| At present,with the in-depth development of a new round of scientific and technological revolution,it has become a general trend to comprehensively promote the integration of new information and communication technologies such as Internet of things and big data with the development strategy of new intelligent cities.Among them,the construction of smart buildings is one of the important ways to promote the intellectualization of smart building design,construction,operation and service by using new technologies such as the Internet of things.However,the existing smart building construction presents problems such as unclear thinking and blind construction,and insufficient consideration in information utilization,people-oriented and other aspects.With the development of the city and the catalysis of the epidemic,the problem of personnel congestion in the morning rush hours of office buildings is becoming more and more prominent.Therefore,it is of great significance to predict the flow of people in office buildings in the construction of smart buildings.Based on the above problems,this topic builds a smart building system based on the existing Internet of things infrastructure and various platforms,establishes the pedestrian flow characteristic data set based on the collected personnel access data and the collected impact characteristics,and then constructs four pedestrian flow prediction models.The first mock exam shows that the combined model based on variance reciprocal method has a significant improvement over other single models,and provides a basis for decision-making data for staff optimal travel time,improves user experience and reduces enterprise costs.The main work is as follows:（1）After a lot of research,literature research and system requirements analysis,this topic designs the overall architecture and technical composition of the system,and shows the functions and implementation steps of key subsystems.The system data processing flow is analyzed,the corresponding database is designed to access various building data including personnel access data,comprehensively considering the impact of various practical factors on the passenger flow prediction,and finally eight impact characteristics are selected,and a number of pre-processing operations are carried out on the relevant data,and finally the passenger flow characteristic data set is established.（2）Aiming at the problem of people flow prediction in office buildings,this thesis studies the existing people flow prediction methods,uses the most common ARIMA model and support vector regression model,and predicts the people flow in office buildings according to the modeling process of the two methods.The results show that the traditional ARIMA model is difficult to describe the randomness and complexity of people flow fluctuation,and the prediction effect of people flow is poor.Based on the above experimental results,considering that the neural network prediction method has excellent fitting ability to the non-linear pedestrian flow series,the LSTM model is introduced to predict the pedestrian flow.The experimental results show that both SVR and LSTM have better prediction effects on pedestrian flow,and the combined model can improve the prediction accuracy in theory.Therefore,a linear weighted combined prediction model（lstm-svr）is proposed to predict pedestrian flow.The results show that the combined model has better prediction accuracy.（3）This thesis tests the function and performance of smart buildings,expounds the testing process of some core functions of smart buildings,takes root mean square error（RMSE）,mean absolute error（MAE）and mean absolute percentage error（MAPE）as the evaluation indexes of this experiment,and demonstrates that the lstm-svr combined prediction model has practical application value in the prediction of pedestrian flow in office buildings,It can provide effective decision analysis support for peak passenger flow detection of intelligent building system. |
PDF Full Text Request