found that this is a low-cost,efficient,non-invasive,and high-quality data collection method.Compared with the direct monitoring of occupants,the method has obvious advantages.Secondly,based on CO₂ concentration,ARIMA model was used to predict CO₂ variation trend and seasonal decomposition model was used to predict occupants variation trend,which realized the prediction of the trend of CO₂ concentration change and the recognition and prediction of occupancy in the graduate's workplace.The ARIMA model results show that the MAE of 1 day and 3 day prediction is about 60 ppm and MAPE is about 7%.The seasonal decomposition model results show that the average accuracy of occupants' status recognition reaches more than 97%,the accuracy of occupants' number level recognition is close to 90%,and the accuracy of occupants' number recognition is about 60%.Lastly,based on multiple environmental parameters,the data mining models were proposed to identify occupants behaviors include opening or turning off the lights and opening or turning off the door in the workshop.The recognition results show that the C5.0 model has a high accuracy which is about 85% in occupants behaviors identification.And this paper studied the relationship between the two behaviors and obtained the association of rules of opening the door???opening the lights.