Energy consumption calculation of campus buildings based on the law of personnel flow

For a certain functional building group, the coupling relationship between the load of a single unit and entire building groups lies in the spatial and temporal patterns personnel distribution. Using the DeST (Designer's Simulation Toolkit) method, the personnel flow trajectory in a university campus in Shanghai was analyzed and the established energy consumption calculation model based on personnel flow trajectory of university buildings. It is proposed that personnel mobility should be included in the application of DeST to predict the total load of building groups. The results show that after the people flow trajectory was included, the total annual accumulated load decreased from 27.01 to 19.39  GWh, reducing by 28.20%; the hourly load fluctuation became more stable, the daily peak load decreased by 22.57%, and the daily peak valley difference decreased by 43.74%. The results suggest that the prediction for the air conditioning load of building complexes would be better when the trajectory of human flow is considered, minimizing unnecessary system scale and investments. Under the “double carbon” policy, the personnel flow trajectory model can effectively reduce waste in air conditioning energy consumption in campus buildings.


| INTRODUCTION
Construction, industry, and transportation are the three major sources of carbon emissions. 1,2 Currently, buildings alone are responsible for 30%-40% of the overall energy consumption of world and growing with a rate of 2.7% per year. 3,4 In recent years, China's emphasis on education has increased the number of universities. With the rapid growth in academic campuses, building energy consumption has drawn greater public attention. 5 According to the Ministry of Education, at the end of 2021, China had 3012 institutions of higher learning and adult higher education, with a total scale of more than 44.3 million people. 6 Reports suggest that the energy consumption per unit area of university buildings is 5-10 times higher than that of residential buildings. 7 As a result, while the number of students accounts for only 3% of the country's population, the energy utilized by university buildings amounts to 10% of the total social energy consumption.
Due to the continuous promotion and implementation of energy-saving policies, energy conservation in university buildings has recently become one of the most pressing issues. The accurate prediction of the energy load of college buildings is one of the major tasks in energy conservation among academic institutions. In the Buildings Report of the International Energy Agency, six main factors influencing building energy consumption were discussed, which included personnel activities, considered one of the most important factors. 8 Some studies have suggested that the uncertainty and variability of personnel activities are the main reasons for the difference between building performance and design expectation. Some have found that different human activity patterns may cause up to 60% difference in energy consumption. 9,10 Compared to commercial and residential buildings, campus buildings are closed. The crowd inside a campus only flows in different functional buildings and has little total population changes. This means that the flow of personnel in different buildings has special significance for the air conditioning load composition of campus buildings compared with other buildings.
Zhao et al. 11 investigated the occupancy rate of single buildings on different working days and established 17 models of different climate zones. They found that the occupancy rate had the greatest impact on each factor model of energy consumption. Markovic et al. 12 studied the personnel behavior and the control of window switches in office buildings. The calculation results showed that the model's accuracy increased between 86%-89%. Zheng et al. 13 established the staff rest model of survey data based on the flow trajectory of large shopping malls. Compared with the conventional model, the average deviation of the hourly cooling load was 3.29%, and the average deviation of the daily cumulative cooling load was reduced to 11.8%. Ahn et al. 14 proposed the "random walk" model in line with the fixed human behavior pattern and predicted the impact of human movement behavior on the energy consumption of two laboratories and reading room buildings using the Markov chain.
They found that compared with single buildings, the load curves of regional building groups are more closely related to personnel in time and space. Brounen et al. 15 investigated the energy use of 305,001 residential buildings and found large differences in the load curve between households in terms of time and change trends, proving varied universality. Fonseca et al. 16 concluded that the time of user load generation should correspond to the different air conditioning spaces in the regional system, which is crucial in equipment selection and control strategy. Weissmann et al. 17 established the load curves of 144 buildings and found that the differences between buildings have a peaking effect on the centralized cooling systems, especially in high-density building areas, and that the diversity of load curves should be used. An et al. 18 proposed the random human behavior model for regional load calculation of residential buildings. Using survey results of regional personnel activities, the DeST (Designer's Simulation Toolkit) modeling was applied to calculate the load, greatly improving the result accuracy. A number of studies on Modelica modularized the construction of human behavior, providing wider applications and better reflecting the attributes of human behavior. 19 When predicting the total load of regional building complexes, most studies have only considered the simple superposition of peak loads of single buildings, while the people movement has largely been overlooked in the room situation. The same is true for university buildings. For instance, Xiang 20 used the area index method to estimate the total load of the whole region and found that hotels and offices are established after the typical building models of commerce. Zhang 21 determined a total load of hourly load superposition method based on the model of campus buildings, including a dormitory, academic building, and canteen. Du et al. 22 indicated that the functions and work schedules of each building in the campus complex were not completely consistent. They concluded that the load difference between each building and the whole should be fully considered.
The above literature shows that human behavior can significantly impact building energy consumption. It is easier to investigate and simulate the movement trajectory of people in a single building. It proved that studying the behavior of personnel in buildings can provide more practical guidance and analysis of energy consumption. The important role of personnel has been proven in building load forecasting and improving simulation accuracy. However, the literatures lack quantitative research on the impact of human behavior on building energy consumption. The literature often uses default work and rest in simulation. In the literature, when calculating the total load in a relatively closed building group, the impact of personnel factors on the total load is not considered, and only conventional superposition is used.
This paper further discusses the defects of existing research that only qualitative but not quantitative factors affect people and the defects of DeST software in predicting group building loads. In this study, a university building complex in Shanghai was used as the research object to explore and analyze the work and rest rules of campus personnel flow based on the activity model of single building personnel. The total load and dynamic load curve of regional buildings were calculated using the DeST model, which is affected by personnel flow. By incorporating the compression and analysis of energy-saving measures, the accuracy of load forecasting of regional buildings would be improved.

| Research method
The building energy consumption simulation software, DeST, independently developed by Tsinghua University, was selected as the simulation tool. DeST is characterized by "phased design and phased simulation." Connect the environment and buildings by natural room temperature. It is more suitable for Chinese buildings. It is used to calculate hourly load of various typical buildings on campus. After investigating the hourly flow trend of all kinds of building personnel in the campus, the staff work and rest will be introduced into the established models. Compare the hourly superimposed total building load method with the total building load method of adding people flow trajectory model to prove the impact of people flow on the total building load in the building groups.

| Physical model
Based on the technical guidelines for the construction of an energy conservation supervision system of campus buildings in college, 23 the buildings in colleges are divided into 13 categories. The main building types are academic buildings, office buildings, libraries, canteens and dormitories. Based on the ministry of housing and urban rural development and the national development and reform commission in 2018, the average area of a comprehensive college student is 26.60 m 2 . The building structure in this paper comes from the actual building, and the thermal parameters and area meet the specification requirements. It is a universal building group that can reflect the current situation of colleges and universities. The main thermal parameters of typical model buildings is listed in Table 1.
Established five types of typical building model diagrams, as shown in Figure 1.

| Mathematical model
(1) The cooling load comprises non-transparent envelope cooling load, air cooling load, solar radiation cooling load, and personnel load. The design value of the hourly total cooling load is formed by the hourly superposition of various loads. The building cooling load is expressed as follows: where Q τ c( ) is the architectural cooling load, W; q τ 1( ) is the non-transparent envelope cooling load, W; q τ 2( ) is the air cooling load, W; q τ 3( ) is the solar radiation cooling load, W; q τ 4( ) is the personnel cooling load, W; T A B L E 1 Thermal parameters of typical model buildings. q τ 5( ) is the lighting cooling load, W; q τ 6( ) is the equipment cooling load, W.

Building types Unit
(2) The heat dissipation of the human body is affected by many factors, such as gender, age, clothing, and labor intensity. The latent heat radiated by the human body and the convective heat dissipation directly form the instantaneous cooling load. The amount of radiated heat is delayed in time and can be calculated by the cooling load coefficient method. The heat dissipation among adult men is used as the calculation basis, and the clustering coefficient is introduced for correction. The sensible heat dissipation caused by the human body is expressed as follows: where q′ τ 4( ) is the cooling load is formed by the sensible heat of the human body, W; q s is the sensible heat dissipation of adult men, W; n is the number of people; φ is the clustering factor; C LQ is the cooling load coefficient of sensible heat dissipation of the human body.
The latent heat dissipation caused by the human body could be expressed as follows: where q″ τ 4( ) is the cooling load formed by latent heat dissipation from the human body, W; q l -the latent heat dissipation of adult men, W.
The hourly cooling load of personnel is composed of the cooling load formed by sensible heat dissipation and the latent heat dissipation of the human body.
(3) The occupancy rate of each building staff on the campus meets the matrix distribution. The matrix can be represented by the time and personnel-inroom coefficients. The room occupancy coefficient of each building on campus is set as x, and the air conditioning operation time is set as t. After sorting and summarizing, the relationship between the final person in the room y could be expressed as follows:

| Analysis of work and rest rules of personnel
Indoor personnel status (identifying when and how many people were in the room) is a fundamental input parameter in building energy consumption simulation. Directly determined by the movement of personnel, the flow of indoor personnel can be precisely characterized by an accurate simulation of the room. 24 The load depends on the presence of indoor personnel. For regional buildings, the flow of people in different buildings determines the room occupancy rate. This study focused on the changes in the number of people inside each building at different periods and did not include the reasons for the movement to concentrate on the impact of indoor staff activities on building energy consumption.
Three methods were used: infrared monitoring, academic affairs course plan review, and investigation of the personnel flow trajectory of campus buildings. The duration was set to one semester, and the analysis focused on the hourly personnel density of five campus buildings. The comparison results between the DeST default staff schedule and the survey schedule are shown in Figure 2.
There were particular differences between the changing trend of DeST default staff and the actual investigators in the survey results. Except for the dormitory building, the DeST default staff work and rest for the other building types were greater than the actual survey results. The largest difference was found in the academic building, with its maximum staff work and rest decreasing from 0.445 to 0.20/m 2 . There was no peak from 16:00 to 22:00, which is a stable downward trend. Since most of the daily courses are given in the morning, a big gap can be found between the staff's work and rest time in the academic building, which is true for both working and nonworking days. Such buildings are significantly affected by time.
The library staff schedule was stable, with established fluctuations and strong regularity. The maximum daily work and rest was 0.14/m 2 . Compared to the default setting in DeST, the work and rest for the library staff on nonworking days increased to 0.17/m 2 . Since there are no classes on nonworking days, students prefer to study in the library. Since the office has master's and doctoral studies from 16:00 to 22:00, it does not drop to 0. For the canteen building, its personnel density rapidly increased during particular periods (i.e., 12:00-13:00 and 18:00-19:00), with numbers greater than the DeST default values. The dormitory building was significantly affected by the other buildings; most students prefer going to the dormitory to rest when they have no class and after classes.
In general, significant differences were found in the work and rest among the construction personnel in the same time period. This means that when the influence of personnel flow is not considered, the campus simulation will yield inaccurate results.

| RESULTS AND DISCUSSION
The data of campus staff investigated were imported into the typical building model personnel thermal disturbance module in the form of staff work and rest. Change the room occupancy rate of the original building personnel, and build the energy consumption model of personnel flow trajectory. The DeST simulation calculations were adopted. The hourly cooling and heating load and accumulated cooling and heating load indexes were computed, and the results were extended to the entire campus building complex. Without considering the personnel flow trajectory, the calculation results are illustrated in Figure 3.

| Comparison of hourly cooling load on typical days
DeST calculated the hourly cooling and heating load of 8760 h for the entire year. Take the hourly cooling load of typical summer day (6.29) as an example, compare the hourly cooling load of various buildings and campus buildings under the two models, as shown in Figure 3.
The calculation results in Figure 3 suggest that the hourly cooling load value and load curve of a typical day changed to a certain extent after adding the personnel flow trajectory. For the office building, the hourly cooling load decreased by 21.26%, reaching 30.19% at 14:00. The peak was found at 9:00, and there were no obvious troughs in the personnel flow trajectory model. Because the period of 13:00-14:00 is the lunch break time for master's and doctoral students and faculty members, the difference between peak and trough decreased from 27.5% to 20.2%, resulting in a smoother curve. For the academic building, the hourly cooling load of the personnel flow trajectory model was 48.17% lower than that of the conventional model. Since some classrooms remained empty during class time, the daily average staff density was less than the original setting. The typical daily load could be divided into three parts: 8:00-12:00 (class time), 14:00-19:00 (class time), and the rest of the evening (self-study time). The three parts load was roughly equal to 450 kW, and the relative deviation was within 10%. When the lights were turned on at night, there was an increase in motion.
For the library building, the load rapidly increased to its peak within an hour of opening. The difference between the two peaks was about 675 kW; the cooling load trend from 8:00 to 8:00 slightly differed from that of the conventional model. The personnel flow trajectory model exhibited a slow upward trend and had no peak valley alternation. The peak times for students studying in the library were in the afternoon and evening. The dormitory differed from the other buildings, with the air conditioning turned on the entire day. There was no significant difference between the survey results and the default DeST calculation results. The room occupancy rate for the dormitory building remained constant throughout the day, and the hourly load fluctuation was minimal. For the canteen building, the total cooling load decreased by 45.82%, differentiating a school canteen from a normal one.
Generally, the traditional area index estimation method does not reflect hourly load change trends, which causes energy waste from the overestimation of the total load. The hourly dynamic load curve of regional buildings is presented when the conventional dynamic load is superimposed hourly. As shown in the figure, the daily load of campus buildings had "three peaks" occurring at 8:00, 17:00, and 21:00. The maximum value was 36710.64 kWh, with an average load of 22126.75 kWh. The low load time occurred at 13:00 and 20:00. The minimum value was 27674.40 kWh, with an average load of 16304.14 kWh. When the personnel flow trajectory was added, the peak time appeared only at 8:00, and the load fluctuation (about 20,000 kWh) for the rest of the time was small. The daily peak load decreased by 22.57%, and the average load decreased by 26.32%. The daily peak valley difference and the daily peak valley difference ratio were both reduced, and the load curve was more stable between 8:00 and 17:00. The results suggest that, to a certain extent, the hourly load is influenced significantly by personnel density at different times.

| Comparison of hourly cooling load in typical weeks
The hourly cooling load of a typical week in summer (6.23-6.29) was selected for analysis. The comparison diagram is shown in Figure 4. The figure shows that the cumulative total load for the academic building at weekends was only 7% of the total load of the teaching week. Among the different campus building types, academic buildings are the most significantly affected by whether the days are workdays. The total load during nonwork days for the dormitory, office building, and library accounted for 59%, 45%, and 40% of the workday load. The canteen was the least affected by the work-holiday schedule. The overall load change trend was largely the same, and the total accumulated cooling load at weekends was about 53.7% of the work day.

| Comparison of accumulated cooling and heating loads on campus
It can be seen from Figure 5 that the cumulative load index per unit area can reflect the influence of personnel on the cooling and heating load of buildings and plays an important role in architectural design and planning. After the personnel flow trajectory was added, the accumulated cooling load index per unit area of the canteen was reduced by 45.23%, from 120.55 to 66.02 kWh/m 2 . The cumulative heat load index per unit area decreased by about 48.29%, from 27.54 to 14.24 kWh/m 2 .
The campus canteen differs from other canteens, having pronounced time limits since most people eat at a fixed time. Short use time and short meal time are the main reasons for the significant load reduction. Among the campus buildings, the canteen is the most affected by the trajectory of personnel flow. In comparison, the relatively fixed personnel and the extremely light activity intensity cause the minimum decrease in the dormitory building.
From Figure 6, we can see that the accumulated total load for the entire year decreased from 27.01 to 19.39 GWh, reducing by 7.62 GWh or about 28.20%. The accumulative total load gap for the academic building was the largest, followed by the canteen building. Their cumulative load comprised about 40% of the total campus load. These two building types are the focus of energy conservation on academic campuses.
Most campus buildings have only one library, and the flow of librarians is relatively constant. The dormitory had the lowest cumulative heat and cold load indices, varying the most compared to other buildings. The total dormitory area was larger than the others. Therefore, the proportion of the cumulative total load on the campus was only lower than that of the academic building. The cumulative load for dormitories accounted for the smallest proportion, and the effect of the personnel flow trajectory is not as pronounced. It is worth noting that the office buildings account for the work of the faculty, staff, and graduate students.
To verify the scientificity of the simulation data, the dormitory building is taken as an example. Referring to the literature, 25 the measured cooling load per unit area is about 32 kWh/m 2 and the simulation data in this paper is 34.65 kWh/m 2 . They have the same building type and climate areas. It is believed that the gap between them is small and the simulation data are accurate.

| CONCLUSION
In the actual building operation process, there are changes in regional load caused by different personnel flow patterns. This should be fully considered when using DeST to calculate the total load of the building group, reflecting the relationship between the ebb and flow of personnel in the building. College campus is not a simple superposition of multiple buildings, but an organic whole. There is a close relationship between the total load and each building load. The DeST software was used to explore and model the total load and the impact of personnel flow trajectory for different building types on a university campus in Shanghai. The analysis also included the total cumulative load and the typical hourly cooling load change trends for the various campus buildings.
With the addition of the personnel flow track, the total annual cumulative load was reduced by 28.20%. During summer, the daily dynamic cooling load curve was gentler; the daily peak load decreased by 22.57%, the average load decreased by 26.32%, and the daily peakvalley load difference decreased by 43.79%. Among the campus buildings, the academic building had the biggest contrast in energy load between workday and nonworkday. The load in dormitory and canteen buildings was found to fluctuate steadily. As for library and office buildings, the load was regular and not easily affected.
The number of people among the various building types fluctuated, significantly affecting the campus energy load. The results indicate that when DeST is used for regional loads forecasting, personnel flow trajectory is an important factor that must be considered.
Based on the conclusion, in future research, it needs to determine whether we can find the internal law of personnel mobility, summarize them into formula to replace the existing research work, and improve the efficiency of prediction methods. This is the direction to improve the scientificity and accuracy of building groups energy consumption prediction in the future.