Field Study and Analysis of Passenger Density in the Beijing Subway Transfer Hall

Abstract

Passenger density is a major factor influencing indoor climate conditions and the corresponding energy consumption in the transfer halls of metro transportation hubs. In this study, passenger flow in a subway transfer hall in Beijing was measured across three typical seasons. A differential equation model was established to account for the characteristics of passenger density. Based on this, a passenger density prediction model was developed and utilized for analysis. The results show that passenger density is primarily affected by operation time and the surrounding population composition, with a relatively weak correlation to outdoor weather conditions. Additionally, passenger density differs significantly between weekdays and weekends, with the weekday peak being 2.5 times higher than that on weekends. During the four peak hours on weekdays, both in the morning and evening, passenger density exceeds the design capacity, resulting in insufficient fresh air supply. However, for most of the weekday hours and for all hours on weekends, passenger density remains below the design capacity. This indicates a substantial potential for energy savings by adjusting the fresh air supply according to actual passenger density.

1. Introduction

With increasing urbanization and the evolution of the technological sector in China, urban integrated transportation hubs have experienced a rapid development trend in terms of design and technologies. In this regard, based on the Beijing Municipal Transportation Commission plan, the city will build and expand nearly 40 integrated transportation hubs by 2038. In Beijing, about 9 million commuters use the subway as a daily means of transportation [1]. Overall, the transfer hall of the transportation hub is characterized by a large space span, complex functions, high population density, and relatively long operation times. This results in a large energy consumption of the corresponding buildings compared with ordinary public buildings [2,3,4,5]. Moreover, in the design stage of these buildings, passenger density is a major parameter used to estimate the human load as part of air conditioning load calculation and to determine the energy consumption of the fresh air system. Generally, in the transfer halls of transportation hubs, high densities of passengers flow quickly and vary greatly with time. Thus, these passengers hardly interact with the buildings, as is usually the case in other buildings [3]. However, this parameter is often regarded as a major cause of uncertainty in building performance evaluation and, in some cases, results in high energy consumption levels [6,7]. Therefore, research and analysis of the passenger density in transfer halls is crucial to reduce the energy cost of air conditioning systems without sacrificing indoor air quality and comfort levels [8].

Passenger flow is the amount of passenger movement in a certain environment. In most cases, the directional distribution is relatively balanced and steady, as passengers typically move in a uniform pattern in terms of location. On the other hand, passenger flow is very uneven when considering the time distribution, but shows some general regularity [9]. In recent research conducted by scholars, three methods have been used to count the number of people in a certain room at a specific time: (i) surveillance video [10,11,12,13]; (ii) indoor environment sensors, such as temperature, humidity, and CO₂ concentration sensors [14,15]; and (iii) wireless signals, including Wi-Fi, cell phone signals, Bluetooth, and others [16,17,18,19]. Based on the field investigation methods and results presented above, the recorded passenger flow data has been used in many studies to develop traffic hub passenger flow prediction models. Some of the widely employed big data analysis methods include statistical analysis, deep learning, genetic algorithms, and neural networks [20,21,22,23,24,25]. In their study, Liu et al. [26] collected passenger flow data from the Dongzhimen subway transportation hub and conducted statistical data analysis to derive monthly, weekly, and hourly passenger flow patterns for different modes of transportation at the hub, which has important guiding significance for its scientific operation management and safety prevention. In another study, Zhong et al. [27] used cellphone signals to detect passenger flow patterns at the Shanghai Hongqiao traffic hub and proposed a passenger flow classification extraction method to study the passenger movement characteristics at different time intervals. The results further verify the potential of using mobile phone data to monitor and characterize passenger flow related to transportation hubs. Additionally, Xu et al. [28] examined the passenger flow in the closing period of subway stations. They proposed a spatial–temporal ensemble prediction model to capture the temporal and spatial characteristics of the passenger flow at each station as well as the passenger flow between different station groups. In order to address the effects of station closures, a support vector regression method was designed for studying the residual time series of the integrated model. The proposed model was found to be superior to other advanced models in estimating the accuracy and reliability of the station closure effect. Gu et al. [29] established a spatial–temporal passenger distribution model to describe the transient passenger distribution in 14 characteristic areas of an airport terminal. The airport operation data and the results of the passenger flow simulation using AnyLogic software 8.5 were used to simulate the energy consumption. The comparison results validate the proposed model’s ability to optimize the HVAC system of the future airport design.

These previous studies into the relationship between building energy consumption and indoor occupants were mainly performed in office buildings and airport terminals [30,31,32,33,34]. In office buildings, the total occupancy rate is basically determined by work schedule [32,35]. Furthermore, the occupancy distribution in offices is mainly affected by occupants’ preferences and diverse space functionalities [33]. In airport terminals, the total occupancy rate is influenced by flight schedule. However, the characteristics of passenger density in subway transfer halls are neither entirely analogous to those in office buildings, which are determined by work schedules, nor to those in airport terminals, which are closely tied to flight schedules. With regard to the transfer halls of subway transportation hubs, the specific passenger flow characteristics, the variation in the number of people in the room with time, and the implementation of passenger density as a key parameter in the design and operation of building energy systems have not been studied in detail. Additionally, the present regulations and scientific basis for considering passenger density in the design process of transportation hub buildings need a drastic improvement. In this regard, when designing these facilities, designers often rely on the number of evacuees provided by the building professional or estimated based on experience. Most of the time, this value is often large, resulting in a waste of energy and the extensive unneeded operation of air conditioning systems and fresh air supply.

In this study, the focus is on the accurate calculation of passenger density in order to passively reduce the energy consumption of an air conditioning system. This approach is similar to the method used by Wang [36,37], which passively reduces building energy consumption by enhancing the performance of recycled aggregate concrete. An analysis is carried out for a typical three-month period of measured passenger flow data over different seasons in the transfer hall of a subway transportation hub in Beijing. On this basis, an hour-by-hour prediction model of the passenger density in the transfer hall is established and verified, considering the characteristics of passenger flow and the key influencing factors. we compared the design and calculation values of passenger density and discuss the occupancy rates in the transfer hall. The results can be useful both in the design and operation processes of HVAC systems in the subway transfer hall.

2. Methodology

2.1. The Subway Transport Hub Case Study

This study examines the case study of a metro transportation hub in Beijing (39°54′ N, 116°25′ E) as the object of research. In the considered site for investigation, the transfer hall is the interchange area of the metro, bus, and coach on the ground. The hall is a single-story building with a construction area of 8300 m² and a height of 15.3 m. The northern part of the building is connected to an office building, which has four floors above ground. The southern part of the site is connected to another building with six floors above ground, mainly comprising an arrival area, a commercial area, and an office area. The different access points to the transfer hall are illustrated in Figure 1, as follows: entrance (a) and exit (b) of the long-distance locomotive and north building; entrance and exit of the north side (c); entrance and exit of the subway interchange (d); entrance and exit of the south side (e); entrance of the bus and south building (f); and exit of the bus interchange (g). These entrances and exits are also considered the locations of the passenger flow measurement points in the transfer hall.

Furthermore, the transfer hall brings together different passenger flows from multiple transportation interchanges with direct interaction with the office and commercial areas and their corresponding personnel. Thus, the transfer hall passenger flow line has a significant impact on the efficiency and function of the whole city [38]. This transfer hall (Red line area in Figure 1a) is the main research object in this paper, and it will be examined and analyzed with a focus on the variation in passenger flow at each of its entrances and exits listed above. Following the transfer hall plan shown in Figure 1, it is noted that passengers who change to the subway will leave and enter the transfer hall from exit/entrance (c). Then, they can change to take a bus or a coach or enter the office and commercial areas of the south and north buildings. Additionally, passengers who exit the bus in addition to the office and commercial areas’ personnel of the south building will leave or enter the transfer hall via exit/entrance (e) of the south building. In this way, they can change to the subway, take a coach, or even leave this transportation hub for the nearby areas. Moreover, passengers who take the long-distance locomotive and use the office building in the north will enter the transfer hall through entrance (a) and leave through exit (g). Finally, passengers can enter or exit the transfer hall through the outdoor north entrance/exit (b) and south entrance/exit (d).

2.2. Key Parameters Affecting Passenger Density

Overall, this study considers three main factors that may affect the passenger density in the investigated transfer hall:

(1)

Surrounding land type: According to a previous study [39], passenger flows in transportation hubs are directly connected to the surrounding land type. Depending on the type of surrounding land, the passenger flow exhibits different time distribution characteristics, as follows:

(a)

Residential and office type: In this case, the transportation hub is mainly surrounded by residential districts and areas with houses and apartments, in addition to areas with office buildings and facilities. Under this condition, and following the characteristic of committing hours, the passenger flow in the corresponding transportation hub exhibits obvious morning and evening peaks.

(b)

Commercial and large hub types: In this case, the transportation hub is mainly surrounded by commercial buildings and facilities, airplanes, trains, and other rail transit systems. Due to the nature of these facilities and surrounding areas, there is no peak passenger flow on weekdays, and the transportation hub experiences a relatively consistent daily passenger flow distribution. On the other hand, passenger flow exhibits a significant peak on weekends or holidays compared with weekdays.

(2)

Dwell time: According to the study presented by Zhao et al. [5], dwell time is defined as the time that passengers will spend in a specific hall. Thus, it could be represented by a probability distribution due to the diversity of passengers’ preferences. It is noted here that the dwell time is influenced by the characteristics of different halls and areas, such as the floor areas and the service counters.

(3)

Population density: The population density refers to the number of people per unit area of the area or district surrounding the transportation hub or in a city. On this basis, it restricts the passenger flow of the transportation hub to a certain extent.

2.3. Questionnaire Survey

As for the main factor of dwell time, the questionnaire surveys were adopted. In order to ensure the accuracy and rigor of the calculation results, the author conducted the questionnaire surveys while measuring the passenger flow in the transfer hall. The main contents of the questionnaire survey include classification of the respondents, dwell time, and indoor thermal comfort. The field investigations of the passenger flow patterns in the considered transportation hub’s transfer hall were carried out in January, April, and July 2022. In the questionnaire surveys, completed questionnaires were collected from 1424 respondents.

2.4. Field Measurement

To conduct passenger flow measurement in this study, passenger flow laser detection equipment was installed above each entrance/exit of the transfer hall, as shown in Figure 2. The exit/entrance (a)–(f) in Figure 2 correspond to the ones in Figure 1b. This equipment performs vertical scans. When pedestrians pass through the laser curtain, peak packets appear in the laser scan data of the specific frame, reflecting the pattern of the travelers. At any instant during the test, the number of peak packets recorded represents the number of pedestrians passing through the laser scanning surface at that instant. On this basis, real-time passenger inward/outward flow data from the transfer hall is obtained. Then, the data are transferred to an intelligent management platform for data collection and evaluation. It shall be noted here that the installed laser detection equipment records all of the people passing through the laser curtain and that there is no distinction between the number of passengers entering or leaving the transfer hall.

2.5. Separation of Passengers Entering and Leaving

Due to the need for research, the data obtained from the passenger flow laser detection equipment needs to be distinguished from the passenger entering and leaving. In the case of the transportation hub considered for investigation in this work, the surrounding land is mainly an area with residential buildings. Based on the data presented in a previous study on the hourly profile characteristics of the passenger flows inward and outward to the surrounding residential transportation hub, the data regarding passengers entering and leaving were divided in order to obtain the inbound and outbound hourly passenger flows to the transfer hall on weekdays and weekends. Overall, a cycle of 24 h should be satisfied, meaning that the total number of people entering the transfer hall should be equal to the number of people leaving over 24 h, as represented in Equation (1).

$${\sum }_{i=0}^{i=23}\mathrm{I}\mathrm{i}={\sum }_{i=0}^{i=23}\mathrm{O}\mathrm{i}$$

(1)

Ii represents the number of people entering the transfer hall per hour and Oi represents the number of people leaving the transfer hall per hour.

Figure 3 presents the flow chart of the research methodology. This study focuses on a metro transportation hub in Beijing, obtaining the dwell time and passenger flow in three different seasons through questionnaire surveys and field measurements. By combining the passenger flow characteristics indicated by the surrounding land types and Equation (1) to separate the inflow and outflow of passenger volumes in the transfer hall, a differential equation for micro-unit passenger flow density was constructed and verified using measured data. The small error margin demonstrates the model’s accuracy. The resulting findings can provide valuable guidance for the design and operational management of the fresh air system in the transfer hall.

3. Results and Discussion

3.1. Dwell Time

Figure 4 presents the demographic characteristics of the respondents who have filled out the questionnaire. It is noted that, because the considered transfer hall is part of a transportation hub for subway and bus connections, the passenger’s dwell time is generally short. As shown in Figure 4c, 73.4% of the passengers who took part in the field testing stayed in the transfer hall for less than 15 min, 23.4% of the passengers stayed for 10–15 min, and only 3.1% stayed in the hub for more than 15 min.

3.2. Passenger Flow

Figure 5 presents the profile of a one-day passenger flow at each entrance and exit of the transfer hall as recorded by the measuring equipment. It is noted that the traffic flow at the subway entrance and exit (c) is the largest among the different investigated accesses, followed by the north entrance and exit (b), the long-distance locomotive and north build-ing exit (g), exit/entrance (e), entrance (f), south entrance/exit (d) and entrance (a). The smallest passenger flow was experienced at the long-distance locomotive entrance and exit (a). This indicates that there were very few passengers taking long-distance locomotives on the day of the investigation. Additionally, the results highlight that the traffic flow of the considered transportation hub is mainly composed of three parts: traffic flow related to subway interchange personnel, traffic flow for the nearby residence, and traffic flow for the north building staff. These three traffic flows exhibit an obvious morning and evening “tide pattern.” The morning peak of passenger flow occurs from 06:00 to 09:00, and the evening peak occurs from 17:00 to 20:00. Overall, it can be reported that the passenger flow is closely related to the size of the office and the population of residential areas around the considered interchange hub.

In addition, Figure 6 presents the average hour-by-hour passenger flow profile recorded by the laser equipment detection devices on weekdays and weekends in January, April, and July. Considering that these three months allow for the characterization of passenger flow in the winter season, transitional season, and summer season, respectively, it can be seen from the results presented in Figure 6 that in January, April, and July, the average hourly passenger flow on weekdays is nearly identical, and that a similar pattern is observed for weekends. This indicates that seasonal change has little impact on passenger flow. In addition, regardless of the season investigated, there is a clear difference between the passenger flow characteristics of weekdays and weekends. On weekdays, there are distinct morning and evening peaks in passenger flow, with the highest volume occurring at 07:00, reaching up to 22,597 person/h. In contrast, the weekend passenger flow is more stable, with the highest volume also at 07:00 but only reaching 7821 person/h, and lower than the weekday peak. The daily passenger flow on weekends is around 57.2% of the daily passenger flow on weekdays.

Because the passenger flow is not related to the season, but only to the work schedule, the passenger flow characteristics of working days and weekends are further analyzed separately. Figure 7 shows the hour-by-hour passenger flow characteristics of weekdays and weekends. As shown in Figure 7a, the passenger flow increases sharply from 05:00 to 07:00 on weekdays. The passenger flow at 07:00 accounts for 21.34% of the passenger flow throughout the day, and the cumulative passenger flow by 08:00 has reached 54.30%. Then, the traffic flow decreases after 09:00. In the afternoon period, starting at 16:00, the traffic gradually increases until reaching a small peak at around 18:00. These characteristics are closely related to the working schedule on weekdays. According to the analysis of the usual working time from 09:00 to 17:00, the morning peak of passenger flow appears 2 h earlier than the working time, and the evening peak appears one hour later than the working time. It is also shown in Figure 7b that the passenger flow also starts to increase from 05:00 to 07:00 on weekends. By 08:00, the passenger flow is only 37.05% of the passenger flow throughout the day, and the distribution of passenger flow throughout the day is more consistent. This is mainly because the purpose of travel is different on weekends, and the travel time is more random. Furthermore, it can be seen from the box diagram in Figure 7 that, whether it is a working day or a weekend, the passenger flow in the morning peak period has a large change, while the passenger flow in the evening peak and other hours has a small change. This is mainly due to the strict requirements of the morning work schedule, and because there is more uncertainty in passenger flow by transportation and transfer than in the late peak.

3.3. Entering and Leaving Passenger Flow

According to the method described in Section 2.5, the entering and leaving passenger flows were further distinguished. The separation results are shown in Figure 8. From the Figure, it can be observed that, on weekdays, the entering passengers peak at 08:00, with another smaller peak at 18:00, which is 27.0% of the maximum value. The leaving passengers peak at 18:00, with another smaller peak at 08:00, which is 30.3% of the maximum value. This pattern is closely related to the typical work schedule. On weekends, the hourly variation in both entering and leaving passengers is less pronounced. The entering passengers peak between 07:00 and 08:00, reaching approximately 29.1% of the weekday peak value, and the leaving passengers peak between 18:00 and 19:00, reaching approximately 34.4% of the weekday peak value.

4. Prediction of Passenger Density

4.1. Prediction Model

Based on the concept of energy conservation and the derivation principle of the heat conduction differential equation representing the heat transfer process, the transfer hall is divided into a micro-unit of a volume of dv = dxdydz. The three sides of the micro-unit are lines across the x-axis, y-axis, and z-axis, as depicted in Figure 9. Similarly to the differential equation of heat conduction, the variation in the micro-unit passenger density in time and space can be expressed by the differential equation provided in Equation (2).

$${\displaystyle \frac{\partial P\left(x,y,z,\tau \right)}{\partial \tau }}=D\left({\displaystyle \frac{{\partial }^{2}P\left(x,y,z,\tau \right)}{{\partial x}^2}}+{\displaystyle \frac{{\partial }^{2}P\left(x,y,z,\tau \right)}{{\partial y}^2}}+{\displaystyle \frac{{\partial }^{2}P\left(x,y,z,\tau \right)}{{\partial z}^2}}\right)+S\left(x,y,z,\tau \right)$$

(2)

In the above equation, $P\left(x,y,z,\tau \right)$ denotes the passenger density; x, y, and z represent the spatial coordinates; τ denotes the time coordinates; ${\displaystyle \frac{\partial P\left(x,y,z,t\right)}{\partial \tau }}$ is the rate of change of personnel density over time; D is the diffusion coefficient that represents the diffusion rate of personnel; ${\displaystyle \frac{{\partial }^{2}P\left(x,y,z,\tau \right)}{{\partial x}^2}}+{\displaystyle \frac{{\partial }^{2}P\left(x,y,z,\tau \right)}{{\partial y}^2}}+{\displaystyle \frac{{\partial }^{2}P\left(x,y,z,\tau \right)}{{\partial z}^2}}$ denotes the change rate of passenger density with space, which is the Laplace operator of space coordinates $x$, $y$, and $z$; and $S\left(x,y,z,\tau \right)$ denotes the source term of passenger density, which represents the variation in the passenger density in the transfer hall per unit time.

To simplify the calculations and evaluation in this study, the following assumptions are made:

• The passengers in the transfer hall are assumed to be evenly distributed, with a consistent passenger density. On this basis, the passenger density $P$ is only related to the time parameter $\tau$ and is not related to the spatial parameters $x$, $y$, and $z$;
• It is assumed that all passengers in the transfer hall have the same diffusion rate, meaning that the diffusion is uniform;
• The passenger dwelling time is assumed to be consistent, thus the weighted average of the dwelling time of the passenger can be taken as the average residence time according to the survey data reported in Figure 3.

Based on the above assumptions, Equation (2) is simplified to obtain Equation (3).

$${\displaystyle \frac{dP\left(\tau \right)}{d\tau }}=S\left(\tau \right)$$

(3)

In Equation (3), $S\left(\tau \right)$ represents the source term of passenger density, which is only related to time, as mentioned above. As shown in Figure 10, the source term of passenger density at time τ in the transfer hall should be equal to the difference between the passenger density entering and leaving the transfer hall at that time plus the passenger density at the previous time $\tau -1$ [40]. This can be expressed by Equation (4).

$$S\left(\tau \right)={\displaystyle \frac{I\left(\tau \right)}{A}}-{\displaystyle \frac{O\left(\tau \right)}{A}}+{\displaystyle \frac{dP\left(\tau -1\right)}{d(\tau -1)}}$$

(4)

$${\displaystyle \frac{dP\left(\tau -1\right)}{d(\tau -1)}}={\displaystyle \frac{dP\left(\tau -1\right)}{\overline{R}}}$$

(5)

Introducing Equations (4) and (5) into Equation (3), Equation (6) is obtained as follows:

$${\displaystyle \frac{dP\left(\tau \right)}{d\tau }}={\displaystyle \frac{I\left(\tau \right)}{A}}-{\displaystyle \frac{O\left(\tau \right)}{A}}+{\displaystyle \frac{dP\left(\tau -1\right)}{\overline{R}}}$$

(6)

In Equation (6), $P(\tau -1)$ represents the passenger density at time $\tau$ − 1; $I\left(\tau \right)$ represents the number of people entering the transfer hall at time τ;$O\left(\tau \right)$ represents the number of people leaving the transfer hall at time $\tau$; $\overline{R}$ is the average dwelling time of passengers in the transfer hall; and $A$ is the area of the transfer hall.

4.2. Simplified Calculation Model

From the above analysis, it is evident that passenger density exhibits distinct characteristics on weekdays and weekends, with variations solely related to time. To simplify calculations and facilitate engineering applications, this study conducts univariate linear regression on passenger density based on weekday and weekend timeframes. The transportation hub is closed daily from 00:00 to 04:00, resulting in zero passenger flow during this period. Consequently, hourly passenger density fitting is performed for the period from 05:00 to 23:00. Considering the morning and evening peak hour characteristics, univariate linear regression is applied separately for the time intervals from 05:00 to 12:00 and 13:00 to 23:00. The linear regression equations for weekday passenger flow density (Equation (7)) and weekend passenger density (Equation (8)) are derived using univariate linear regression for these two time periods, respectively. Both equations achieve $R^2$ values of 0.92 or higher, indicating a very high degree of fit. Figure 11 illustrates the univariate linear regression results of passenger density in the transfer hall for weekdays and weekends.

$$\left\{\begin{array}{c}y=-0.0022x^5+0.0956x^4-1.6264x^3+13.374x^2-52.907x+80.613(x\in \left(\mathrm{5,12}\right),R^2=0.97)\\ y=0.0002x^5-0.014x^4+0.4531x^3-7.1366x^2+54.785x-163.74(x\in \left(\mathrm{13,23}\right),R^2=0.92)\end{array}\right.$$

(7)

$$\left\{\begin{array}{c}y=-0.0003x^4+0.013x^3-0.2117x^2+1.4706x-3.4801(x\in \left(\mathrm{5,12}\right),R^2=0.997)\\ y=0.0003x^4-0.0225x^3+0.604x^2-7.0551x+30.388(x\in \left(13,23\right),R^2=0.93)\end{array}\right.$$

(8)

4.3. Calculation Results and Model Validation

The Equations (7) and (8) were used to calculate the passenger density in the transfer hall on weekdays and weekends. At the same time, actual measurements of the passenger density in the transfer hall were conducted on March 15 (weekday) and March 18 (weekend), 2022. Figure 12 presents a comparison between the model calculated value, measured value and design value. The error between calculation results and the measured data was evaluated using the error analysis method, proposed in Willmott ‘s research [41]:

$$IA=1-{\displaystyle \frac{{\sum }_{y=0}^N{(X_{Cy}-X_{My})}^2}{{\sum }_{y=0}^N{(\left|X_{Cy}-X_{Cave}\right|+\left|X_{My}-X_{Mave}\right|)}^2}}$$

(9)

In error analysis Equation (9) ${\mathrm{X}}_{Cy}$ and ${\mathrm{X}}_{My}$ are calculation results and the measured data and ${\mathrm{X}}_{Cave}$ and ${\mathrm{X}}_{Mave}$ are calculative and measured hourly mean values, respectively. N stands for the hour, which is the natural number from 0 to 23. In the above, the range of IA can vary between 0 and 1. The IA value being 0 indicates complete inconsistency (calculative and measured results do not match each other). An IA value of 1 means that values of calculative and measured results fully match each other. In this study, the IA values of both weekdays and weekends are 0.97, as shown in Figure 12. This indicates that the proposed mathematical model describes variation, taking place in the passenger density, with a high level of accuracy.

In addition, the results highlight that the maximum passenger density calculated value and measured value are 0.56 and 0.69 people/m², respectively, at 8:00 on weekdays, which are much higher than the design value (0.38 person/m²). The passenger density at 7:00 and 8:00 during the morning peak, as well as at 18:00 and 19:00 during the evening peak on weekdays, exceeds the design capacity, which would lead to insufficient fresh air. However, for most other times on weekdays, the passenger density is below the design value. On weekends, the passenger density varies more smoothly, with the maximum calculated and measured passenger densities both occurring at 18:00, being 0.22 and 0.28 people/m², respectively, both of which are below the design value (0.38 person/m²). It is noted that the maximum passenger density reported on weekdays is 2.5 times higher than on weekends. During the design phase, to accommodate the most unfavorable conditions, the air conditioning load and equipment are usually calculated based on a higher passenger density. It is evident that there is great potential for energy savings in the air conditioning system during off-peak hours of weekdays and throughout the weekends. However, during the morning and evening peak hours on weekdays, the transfer hall usually experiences insufficient fresh air and poor air quality.

On this basis, it is recommended that the new air volume in the transfer hall be controlled by frequency conversion according to the change in carbon dioxide concentration, that is, personnel density. Additionally, based on the results attained in this study, it is recommended to adjust the number of air conditioning units or fresh air units operating according to different needs on weekdays and weekends.

5. Conclusions

Passenger density has a close relation with the energy consumption of HVAC systems in subway transfer halls. In this research, measured passenger flow data in the Beijing metro transportation hub’s transfer hall was considered for analysis and evaluation over three different seasons. On this basis, a model for passenger density calculation was established and the calculated value, measured value and design value were compared, and the following conclusions can be drawn:

(1)

The passenger flow is not affected by the outdoor weather conditions and seasonality but is closely related to the passengers’ working hours and the scale of population in the surrounding office and residential areas. Additionally, there is a major variation between the passenger flow pattern reported on weekdays and that reported on weekends. On weekdays, there is an obvious “tidal” pattern in the morning and evening periods. In this regard, the passenger traffic reaches a peak at 07:00, and the cumulative passenger flow by 08:00 has reached 54.30%. Another small peak is reported at 18:00. It is also shown that the flow of passengers on weekends is much smaller than that on weekdays, with an even change in the flow of passengers. The flow of passengers on full-day weekends is 57.2% of that on full-day weekdays.

(2)

Based on the characteristics of passenger flow and residence time in the transfer hall of the subway transportation hub, a passenger density prediction model was established. The model was validated using actual measurement data, reporting that the proposed calculation method has good precision in terms of capturing peak and average values.

(3)

The characteristics of passenger density have significant influence on the design and operation processes of fresh air systems. The passenger density at 07:00 and 08:00 during the morning peak, as well as at 18:00 and 19:00 during the evening peak on weekdays, exceeds the design capacity, leading to insufficient fresh air. However, for most other times on weekdays and all weekend hours, the passenger density is below the design value, which demonstrates a large potential for energy saving by adjusting the amount of mechanical outdoor air with the variation of actual occupancy rates.

This study provides data and method references for calculating passenger density in subway transfer halls. However, the authors also recognize the limitations of the study’s subjects and empirical data. In the future, more research could be conducted on passenger densities in transfer halls in different regions and at different scales, aiming to establish more generalizable rules.