Passenger flow status identification and control method for urban rail transit stations

Abstract

The present invention proposes a double-layer cloud model as a method for identifying passenger flow within a station, and uses the double-layer cloud model as a method for real-time identification of the predicted passenger flow status within the station, in order to perform real-time identification of the predictive passenger flow status of the station. On this basis, the passenger flow identification and dynamic passenger flow control are interactively fed back with the integrated parent cloud identification results; when performing real-time identification of the predicted passenger flow status of the station, if the control threshold is triggered, the current passenger flow status is used as input, and the passenger flow control within the station is completed through the passenger flow control model to ensure the safety of station operation. Specifically, a standard cloud of the passenger flow status identification sub-cloud model is constructed for each area of ​​the station, and then a parent standard cloud for identifying the global passenger flow status of the station is constructed by integrating the sub-standard clouds; the passenger flow information predicted by the sub-cloud to be tested is obtained by processing the passenger flow information collected by the video camera, and the parent cloud to be tested is integrated to obtain the parent cloud to be tested; the degree of membership between the parent standard cloud and the parent cloud to be tested is compared as a basis for deciding whether to implement passenger flow control measures.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban rail transit operation organization scheme optimization, and specifically relates to a method for identifying and controlling passenger flow status in an urban rail transit station. Background Art

[0002] As rail transit travel develops towards intelligence, by identifying passenger flow at urban rail transit stations, we can understand the passenger flow distribution patterns of stations and explore the travel behavior characteristics of passengers. At the same time, passenger flow information also provides a reference basis for subway scheduling strategies. In recent years, the research on subway passenger flow identification has gradually become a hot research direction.

[0003] In order to alleviate the passenger flow pressure at the station, passenger flow control has also become a necessary means for the daily operation of the station. Shen Liting [Shen Liting. Research on commuter passenger flow identification technology based on subway passenger flow card swiping data [D]. Zhejiang: Zhejiang University, 2018.] proposed to screen and identify the original subway card swiping data obtained by the automatic ticket vending and checking AFC system, and put forward suggestions and improvement measures for the passenger flow organization and station management of large passenger flow stations based on the spatial distribution of commuter passenger flow. Tang Jinjin et al. [Tang Jinjin, Hou Kaiwen, Song Wenhui. A method and system for passenger flow control in urban rail transit: CN201910547200.1 [P]. 2021-06-22.] proposed to obtain the operation diagram data based on the driving and passenger flow in and out of the station monitoring information, and to deduce the passenger flow based on the operation diagram data to generate passenger flow distribution data, so as to realize the combination of operation diagram adjustment and passenger flow control, and to control passenger flow from the three levels of line network, line and station.

[0004] The following problems exist in the methods for real-time status identification of passenger flow in urban rail transit stations and passenger flow control measures:

[0005] First, with regard to the analysis of passenger travel characteristics and passenger flow status in urban rail transit stations, the current mature research still mainly focuses on the analysis of the spatiotemporal characteristics of passenger flow in urban rail transit stations based on AFC historical passenger flow data. This part of the research has the following problems: On the one hand, most of the current passenger flow prediction methods are based on AFC historical passenger flow data and lack real-time data supplementation, resulting in poor prediction results for station passenger flow prediction methods based only on AFC historical passenger flow data; on the other hand, station passenger flow identification lacks spatial distribution information of passenger flow, making it impossible for station managers to grasp the status of passenger flow in various areas within the station.

[0006] Second, the research methods for urban rail transit passenger flow control problems, in terms of process simulation-based methods, are mostly based on theoretical analysis (such as Petri net models, system dynamics models or cellular automaton models, etc.) to establish simulation models, and simulate the process of passenger flow evolution in the station through simulation. The simulation model can restore the self-organization phenomenon of passenger flow in the station, but with the expansion of the passenger scale of the station and the increase in passenger behavior rules, the simulation time is also increasing exponentially, and most of them are based on simulation and optimization to achieve passenger flow control in the station, which is difficult to meet the real-time dynamic passenger flow control.

[0007] Third, in previous studies on passenger flow control at the station level, optimization schemes were mostly designed from the perspective of normal passenger flow congestion. Since normal passenger flow congestion has a relatively accurate regularity, such optimization schemes can achieve good results under the premise of normal passenger flow. However, since the passenger flow at urban rail transit stations is affected by a variety of accidental factors such as passengers' personal behavior habits, station environment, and pedestrians themselves, it has inherent uncertainty. The formulation of passenger flow control strategies often relies more on the manual experience of station managers, which makes the passenger flow control strategies more subjective. At the same time, there may be too much or too little flow control at the station, which reduces the service level of the station. Therefore, scientific and effective control measures are urgently needed to enhance the station's ability to cope with dynamic passenger flow risks.

[0008] Based on the above problems, how to control the passenger flow status in real time is a problem that we need to solve most at present. Summary of the invention

[0009] In order to solve the problems of inaccuracy in passenger flow detection in the existing urban rail transit operation organization, the present invention proposes a double-layer cloud model based on cloud model theory as a method for identifying passenger flow within the station. The present invention uses a double-layer cloud model as a method for real-time identification of the predicted state of passenger flow within the station, in order to perform real-time identification of the predictive passenger flow state of the station. On this basis, with the integrated parent cloud identification results, passenger flow identification and dynamic passenger flow control form interactive feedback. When performing real-time identification of the predicted state of passenger flow at the station, if the control threshold is triggered, the current passenger flow state is used as input, and the passenger flow control within the station is completed through the passenger flow control model to ensure the safety of station operation.

[0010] The present invention discloses a method for identifying and controlling passenger flow status in an urban rail transit station, comprising the following steps:

[0011] S1: Analyze the classification of station passenger flow status evaluation indicators, construct standard clouds of passenger flow status recognition sub-cloud models for each area of ​​the station, and then integrate the sub-standard clouds by weight ratio to construct a parent standard cloud for global passenger flow status recognition of the station;

[0012] S2: Based on the predicted passenger flow information, the sub-cloud to be tested and the parent cloud to be tested are established in the same way; by comparing the membership between the sub-standard cloud and the sub-cloud to be tested, the passenger flow status level of each key area is obtained; by comparing the membership between the parent standard cloud and the parent cloud to be tested, the global passenger flow status of the current station is obtained, and finally the parent cloud membership is used as the basis for deciding whether to implement passenger flow control measures;

[0013] S3: Taking the passenger flow rate entering the station as the control variable and the passenger flow status recognition result as the state variable, when the passenger flow status reaches or exceeds the threshold that triggers passenger flow control, the passenger flow rate of each entrance and exit is finely controlled based on the actual situation.

[0014] Preferably, in step S1, the cloud model is constructed as follows:

[0015] Divide a certain indicator of passenger flow status into n levels, and show that the indicator contains n-1 thresholds, and mark the n-1 thresholds as x1~x n-1 , let the set X={x1,x2,...,x n-1}, for each element in the set X, perform proportional standardization, and then take the passenger flow state clouds of the first and nth levels as half-normal clouds, among which the first-level passenger flow state cloud is a half-descending normal cloud, the nth-level passenger flow state cloud is a half-ascending normal cloud, and the passenger flow state clouds of other levels are full normal clouds.

[0016] As a preferred method, the processing method of equal proportion standardization is: That is the value after proportional normalization, which is the expected value E of the cloud model corresponding to the indicator threshold x ;max(x i ) is the maximum value of the elements in the set X; min(x i ) is the minimum value of the elements in the set X.

[0017] As a preference, the entropy E can be determined from the contribution of a quantitative value to the concept n The calculation formula for extracting the digital features of the full normal cloud model is as follows:

[0018] Therefore, the entropy E can be determined by the contribution of a quantitative value to the concept. n The calculation formula of the full normal cloud model digital feature value is shown in formula 2:

[0019] Where n is the number of grades for the facility-specific indicator;

[0020] The formula for obtaining the digital characteristics of the half-normal cloud model is:

[0021]

[0022] After the sub-cloud model parameters are calculated according to the two formulas, cloud droplets can be generated according to the normal cloud model generation algorithm. The cloud image of the sub-cloud model standard cloud can be obtained by integrating several cloud droplets.

[0023] As a preference, the parent cloud model is formed by integrating the child models according to weights, and the digital characteristic parameters of the cloud model are obtained through step S1. H k , when there are j types of facilities, the k-th type of facility corresponding to the sub-cloud to be tested is The parent cloud to be tested is F(E xF ,E nF ,H F ), then in the process of synthesizing the parent cloud from each sub-cloud, the digital characteristics of the parent cloud are calculated as follows:

[0024]

[0025] Among them, w k The weight of the sub-cloud to be tested.

[0026] As a preferred method, when calculating the membership degree of the passenger flow state corresponding to the cloud to be measured of each facility, it is also necessary to perform multiple calculations and average the average value to reduce the error, and process the average value by proportional standardization, and use the processed average value as the probability that the cloud to be measured belongs to each level of standard cloud. This probability is the weight of the cloud to be measured, and the calculation formula is:

[0027]

[0028] in:

[0029] y i.q Calculate the membership degree of the cloud under test to the i-th level passenger flow state of the standard cloud for the qth time;

[0030] val is the measured passenger flow status after proportional standardization;

[0031] p is the number of repeated calculations;

[0032] y i is the degree of membership of the cloud to be tested to the i-th level passenger flow state of the standard cloud obtained after p calculations.

[0033] As a preferred method, a double-layer standard cloud and a measured sub-cloud for passenger flow recognition are obtained. And the measured parent cloud F(E xF ,E nF ,H F ), let the kth standard subcloud be Standard parent cloud is According to the judgment results of the passenger flow status to be tested cloud and the standard cloud, a real-time identification method for the station passenger flow prediction status is established. The process is as follows:

[0034] 1) Take the cloud to be tested F(E xF ,E nF ,H F ) parameter E nF is the expectation, parameter H F Get the normal random number E' as the standard deviation n ;

[0035] 2) Take the cloud to be tested F(E xF ,E nF ,H F ) parameter E xF is the expectation, the normal random number E' obtained in step 1) n Get a normal random number x as the standard deviation practice ;

[0036] 3) Let the i-th level cloud image of the parent standard cloud be F i criterion Parameters For expectations, parameters Get normal random numbers for the standard deviation

[0037] 4) Calculate the degree of membership to the i-th level service level

[0038] 5) Repeat steps 1) to 4) N times, and set the similarity between the final cloud to be tested and the i-th level cloud image of the standard cloud as β i , then β i Can be

[0039]

[0040] Repeat steps 1) to 5) to obtain the similarity between the cloud to be tested and the standard clouds of each level, and finally perform normalization processing to calculate the percentage of similarity at each level.

[0041] Preferably, in step S3, the passenger flow rate of the station is used as the control variable, and the passenger flow state identification result is used as the state variable. When the passenger flow state reaches or exceeds the threshold value that triggers passenger flow control, closed-loop feedback control is implemented to establish dynamic passenger flow control of the station. The global passenger flow density of the station is determined by the weighted passenger flow density of each area, which is ρ(t)=∑ω i ρ i (t), where ρ(t) is the global passenger flow density of the station during period t (person / m 2 );ω i When calculating the global passenger flow density, the weight of the i-th area, ρ i (t) is the passenger flow density of the ith area in period t (person / m 2 ).

[0042] The beneficial effects of the present invention are as follows: the urban rail transit station passenger flow status identification and control solution provided by the present invention can achieve the following technical effects compared with the prior art:

[0043] 1. The double-layer cloud model can, to a certain extent, make up for the current relatively thin station operation organization and management method based only on AFC historical passenger flow data. At the same time, it is in line with the complex characteristics of urban rail transit passenger flow demand, which is time-varying and unbalanced, so that station managers can grasp the status of passenger flow in each area of ​​the station. In the construction method of the double-layer cloud model, the sub-standard cloud reflects the status classification of each area of ​​the station, and the parent standard cloud is integrated by the sub-cloud according to the weight of the service capacity allocation, reflecting the overall status classification of the station and realizing the reflection of the passenger flow status of each area. Through model identification, the current passenger flow status level of each area and the whole of the station can be accurately obtained, and actual cases have verified that the double-layer cloud model has higher accuracy, wider application range and stronger real-time performance than traditional identification methods for the station passenger flow status.

[0044] 2. The present invention constructs a real-time identification model for passenger flow status, providing a decision-making basis for the station to achieve good passenger flow control effects. At the same time, the station passenger flow control model based on feedback control is improved to achieve refined real-time passenger flow control, and form interactive feedback between the real-time identification of station passenger flow status and dynamic passenger flow control. The passenger flow control closed-loop feedback model is improved, and the passenger flow control error compensation link is introduced to adapt to the dynamic changes in the passenger flow of urban rail transit, seek the optimal control strategy for the passenger flow entering the station, and give full play to the station's distribution capacity while ensuring the safety of the passenger flow status, which is conducive to the comprehensive improvement of the passenger service level and operational safety of urban rail transit stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of standard cloud diagrams of various levels of accommodation type facilities provided by an embodiment of the present invention;

[0046] Figure 2 It is a schematic diagram of a standard cloud diagram of each level of facilities of a class type provided by an embodiment of the present invention;

[0047] Figure 3 It is a station parent standard cloud diagram provided by an embodiment of the present invention;

[0048] Figure 4 is a schematic diagram of the service level span of different types of facilities provided by an embodiment of the present invention;

[0049] Figure 5 Picture to Figure 8 The parent standard cloud of the passenger flow status of the station and the parent cloud diagram of the four passenger flow situations to be tested provided by the embodiment of the present invention;

[0050] Fig. 9 is the passenger flow speed at each entrance and exit without control provided by the embodiment of the present invention;

[0051] Fig.10 It is a time-varying graph of the global passenger flow density of a station without control provided by an embodiment of the present invention;

[0052] Fig.11 It is a comparison diagram of passenger flow speeds at various entrances and exits under passenger flow control provided by an embodiment of the present invention;

[0053] Fig.12 is a time-varying graph of the global passenger flow density of a station under different passenger flow control methods provided by an embodiment of the present invention DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the present application will be described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are partial embodiments of the present application, not all embodiments. Instead, they only represent selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0055] 1. Real-time identification of passenger flow status based on the double-layer cloud model

[0056] S1: Construction of cloud model

[0057] Assuming that a certain indicator of the passenger flow status of a facility can be divided into n levels, it is obvious that the indicator should contain n-1 thresholds, and the n-1 thresholds are marked as x1~x n-1 , let the set X={x1,x2,...,x n-1}, for each element in the set X, perform proportional normalization processing using Formula 1.

[0058]

[0059] in:

[0060] ——x i The value after proportional normalization is the expected value E of the cloud model corresponding to the indicator threshold. x ;

[0061] max(x i )——the maximum value of elements in set X;

[0062] min(x i )——the minimum value of the elements in the set X.

[0063] After proportional normalization, the expected value of the minimum element is 0, the expected value of the maximum element is 1, and the expected values ​​of other elements are between 0 and 1. Therefore, the passenger flow state cloud of the first level (corresponding to the minimum threshold of the indicator) and the nth level (corresponding to the maximum indicator) are half-normal clouds, among which the first-level passenger flow state cloud is a half-descending normal cloud, the nth-level passenger flow state cloud is a half-ascending normal cloud, and the passenger flow state clouds of other levels are full normal clouds.

[0064] Since the normal cloud model has normal distribution characteristics, similar to the interval and area relationship of normal distribution, according to the contribution of a quantitative value to the concept, assuming that the sum of the contribution values ​​of all elements in set X to the concept is 100%, the following characteristics are obtained:

[0065] 1) Interval [E x -0.674E n ,E x +0.674E n ]The sum of the contribution values ​​to the concept is 50.0%;

[0066] 2) Interval [E x -E n ,E x +E n ]The total contribution value to the concept is 68.3%;

[0067] 3) Interval [E x -2E n ,E x +2E n ]The total contribution value to the concept is 95.4%;

[0068] 4) Interval [E x -3E n ,E x +3E n ]The total contribution value to the concept is 99.7%.

[0069] This is true for all fully normal clouds. Therefore, the entropy E can be determined by the contribution of a quantitative value to the concept. n The calculation formula of the full normal cloud model digital feature value is shown in Formula 2.

[0070]

[0071] Where n is the number of grades for the facility-specific indicator.

[0072] The formula for obtaining the digital characteristic value of the half-normal cloud model is shown in Equation 3

[0073]

[0074] After calculating the sub-cloud model parameters according to Formula 2 and Formula 3, cloud droplets can be generated according to the normal cloud model generation algorithm. The cloud image of the sub-cloud model standard cloud can be obtained by integrating several cloud droplets.

[0075] S2: Double-layer cloud model for real-time identification of passenger flow at stations

[0076] The parent cloud model is relative to the child cloud model. The child cloud model in the previous article refers to the cloud model generated by grading the service levels of various indicators of various facilities in the station. It is the quantification of the service level grading of each facility. The parent cloud model is the integration of child models according to weights, which is the quantification of the service level grading of the station.

[0077] The passenger flow status of the facility is identified by the camera in the station, and the service level is defined and standardized in proportion. Combining the above formula 2 and formula 3, the digital characteristic parameters of the cloud model can be obtained respectively. H k Assume that there are j types of facilities in the station, and the sub-cloud to be tested corresponding to the k-th type of facility is The parent cloud to be tested is F(E xF ,E nF ,H F ), then in the process of synthesizing the parent cloud from each sub-cloud, the digital characteristics of the parent cloud can be determined by Formula 4.

[0078] The digital characteristics of the parent cloud can be determined by Equation 4.

[0079]

[0080] Among them, w k The weight of the sub-cloud to be tested.

[0081] The station cloud to be measured is obtained by the cloud to be measured at each facility through formula (4-4). When calculating the membership degree of the cloud to be measured at each facility corresponding to the passenger flow state, it is necessary to perform multiple calculations and average the average to reduce the error, and process the average value by proportional standardization. The average value after processing is used as the probability that the cloud to be measured belongs to each level of standard cloud. This probability is the weight of the cloud to be measured, as shown in formula 5.

[0082]

[0083] in:

[0084] y i.q ——Calculate the membership degree of the cloud under test to the i-th level passenger flow state of the standard cloud for the qth time;

[0085] val——actual passenger flow status after proportional standardization;

[0086] p——number of repeated calculations;

[0087] y i——The degree of membership of the cloud to be tested to the i-th level passenger flow state of the standard cloud obtained after p calculations.

[0088] It should be pointed out that since the bottleneck area in the station is more likely to affect the service level of the entire station, and the bottleneck is related to the service capacity of the facility, this paper proposes a method for determining the weight of the standard cloud based on the static service capacity of the facility, as shown in Equation 6. This method solves the problem of weight differences in synthesizing the parent standard cloud from the sub-standard cloud.

[0089]

[0090] in:

[0091] w k criterion ——The weight of the sub-standard cloud corresponding to the k-th type of facility in the synthesis of the parent standard cloud;

[0092] C k ——Service capacity of category k facilities (person / h).

[0093] After calculating the standard cloud weight, the parent standard cloud can be integrated using formula (4) to obtain the weighted station passenger flow recognition standard cloud.

[0094] At this point, we can get the double-layer standard cloud and measured sub-cloud for passenger flow recognition. And the measured parent cloud F(E xF ,E nF ,H F ), let the kth standard subcloud be Standard parent cloud is Next, we need to establish a real-time identification method for the station passenger flow prediction status based on the judgment results of the passenger flow status test cloud and the standard cloud. Taking the parent cloud as an example, the identification method is as follows:

[0095] 1) Take the cloud to be tested F(E xF ,E nF ,H F ) parameter E nF is the expectation, parameter H F Get the normal random number E' as the standard deviation n ;

[0096] 2) Take the cloud to be tested F(E xF ,E nF ,H F ) parameter E xF is the expectation, the normal random number E' obtained in step 1) n Get a normal random number x as the standard deviation practice ;

[0097] 3) Let the i-th level cloud image of the parent standard cloud be F i criterion Parameters For expectations, parameters Get normal random numbers for the standard deviation

[0098] 4) Calculate the degree of membership to the i-th level service level

[0099] 5) Repeat steps 1) to 4) N times, and set the similarity between the final cloud to be tested and the i-th level cloud image of the standard cloud as β i , then β i It can be determined by formula 7;

[0100]

[0101] 6) Repeat steps 1) to 5) to obtain the similarity between the cloud to be tested and the standard clouds of each level, and finally perform normalization processing to calculate the percentage of similarity at each level.

[0102] 2. Passenger flow control model considering the maximum distribution capacity

[0103] After the station passenger flow status is identified, when the control threshold is triggered, it is necessary to control the passenger flow in the station. Based on the previous analysis of the station passenger flow status and passenger flow control, this section proposes an urban rail transit station passenger flow control model that takes into account refined passenger flow control to meet the service quality and safety requirements of station operations. Therefore, in the simulation case simulation of the present invention, the station passenger flow density, passenger flow and passenger flow in each direction are used as state variables, and the passenger flow velocity entering the station is used as the control variable to implement closed-loop feedback control to establish a station dynamic passenger flow control strategy.

[0104] S1: Passenger flow control model

[0105] The closed-loop feedback control is based on the state variable, that is, the global passenger flow density in the station and the safe passenger flow density threshold to evaluate the passenger flow safety status, thereby adjusting the flow rate of the incoming passenger flow. The global passenger flow density of the station is determined by the weighted passenger flow density of each area, as shown in Equation 8.

[0106] ρ(t)=∑ω i ρ i (t) (8)

[0107] in:

[0108] ρ(t)——the overall passenger flow density of the station during period t (person / m 2 );

[0109] ω i ——When calculating the global passenger flow density, the weight of the ith area is determined according to formula (4-6);

[0110] ρ i (t)——passenger flow density of the ith area during period t (person / m 2 ).

[0111] Assume that a station has only one entrance and exit, and the passenger flow rate entering the station is as shown in Formula 9.

[0112]

[0113] in:

[0114] v(t)——the passenger flow rate entering the station during time period t (person / h);

[0115] ρ * ——Passenger flow density threshold that triggers flow control (person / m 2 );

[0116] ——Adjustment coefficient.

[0117] The above passenger flow control model can be regarded as a proportional regulation feedback control model. Simple proportional regulation may have steady-state errors, which is reflected in the actual passenger flow control, that is, the passenger flow density will be somewhat different from the expected passenger flow density, resulting in the station failing to fully utilize its distribution capacity. From the perspective of automatic control, it is possible to consider introducing integral control to eliminate steady-state errors, thereby constructing the passenger flow velocity entering the station as shown in Equation 10.

[0118]

[0119] in:

[0120] —— adjustment coefficient;

[0121] Δρ i ——Δρ i =ρ * -ρ(i).

[0122] It can be seen from formula 10 that in the feedback control model, the passenger flow rate in period t is determined by the passenger flow density in period t-1, that is, there is a time lag in passenger flow control, which cannot meet the timeliness requirements of passenger flow control. In order to reduce the time lag effect and control deviation of feedback control, an improved feedback control method based on predicted passenger flow density is proposed, as shown in formula 11.

[0123]

[0124] in:

[0125] ρ p (t)——predicted passenger flow density during time period t (person / m 2 ).

[0126] S2: Determination of adjustment coefficient

[0127] Adjustment coefficient Indicates the magnitude of the inbound passenger flow rate that needs to be adjusted when the global passenger flow density changes by one unit. It represents the influence of the cumulative error of global passenger flow density on the flow rate of incoming passenger flow. Assuming that all passengers have the same speed and are evenly distributed, the changes in passenger flow and passenger flow density in time period t are shown in equations 12 and 13:

[0128] Δq(t)=v(t)Δt (12)

[0129]

[0130] in:

[0131] Δq(t)——passenger flow into the station (person-times);

[0132] Δt——passenger flow control time granularity (min);

[0133] Δρ(t)——Change in passenger flow density at the station (person / m 2 );

[0134] S——Usable area of ​​the station (m 2 ).

[0135] According to the adjustment factor The definition of , its value can be determined according to formula (4-14).

[0136]

[0137] Adjustment coefficient According to formula 14, the parameters It needs to be determined according to the actual situation. Next, the passenger flow control model in this paper is extended to the situation of transfer stations.

[0138] In a transfer station, the incoming passenger flow will go to different lines according to the destination platform, and usually the passenger flow of different lines is different, which will have different impacts on the passenger flow density of the station. Now suppose a transfer station is r-line transfer, and the proportion of incoming passenger flow to each line in time period t is β1(t), β2(t),...,β r (t), when the time granularity is small, the proportion of the passenger flow direction in time period t can be obtained from the direction of the passenger flow in the previous time period, as shown in Formula 15.

[0139]

[0140] in:

[0141] qi (t-1)——the passenger flow (number of passengers) entering the station with the destination route i during the period t-1;

[0142] q 总 (t-1)——Total passenger flow entering the station during the t-1 period (person-times).

[0143] Assume that when controlling the passenger flow entering the station during period t, the adjustment parameters of each line are and According to the definition, the value of each parameter can be determined according to Equation 16 and Equation 17.

[0144]

[0145] Assume that the station has s available entrances and exits. Now merge all the entrances and exits, that is, regard the generation of incoming passenger flow as a virtual passenger flow source. At this time, it is transformed into the problem of inbound passenger flow velocity control at a single entrance and exit, and passenger flow control can be performed according to formula 11. Under ideal conditions, it is necessary to control the passenger flow of different lines in the inbound passenger flow differently according to the specific situation of the station passenger flow density. The station controls the inbound passenger flow velocity to the jth line in time period t as shown in formula 18.

[0146]

[0147] In theory, passenger flow control should be carried out on the passenger flow of different lines separately. However, since it is impossible to distinguish the specific flow direction of the passenger flow when entering the station, in order to ensure the safety of passenger transportation at the station, the passenger flow rate control of each entrance is temporarily set to a safe value, which is set to but:

[0148]

[0149] Since the passenger flow entering the station at each entrance and exit is not equal, it is necessary to make differentiated flow rate adjustments for the passenger flow entering the station at different entrances and exits. Now, we consider determining the weight based on the number of passengers entering the station at each entrance and exit to obtain the expected passenger flow rate of each entrance and exit, as shown in Formula 20.

[0150]

[0151] in:

[0152] v k (t)——the expected passenger flow rate of the kth entrance and exit in time period t (person / h);

[0153] q k (t)——The passenger flow (people) at the kth entrance and exit during time period t.

[0154] Finally, by comparing the passenger flow speed with that of the previous control period, it can be determined whether the target entrance and exit is controlled during this period: if the passenger flow speed of the entrance and exit in the previous period is lower than the expected flow speed during this period, the inbound passenger flow speed of the entrance and exit will be controlled; otherwise, the inbound passenger flow speed of the entrance and exit will not be adjusted.

[0155] 3. Application analysis of case optimization methods

[0156] The following is a practical example of a subway station in Shenzhen. A double-layer cloud model is used to identify the passenger flow status of urban rail transit stations in various regions and the entire station. First, a standard cloud of the sub-cloud model is constructed for each facility in each station, and a standard cloud of the parent cloud model is constructed for the entire station in combination with the service capabilities of the facilities. Secondly, by comparing the similarity and membership between the standard cloud and the cloud to be tested, the passenger flow status level of each region is obtained; finally, by integrating the passenger flow status of each region and comparing it with the parent standard cloud, the overall passenger flow status level of the station is obtained. In terms of passenger flow control, the passenger flow control error compensation link is introduced to eliminate the problem of insufficient or excessive control strength that may exist in the original passenger flow control model. When the passenger flow status reaches or exceeds the threshold that triggers passenger flow control, the passenger flow velocity of each entrance and exit is finely controlled in combination with the actual situation to form interactive feedback between passenger flow identification and passenger flow control.

[0157] S1, Case Background

[0158] This station is one of the transfer stations for Shenzhen Metro Line 3 and Line 4. It has a huge passenger flow, with an average daily passenger flow of 40,000 people. The super peak hour passenger flow during the evening peak period can be close to 10,000 people. The passenger flow on weekdays is larger than that on weekends. Line 3 is arranged in an east-west direction, and Line 4 is arranged in a north-south direction. The platform has 9 entrances and exits, of which E, F1, and F2 entrances and exits lead to the Line 3 station hall, and A1, A2, B, C1, C2, and D entrances and exits lead to the Line 4 station hall. The underground first floor of the station is the Line 3 station hall level and the Line 4 station hall level. The two station halls are connected by an L-shaped transfer passage with a length of about 130 meters; the underground second floor is the Line 4 platform level, and the platform form is a side platform. The underground third floor is the Line 3 platform level, and the platform form is an island platform. The station hall of Line 3 is equipped with 4 security inspection machines, 8 entry gates, 9 exit gates, and 2 sets of two-way escalators; the station hall of Line 4 is equipped with 4 security inspection machines, 6 entry gates, 11 exit gates, 4 one-way escalators, and 4 staircases.

[0159] AnyLogic has the advantages of reflecting the movement characteristics of passengers and group self-organization phenomena, realizing the physical environment of various situations, fitting the actual operation of subway stations, and having visual data analysis tools. Therefore, AnyLogic was used as the simulation software. Based on the platform data characteristics, a two-layer cloud model was used to verify the real-time status recognition capability of passenger flow and the interactive feedback effect of passenger flow recognition and passenger flow control.

[0160] S2, real-time identification of passenger flow status

[0161] The real-time identification of platform passenger flow prediction status includes three aspects: the construction of the standard cloud, the construction of the cloud to be tested, and the determination of the membership of the cloud to be tested with respect to the standard cloud.

[0162] According to the classification of the accommodation and passing facilities shown in Tables 1 and 2, and after processing according to the proportional standard, the digital features of the standard clouds at each level are shown in Table 3. Among them, the values ​​of the cloud digital features greater than or equal to 1 after the proportional standard processing are set to 1, and the values ​​less than or equal to 0 are set to 0.

[0163] Table 1 Classification of status of accommodation facilities

[0164]

[0165] Table 2 Classification of entrance and exit status of through-type facilities

[0166]

[0167] Table 3 Digital characteristics of standard clouds at all levels of stations

[0168]

[0169] According to the digital characteristics of the cloud model of the known station data, the number of cloud drops is set to 1000, and the standard cloud maps of different types of facilities at all levels can be obtained, such as Figure 1 , Figure 2 shown.

[0170] The above constructs the sub-standard cloud of the station. The following constructs the parent standard cloud based on the actual layout of the subway station facilities. Since it is a dual-line transfer station for Line 3 and Line 4, the facilities can be divided into 9 sub-facilities, including Line 3 entrance channel, Line 3 station hall, Line 3 escalator, Line 3 platform, Line 4 entrance channel, Line 4 station hall, Line 4 escalator, Line 4 platform, and transfer channel.

[0171] According to the above analysis of the basic digital characteristics of the double-layer cloud model, the contribution of each sub-facility to the parent standard cloud is determined by the service capacity of each sub-facility, and the cumulative service capacity of each sub-facility is calculated by combining Formula 6. Assuming that the contribution value of the two major types of sub-facilities to the parent standard cloud is 50%, the cumulative passing capacity of each sub-facility at the Children's Palace Station and the weight to the parent standard cloud are calculated as shown in Tables 4 and 5. Then, Tables 4 and 5 are combined with Formula 1 to obtain the digital characteristics of the parent standard cloud of the station as shown in Table 6.

[0172] Table 4 Calculation value of the cumulative service capacity of each type of sub-facilities at a station

[0173]

[0174] Table 5 Calculation value of the cumulative service capacity of each type of sub-facilities at a station

[0175]

[0176]

[0177] Table 6 Digital characteristics of the parent standard cloud of a station

[0178]

[0179] Finally, the station parent standard is obtained as shown in the figure below: Figure 3 Shown

[0180] S3, simulation parameter setting

[0181] The parameters of the station facilities and equipment are set with reference to the AnyLogic simulation model of the station. The passenger flow data is obtained from the images of the station passenger flow collected by the camera, and the passenger flow is obtained through the deep learning algorithm recognition situation. The passenger flow parameters are determined based on the passenger flow OD data during the evening peak period on weekdays on July 30, 2021. The passenger flow parameters are calculated by counting the passenger flow in and out of the station and the transfer passenger flow.

[0182] 1) Passenger flow in and out of the station. The passenger flow in and out of the station during the peak hour on the weekday evening on July 30, 2021 is shown in Table 7.

[0183] Table 7 Passenger flow in and out of the station

[0184]

[0185] 2) Transfer passenger flow. The hourly transfer passenger flow during the evening peak on weekdays on July 30, 2021 is shown in Table 8.

[0186] Table 8 Transfer passenger flow

[0187]

[0188]

[0189] 3) Input parameters for pedestrians entering and getting off the train. According to the passenger flow entering the station, the passenger flow entering the station is evenly distributed to 9 entrances and exits, and the arrival rate can be set to 986 people / hour; according to the passenger flow leaving the station and transferring, the passenger arrival rate of trains in two lines and four directions is set as shown in Table 9.

[0190] Table 9 Arrival rate of passengers getting off the bus

[0191]

[0192] S4, simulation results

[0193] Since the passenger flow on the forecast day is extremely large, in order to obtain real-time identification results of passenger flow forecast status under different passenger flow situations, the basic passenger flow parameter setting base for simulation analysis is 0.25 times, 0.5 times, 0.75 times and 1 times the passenger flow on that day, corresponding to scenario one, scenario two, and so on.

[0194] The simulation results show that in scenarios 1 and 2, the passenger flow of the station is relatively smooth, and there is no passenger congestion; in scenario 3, there is a small-scale passenger congestion problem in the north of the Line 4 station hall and the connection between the transfer passage and the Line 3 station hall, which poses a certain operational safety risk; in scenario 4, there is a large-scale passenger congestion problem at the north and south entrance gates of the Line 4 station hall, the east and west entrance gates of the Line 3 station hall, the escalator of the Line 3, and the connection between the transfer passage and Line 3, and the station operation is in a high-risk state. The passenger flow density is shown in Table 10.

[0195] Table 10 Passenger flow density of stations under different scenarios

[0196]

[0197]

[0198] The data in Table 10 are processed by the above-mentioned proportional standards such as Formula 1, Formula 2, and Formula 3, and the calculation results are shown in Table 11.

[0199] Table 11 Passenger flow density after proportion standard processing

[0200]

[0201] The data in Table 11 is combined with Formula 5. The number of repeated calculations is set to 200 to reduce the error. The weights of the passenger flow status of each area belonging to each level of status are calculated, as shown in Tables 12 and 13.

[0202] Table 12 Facilities by Class

[0203]

[0204]

[0205] Table 13 Accommodation facilities

[0206]

[0207]

[0208] Tables 12 and 13 reflect the passenger flow status of different facilities and areas, which can be used as the basis for passenger flow control in different facilities or areas. From the calculation results, it can be seen that the service level of through facilities does not change significantly under different circumstances, while the service level of accommodation facilities changes significantly. This is because when referring to the "U.S. Public Transportation Capacity and Service Quality Manual (Second Edition)" (TCQSM) to classify the service level of urban rail transit station facilities, the span of the D and F levels of through facilities is relatively large, such as Figure 4 shown.

[0209] In order to solve the problem that the overall passenger flow status in the station is difficult to determine due to the differences in the classification of service levels of different types of facilities, the child clouds to be tested are integrated into a parent cloud, and the passenger flow density status in the station is determined by calculating the degree of membership of the parent cloud to be tested to the parent standard cloud.

[0210] Substitute the calculated measured weight value into Formula 2 to synthesize the cloud to be measured, and finally obtain the digital characteristics of the synthetic cloud to be measured as shown in Table 14.

[0211] Table 14 Digital characteristics of the cloud to be tested

[0212]

[0213] According to the cloud digital features in Table 14, the cloud map of the cloud to be tested is constructed and compared with the cloud map of the standard cloud. Figures 5 to 8 shown.

[0214] Finally, the number of repeated calculations was set to 200 to reduce the error, and the similarity between the cloud to be tested and the standard clouds of each level was calculated. The calculation results are shown in Table 15.

[0215] Table 15 Similarity calculation results between the parent cloud to be tested and the standard parent cloud

[0216]

[0217] In case 1, the station passenger flow state and the C-level state membership reached 69.40%; in case 2, the station passenger flow state and the D-level state membership reached 98.77%; in cases 3 and 4, the station passenger flow state and the E-level state membership reached 95.24% and 98.62% respectively. In summary, it can be considered that the double-layer cloud model has a good recognition effect on the station passenger flow state, and the passenger flow state recognition combined with passenger flow information prediction can effectively serve as a basis for the subsequent station passenger flow control.

[0218] 4. Refined dynamic passenger flow control

[0219] The control object of the station passenger flow dynamic control method in this paper is the passenger flow rate at each entrance and exit. In the actual station passenger transportation process, the passenger flow rate control at each entrance and exit is achieved by setting railings and iron horses outside the entrances and exits, restricting the walking direction of the entrances and exits, and adjusting the number of ticket machines and gates. It is a dynamic process, and AnyLogic simulation is difficult to achieve dynamic facility change adjustment. Therefore, the simulation process achieves flow control at each entrance and exit by reducing the passenger flow demand generation rate.

[0220] The facility parameters for the subway station passenger flow control simulation refer to the settings in the power requirements. To simulate a large passenger flow situation, the passenger flow parameters are set to 1.3 times the actual passenger flow on the forecast day. The control time granularity is consistent with the passenger flow information forecast time granularity, which is 5 minutes. The adjustment coefficient According to formula 14, the value is 975. According to multiple simulation experiments, the value range is This paper takes the value of 80. The passenger flow data used above is obtained by recognizing the image information collected by the camera.

[0221] The station has a total of 8 entrances and exits, which can be divided into three groups based on urban roads. Among them, entrances F1 and F2 are close to Lianhuashan Park and have the smallest passenger flow; entrances B, A1, and A2 are close to the Children's Palace and have slightly more passenger flow; entrances E, C1, C2, and D are close to the commercial area and have the largest passenger flow. According to actual conditions and field observation statistics, the proportion of passenger flow entering the station at each entrance and exit is shown in Table 16.

[0222] Table 16 Passenger flow ratio of each entrance and exit during the evening peak period in the station

[0223]

[0224]

[0225] According to the simulation results above, there is already a certain degree of passenger congestion in the passenger flow in situation three. At this time, the membership degree of the global passenger flow state level E of the station exceeds 95%, reaching 95.24%. It can be considered that the passenger flow is in a high-risk state at this time. The corresponding passenger flow density is obtained by weighted summing the passenger flow density corresponding to each facility level E with the weights calculated in Tables 5-10 and 5-11. The calculated result is 2.22 people / m 2 , taking this value as the corresponding passenger flow density of the station carrying capacity limit, according to the "Urban Rail Transit Operation Safety Management Specifications", when the passenger flow reaches 70% of the station carrying capacity, the flow control measures should be initiated. Therefore, the Children's Palace Station triggers the passenger flow control threshold ρ * =1.56 people / m 2The simulation duration is set to 1 hour, corresponding to the evening peak hour. Without taking passenger flow control measures, the passenger flow speed at each entrance and exit of the station and the overall passenger flow density of the station change with time as shown in the figure below: Fig. 9 and Fig.10 shown.

[0226] The dynamic passenger flow control method of rail transit stations under occasional congestion proposed by Yu Dandan and other scholars is compared with the control method in this paper for the above scenarios. The comparison chart of passenger flow speed at each entrance and exit of the station is obtained, as shown in the figure below: Fig.11 As shown in the figure; and the comparison chart of the change of the overall passenger flow density of the station over time, as shown in the figure Fig.12 shown.

[0227] From the comparative analysis of the two discounted changes, it can be obtained that both passenger flow control methods can complete the dynamic control of the station passenger flow. Starting from the 10th minute, the global passenger flow density of the station fluctuates around the critical value. On the one hand, the passenger flow control method proposed in the present invention has never exceeded the passenger flow density threshold, while the passenger flow control method proposed by scholar Yu Dandan slightly exceeded the passenger flow density threshold in time periods 9 and 10. Therefore, it can be concluded that the passenger flow control method proposed under this model is safer; on the other hand, the passenger flow control method proposed in the present invention is closer to the passenger flow density threshold, so that the station's distribution capacity can be fully utilized within a safe passenger flow density range. Taking all factors into consideration, it is believed that the improved passenger flow control in this article can fully utilize the station's distribution capacity within a safe range, better deal with the passenger transportation risks caused by station passenger congestion, and ensure the safety of station operations.

[0228] 5. Summary

[0229] The present invention takes passenger flow density as a descriptive index of passenger flow status, and proposes a construction method for real-time passenger flow identification of a double-layer cloud model. The sub-standard cloud reflects the status classification of each area of ​​the station. The parent standard cloud is integrated by the sub-cloud according to the weight of service capacity allocation, reflecting the overall status classification of the station. The cloud to be tested is obtained by predicting passenger flow information, which is a reflection of the passenger flow status of each area. Through the identification of the double-layer cloud model, the current passenger flow status level of each area and the whole station can be obtained. At the same time, in terms of dynamic passenger flow control, the passenger flow control error compensation link is introduced to eliminate the problem that the original passenger flow control model may have too much or too little control strength, so that the station can give full play to its distribution capacity within a safe range; secondly, the passenger flow control model is extended to the transfer station, and the passenger flow at each entrance is used as a different control variable, in order to achieve refined station dynamic passenger flow control; finally, through the analysis of the actual operation control strategy, it is determined that 70% of the passenger flow density corresponding to the overall passenger flow status level E of the station is used as the threshold for triggering passenger flow control.

[0230] Taking a Shenzhen Metro station as an example, AnyLogic simulation verifies that the double-layer cloud model proposed in this paper has a good recognition effect on the passenger flow status of the station; and verifies that the refined dynamic passenger flow control model improved in this paper is superior to the original model, which can maximize the passenger flow distribution capacity of the station within a safe range, and at the same time, it can more strictly limit the passenger flow density of the station to below the passenger flow density threshold to ensure the safety of station operation. In summary, it is verified that the double-layer cloud model has a good recognition effect on the passenger flow status of the station, and the passenger flow status recognition combined with passenger flow information prediction can effectively serve as the basis for the subsequent passenger flow control of the station.

[0231] The embodiments described above are only preferred embodiments of the present invention and do not limit the present invention in any form. Any technician familiar with the art, without departing from the scope of the technical solution of the present invention, can make more possible changes and modifications to the technical solution of the present invention using the technical content disclosed above, or the modifications are all equivalent embodiments of the present invention. Therefore, all equivalent and equivalent changes made according to the ideas of the present invention without departing from the content of the technical solution of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for identifying and controlling passenger flow status in urban rail transit stations, characterized in that: The following steps are involved: S1: Analyze the classification of station passenger flow status evaluation indicators, construct standard clouds of passenger flow status recognition sub-cloud models for each area of ​​the station, and then integrate the sub-standard clouds by weight ratio to construct a parent standard cloud for global passenger flow status recognition of the station; S2: Based on the predicted passenger flow information, the sub-cloud to be tested and the parent cloud to be tested are established in the same way; by comparing the membership between the sub-standard cloud and the sub-cloud to be tested, the passenger flow status level of each key area is obtained; by comparing the membership between the parent standard cloud and the parent cloud to be tested, the global passenger flow status of the current station is obtained, and finally the parent cloud membership is used as the basis for deciding whether to implement passenger flow control measures; S3: Taking the passenger flow rate of the station as the control variable and the passenger flow state recognition result as the state variable, when the passenger flow state reaches or exceeds the threshold that triggers passenger flow control, the passenger flow rate of each entrance and exit is finely controlled according to the actual situation; The parent cloud model is formed by integrating the child cloud models according to their weights. The digital characteristic parameters of the cloud model are obtained through step S1. , when there are j types of facilities, the k-th type of facility corresponding to the sub-cloud to be tested is , let the parent cloud to be tested be , then in the process of synthesizing the parent cloud from each sub-cloud, the digital characteristics of the parent cloud are calculated as follows: in, is the weight of the sub-cloud to be tested; when calculating the membership degree of the cloud to be tested of each facility corresponding to the passenger flow state, it is necessary to perform multiple calculations and average the average value to reduce the error, and process the average value by equal proportion standardization. The average value after processing is used as the probability that the cloud to be tested belongs to the standard cloud of each level. This probability is the weight of the cloud to be tested, and the calculation formula is: in: For the The cloud under test belongs to the standard cloud. The membership degree of the passenger flow state; It is the measured passenger flow status after proportional standardization; To calculate the number of repetitions; For passing The cloud to be measured obtained by the calculation belongs to the standard cloud The membership degree of the passenger flow state.

2. The method for identifying and controlling passenger flow status of urban rail transit stations according to claim 1 is characterized in that: In step S1, the cloud model is constructed as follows: Divide a certain indicator of passenger flow status into n levels, showing that the indicator contains n-1 thresholds. The threshold is marked , another set , for the set Each element is standardized in equal proportion, and then the passenger flow state cloud of the first level and the nth level is taken as a half-normal cloud, among which the passenger flow state cloud of the first level is a half-descending normal cloud, the passenger flow state cloud of the nth level is a half-ascending normal cloud, and the passenger flow state clouds of other levels are full normal clouds.

3. The method for identifying and controlling passenger flow status of urban rail transit stations according to claim 2, characterized in that: The processing method of equal proportion standardization is: ,in is the value after proportional normalization, which is the expected value of the indicator threshold corresponding to the cloud model ; For collection Maximum value of inner elements; For collection Minimum value of inner elements.

4. The method for identifying and controlling passenger flow status of urban rail transit stations according to claim 2, characterized in that: The entropy can be determined by the contribution of a quantitative value to the concept The calculation formula for extracting the digital features of the full normal cloud model is as follows: ,in, H is the rating number of the facility-specific indicator; i is the super entropy of the i-th passenger flow state; The formula for obtaining the numerical characteristics of the half-normal cloud model is: , where H1 and H n represents the super entropy of the 1st and nth levels; After the sub-cloud model parameters are calculated according to the two formulas, cloud droplets can be generated according to the normal cloud model generation algorithm. The cloud image of the sub-cloud model standard cloud can be obtained by integrating several cloud droplets.

5. The method for identifying and controlling passenger flow status of urban rail transit stations according to claim 1, characterized in that: Obtain a double-layer standard cloud and measured sub-cloud for passenger flow recognition And measured parent cloud , set The standard subcloud is , the standard parent cloud is ,According to the judgment results of the passenger flow status to be tested cloud and the standard cloud, a real-time identification method for the station passenger flow prediction status is established. The process is as follows: 1) The cloud to be tested Parameters For expectations, parameters Get normal random numbers for the standard deviation ; 2) The cloud to be tested Parameters is the expectation, the normal random number obtained in step 1) Get normal random numbers for the standard deviation ; 3) Set the parent standard cloud The cloud image is ,by Parameters For expectations, parameters Get normal random numbers for the standard deviation ; 4) Calculate the number of Level of service membership ; 5) Repetition In the next steps 1) to 4), let the final cloud to be tested be the same as the standard cloud. The similarity of the level cloud map is ,but Can be Repeat steps 1) to 5) to obtain the similarity between the cloud to be tested and the standard clouds of each level, and finally perform normalization to calculate the percentage of similarity at each level.

6. The method for identifying and controlling passenger flow status of urban rail transit stations according to claim 1, characterized in that: In step S3, the passenger flow rate at the station is used as the control variable, and the passenger flow state recognition result is used as the state variable. When the passenger flow state reaches or exceeds the threshold that triggers passenger flow control, closed-loop feedback control is implemented to establish dynamic passenger flow control at the station. The global passenger flow density of the station is determined by the weighted passenger flow density of each area, which is ,in for Global passenger flow density at the station during the period ; To calculate the global passenger flow density, The weight of the region, for Time period Passenger flow density in each area .

Images