CN108280540B - Method and device for predicting short-time passenger flow state of rail transit station - Google Patents

Abstract

A method and device for predicting short-term passenger flow state of a rail transit station, the method comprising: generating passenger flow and speed time series according to the acquired passenger flow and passenger flow speed of a rail transit target station and upstream and downstream adjacent stations of the target station; The first-order difference operation is performed on the speed time series to obtain the stationary time series; the vector autoregression model is constructed by using the historical sample data of the stationary time series; the error correction term is obtained according to the historical sample data of the original passenger flow, passenger flow and speed time series in the same period ; Establish a vector error correction model for the passenger flow and passenger flow speed of the target site; calculate the predicted value of the passenger flow and passenger flow speed of the target site within the time interval t. The method and device for predicting the short-term passenger flow state of a rail transit station provided by the present invention use the passenger flow and speed parameter data of the relevant stations to establish a vector error correction model to predict the passenger flow and passenger flow speed of the target station, thereby improving the accuracy of the short-term passenger flow prediction. and reliability.

Description

Method and device for predicting short-term passenger flow state of rail transit stations

technical field

The invention relates to the technical field of urban rail transit management, in particular to a method and device for predicting the short-term passenger flow state of rail transit stations.

Background technique

Due to its advantages of speed, comfort and cleanliness, urban rail transit has attracted more and more residents to choose rail transit, which has caused great changes in the spatial and temporal distribution of passenger flow, and put forward higher requirements for operation management. Prediction of rail transit passenger flow status is one of the key technologies for rail transit operation, management and control. Accurate and reliable prediction can not only reflect the real-time changes in passenger flow, provide data support for capacity analysis and capacity matching, but also provide service level and system operation. It is an important decision-making index for state evaluation, and it is of great significance to conduct in-depth research on it.

At present, some researches have also been carried out on the short-term passenger flow prediction of rail transit, such as neural network model and time series model. However, most of the existing rail transit station passenger flow state predictions are based on the historical passenger flow of the station, and the single-input single-output prediction model is mainly used, ignoring the intrinsic relationship between parameters and the spatiotemporal correlation between stations under networked operation. It is difficult to accurately predict the state of passenger flow without effective information. Therefore, it is possible to further improve the accuracy of passenger flow state prediction by in-depth exploration of the spatiotemporal correlation between the passenger flow parameters of rail transit, and on this basis to construct a multivariable time series model of passenger flow state of rail transit stations.

SUMMARY OF THE INVENTION

The purpose of the present invention is to propose a method and device for predicting the state of short-term passenger flow at rail transit stations, so as to improve the accuracy and reliability of short-term passenger flow prediction.

For this purpose, the present invention adopts the following technical solutions:

A method for predicting a short-term passenger flow state of a rail transit station, the method comprising:

According to the obtained passenger flow and passenger flow speed of the target station of rail transit and the adjacent stations upstream and downstream of the target station, generate the passenger flow and speed time series {S _it , i=1, 2, 3}, where S _it =( q _it , v _it ) ^T , i=1, 2, 3, respectively represent the upstream adjacent site of the target site, the target site and the downstream adjacent site of the target site, q _it is the passenger flow, v _it For passenger flow speed:

The first-order difference operation is performed on the passenger flow and speed time series to obtain a stationary time series s' _it = (q' _it , v' _it ) ^T , where s' _it is the ith station within the time interval t. order difference value;

Use the historical sample data of the stationary time series to construct a vector autoregressive model;

According to the original passenger flow in the same period, the historical sample data of the passenger flow and the speed time series, check the cointegration relationship between the passenger flow and the speed time series {S _it , i=1, 2, 3}, and get the error correction term λecm _t-1 ;

According to the vector autoregressive model and the error correction term, a vector error correction model of the passenger flow and passenger flow speed of the target site is established;

According to the vector error correction model and the video data, the predicted value of the passenger flow and the passenger flow speed of the target site within the time interval t is calculated.

In the above scheme, the vector autoregressive model is:

Among them, p, q is the lag order of the vector autoregressive process, Δq _2t is the first-order difference value of the passenger flow of the target site within the time interval t, and Δv _2t is the first-order difference value of the passenger flow speed of the target site within the time interval t , Δq _1(tm) and Δq _3(tm ) are the first-order difference values of the passenger flow between the upstream and downstream adjacent stations within the time interval (tm), and Δv _1(tm) and Δv _3(tm) are the upstream and downstream adjacent stations The first-order difference value of the _{station passenger flow speed in the time interval ( tm ) ;} , c _x , _cy are parameters to be estimated of the vector autoregressive model; ∈ _xt , ∈ _yt are the error terms of the vector autoregressive model.

In the above scheme, the vector error correction model is:

Among them, the error correction coefficients corresponding to λ ₁ and λ ₂ bits, λ ₁ ecm _t-1 and λ ₂ ecm _t-1 are error correction terms.

In the above scheme, according to the vector error correction model and the video data, the predicted value of the passenger flow and passenger flow speed of the target site in the time interval t is calculated, including:

Obtain actual observations at time intervals (t-1), (t-2), (t-3)...(t-p+1) from the video;

Calculate the predicted values Δq _2t and Δv _2t of the first-order difference time series of passenger flow and passenger flow velocity at the target site within the time interval t according to the actual observed value and the vector error correction model;

According to the predicted values Δq _2t and Δv _2t of the first-order difference time series of the passenger flow and passenger flow speed of the site, the predicted value of the passenger flow and passenger flow speed of the target site within the time interval t is calculated.

In the above solution, before generating the passenger flow and speed time series, the method further includes: acquiring the passenger flow and the passenger flow speed according to video data collected by a video detector in the passage of the rail transit station.

In the above solution, the lag orders p and q in the vector autoregressive model are determined by Bayesian criterion.

In the above solution, the cointegration relationship between the original passenger flow and the passenger flow and speed time series {S _it , i=1, 2, 3} is tested by the Johansen cointegration test method.

In the above solution, the parameters to be estimated: α _xm , β _xm , γ _xm , δ _xm , ε _xm , ∈ _xm , α _ym , β _ym , γ _ym , δ _ym , ε _ym , ∈ _ym , c _x , c _y , and the error correction coefficients λ ₁ and λ ₂ are all estimated by the least squares method.

A device for predicting the state of short-term passenger flow at a rail transit station, the device comprising:

The passenger flow and speed time series generation unit is used to generate passenger flow and speed time series {S _it , i=1, 2 according to the obtained passenger flow and passenger flow speed of the target station of rail transit and the upstream and downstream adjacent stations of the target station , 3}, wherein, S _it =(q _it , v _it ) ^T , i=1, 2, 3, respectively represent the upstream adjacent site of the target site, the target site and the downstream adjacent site of the target site, q _it is the passenger flow, v _it is the passenger flow speed;

The difference operation unit is used to perform a first-order difference operation on the passenger flow and speed time series to obtain a stationary time series S′ _it =(q′ _it , v′ _it ) ^T , where S′ _it is the ith station in The first-order difference value within the time interval t;

a first modeling unit, configured to construct a vector autoregressive model using historical sample data including the stationary time series;

A cointegration relationship testing unit, configured to check the relationship between the passenger flow and the speed time series {S _it , i=1, 2, 3} according to the original passenger flow in the same period, the historical sample data of the passenger flow and the speed time series The cointegration relationship of , the error correction term λecm _t-1 is obtained;

The second modeling unit is configured to establish a vector error correction model of the passenger flow and passenger flow speed of the target site according to the vector autoregressive model and the error correction term;

The predicted value calculation unit is configured to calculate the predicted value of the passenger flow and the passenger flow speed of the target site within the time interval t according to the vector error correction model and the video data.

In the above solution, the device further includes a passenger flow information acquisition unit, configured to acquire the passenger flow and the passenger flow speed according to video data collected by a video detector in the passage of the rail transit station.

The method and device for predicting the short-term passenger flow state of a rail transit station provided by the invention use the passenger flow and speed parameter data of the relevant stations to establish a vector error correction model to predict the passenger flow and passenger flow speed of the target station, thereby improving the accuracy of the short-term passenger flow prediction. sturdiness and reliability.

Description of drawings

Fig. 1 is the realization flow chart of the short-term passenger flow state prediction method of rail transit station according to the embodiment of the present invention;

Fig. 2 is the fitting effect diagram of the short-term predicted value and the actual observed value of the station's morning peak passenger flow in the embodiment of the present invention;

Fig. 3 is a fitting effect diagram of the short-term predicted value and the actual observed value of the passenger flow speed in the morning peak of the station in the embodiment of the present invention;

FIG. 4 is a schematic diagram of the composition and structure of an apparatus for predicting the state of short-term passenger flow at a rail transit station according to an embodiment of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, the drawings only show some but not all structures related to the present invention.

Example 1

As shown in FIG. 1 , the method for predicting the short-term passenger flow state of a rail transit station provided by an embodiment of the present invention includes:

Step 110, according to the obtained passenger flow and passenger flow speed of the target station and the adjacent stations upstream and downstream of the target station, generate a passenger flow and speed time series {S _it , i=1, 2, 3}, where S _it = (q _it , v _it ) ^T , i=1, 2, 3, respectively represent the upstream adjacent site of the target site, the target site and the downstream adjacent site of the target site, q _it is the passenger flow, and v _it is the passenger flow speed.

Step 120: Perform a first-order difference operation on the passenger flow and speed time series to obtain a stationary time series S' _it = (q' _it , v' _it ) ^T , where S' _it is the ith station in the time interval t. first-order difference value.

In step 130, a vector autoregressive model is constructed by using the historical sample data of the stationary time series.

Step 140: Check the cointegration relationship between the passenger flow and the speed time series {S _it , i=1, 2, 3} according to the historical sample data of the original passenger flow, passenger flow and speed time series in the same period, and obtain an error correction term λecm _t-1 .

Step 150 , according to the vector autoregression model and the error correction term, establish a vector error correction model of the passenger flow and passenger flow speed of the target site.

Step 160: Calculate the predicted value of the passenger flow and the passenger flow speed of the target site within the time interval t according to the vector error correction model and the video data.

The technical solution provided by the embodiment of the present invention takes into account the cointegration relationship between the site passenger flow and speed parameters, and adopts the passenger flow and speed parameters as the input of the prediction, which makes up for the neglect of the equilibrium relationship between parameters in the single-variable prediction; The spatial correlation of passenger flow between stations, the passenger flow data of relevant stations is introduced, and a vector error correction model is established, which further improves the accuracy and reliability of short-term passenger flow prediction.

Before step 110, the passenger flow and the passenger flow speed are obtained according to the video data collected by the video detector in the passage of the rail transit station.

After that, in step 110, the passenger flow and speed time series generated according to the passenger flow and the passenger flow speed are continuous time series data with a time interval of 5 minutes.

In step 120, the first-order difference operation formula is:

S' _it =S _i(t+1) -S _it

In the formula: _S'it is the first-order difference value of the i-th station in the time interval t, and S _it is the time series of the i-th station in the time interval t.

In step 130, the vector autoregressive model is:

Among them, p, q is the lag order of the vector autoregressive process, Δq _2t is the first-order difference value of the passenger flow of the target site within the time interval t, and Δv _2t is the first-order difference value of the passenger flow speed of the target site within the time interval t , Δq _1(tm) and Δq _3(tm ) are the first-order difference values of the passenger flow between the upstream and downstream adjacent stations within the time interval (tm), and Δv _1(tm) and Δv _3(tm) are the upstream and downstream adjacent stations The first-order difference value of the _{station passenger flow speed in the time interval ( tm ) ;} , c _x , _cy are the parameters to be estimated in the vector autoregressive model; ∈ _xt , ∈ _yt are the error terms of the vector autoregressive model.

The parameters to be estimated: α _xm , β _xm , γ _xm , δ _xm , ε _xm , ∈ _xm , α _ym , β _ym , γ _ym , δ _ym , ε _ym , ∈ _ym , c _x , c _y adopt the least squares method estimated.

Among them, the lag order p and q in the vector autoregressive model are determined by Bayesian criterion.

In step 140, the cointegration relationship between the original passenger flow and the passenger flow and speed time series {S _it , i=1, 2, 3} is tested by the Johansen cointegration test method.

In step 150, the vector error correction model is:

Among them, the error correction coefficients corresponding to λ ₁ and λ ₂ bits, λ ₁ ecm _t-1 and λ ₂ ecm _t-1 are error correction terms. Among them, the error correction coefficients λ ₁ and λ ₂ are estimated by the least square method.

In step 160, actual observations at time intervals (t-1), (t-2), (t-3)...(t-p+1) are obtained from the video;

Calculate the predicted values Δq _2t and Δv _2t of the first-order difference time series of the passenger flow and passenger flow speed of the target site within the time interval t according to the actual and vector error correction model;

According to the predicted values Δq _2t and Δv _2t of the first-order difference time series of the station's passenger flow and passenger flow speed, the predicted value of the target station's passenger flow and passenger flow speed within the time interval t is calculated.

in,

In an example of the embodiment of the present invention, the first station of an urban rail transit is selected as the target station, and the second station and the third station of its adjacent stations are the upstream station and downstream station of the first station, respectively. According to the morning and evening peak characteristics, the embodiment of the present invention only predicts the passenger flow state of the target station, that is, the first station in the morning and evening peaks.

First, according to the images captured by video surveillance during the morning rush hour during the weekdays from August 1 to August 26, 2016, the directional gradient histogram feature descriptor combined with the support vector machine classifier was used to identify pedestrian targets in the subway surveillance video. Continuously adaptive mean shift algorithm (Continuously daptive Mean-SHIFT, referred to as Camshift) algorithm is used to track the target window to realize the statistics of passenger flow and speed parameters. With a time interval of 5 minutes, the data of the morning peak of the first three weeks of working days is used as the sample data for the calibration of the model parameters, and the data of the next week is used for the evaluation of the prediction performance of the model.

After that, according to the obtained passenger flow and passenger flow speed, the original sequence needs to be obtained, and a first-order difference operation is performed to convert it into a stationary time series.

After that, a vector autoregressive model is established for the stationary time series, and the expression is as follows:

Through the Bayesian information criterion, the lag order p, q of the vector autoregressive model of the target site can be determined, and the results are given in Table 1.

Table 1 The lag order of the vector autoregressive model of Gulou station

On the basis of determining the lag order, the Johansen cointegration test method based on regression coefficient is used to test the cointegration relationship between the original series of parameters of the target station and adjacent stations. This embodiment provides the cointegration relationship test results of the upstream and downstream stations of the first train station and its adjacent stations, as shown in Table 2.

Table 2 Cointegration test results of target site and adjacent sites

The principle of the Johansen cointegration test is to use the maximum likelihood estimation method to judge the cointegration relationship between variables. The characteristic root statistic and the critical value are both statistical values at the 95% confidence level. When the characteristic root statistic is greater than the critical value value, reject the null hypothesis, otherwise accept it. It can be seen from the results that there is a co-integration relationship between the target site and the upstream and downstream site parameter sequences, and a Vector Error Correction (VEC) model can be established.

After that, a vector error correction model is constructed, combined with the vector autoregression model and the co-integration relationship, a vector error correction model with a lag order of 3 for the passenger flow variable and a lag order of 2 for the speed variable is established as follows:

All parameters to be estimated are estimated by the ordinary least squares method with sample data. The results are shown in Table 3. Since the estimation results of the systems c _x and _cy are not statistically significant, in the VEC model of the first vehicle station, c _x , _cy are all 0.

Table 3 Estimation results of VEC model parameters

Finally, calculate the first-order difference time series predicted values Δq _2t and Δv _2t of the passenger flow, passenger flow and speed time series of the first station, and further calculate the predicted value of the horizontal series in the time interval t q _2t =q _{2(t- 1)} +Δq _2t , v _2t =v _2(t-1) +Δv _2t , the fitting effect between the predicted value of passenger flow and passenger flow speed and the actual observed value is shown in Figure 2 and Figure 3. Among them, as shown in FIG. 2 , the curve 201 is the passenger flow observed value, and the curve 202 is the passenger flow predicted value. As shown in FIG. 3 , the curve 301 is the observed value of the passenger flow speed, and the curve 302 is the predicted value of the passenger flow speed.

This embodiment uses root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) to evaluate the prediction performance of the target site's passenger flow and speed. In order to compare with traditional forecasting methods, the univariate time series model ARIMA(0, 1, 1) is constructed using the same sample data. This model does not consider the spatial-temporal correlation and the equilibrium relationship between parameters. For example, Table 4.

Table 4. Comparison of prediction performance between VEC and ARIMA models

From the results given in Table 4, it can be seen that the VEC model is significantly better than the ARIMA model in the prediction of passenger flow and speed, and the VEC model constructed by the present invention has better prediction performance.

The method for predicting the short-term passenger flow state of a rail transit station provided by the present invention uses the passenger flow and speed parameter data of the relevant stations to establish a vector error correction model to predict the passenger flow and passenger flow speed of the target station, thereby improving the accuracy of the short-term passenger flow prediction and improving the speed of the passenger flow. reliability.

Embodiment 2

An embodiment of the present invention provides a short-term passenger flow state prediction device at a rail transit station, the device comprising:

The passenger flow and speed time series generating unit 410 is configured to generate the passenger flow and speed time series {S _it , i=1, 2, 3}, wherein, S _it =(q _it , v _it ^)T , i=1, 2, 3, respectively represent the upstream adjacent site of the target site, the target site and the downstream adjacent site of the target site, q _it is the passenger flow , v _it is the passenger flow speed.

The difference operation unit 420 is used to perform a first-order difference operation on the passenger flow and speed time series to obtain a stationary time series S' _it =(q' _it , v' _it ) ^T , where S' _it is the time of the i-th station First-order difference value within interval t.

The first modeling unit 430 is configured to construct a vector autoregressive model by using historical sample data including stationary time series.

The cointegration relationship testing unit 440 is configured to test the cointegration between the passenger flow and the speed time series {S _it , i=1, 2, 3} according to the historical sample data of the original passenger flow, passenger flow and speed time series in the same period relationship, the error correction term λecm _t-1 is obtained.

The second modeling unit 450 is configured to establish a vector error correction model of the passenger flow and passenger flow speed of the target site according to the vector autoregressive model and the error correction term.

The predicted value calculation unit 460 is configured to calculate the predicted value of the passenger flow and the passenger flow speed of the target site within the time interval t according to the vector error correction model and the video data.

The technical solution provided by the embodiment of the present invention takes into account the cointegration relationship between the site passenger flow and speed parameters, and adopts the passenger flow and speed parameters as the input of the prediction, which makes up for the neglect of the equilibrium relationship between parameters in the single-variable prediction; The spatial correlation of passenger flow between stations, the passenger flow data of relevant stations is introduced, and a vector error correction model is established, which further improves the accuracy and reliability of short-term passenger flow prediction.

The short-term passenger flow state prediction device of a rail transit station provided by the present invention further includes a passenger flow information acquisition unit for acquiring passenger flow and passenger flow speed according to video data collected by a video detector in the passage of the rail transit station.

The device for predicting the short-term passenger flow state of rail transit stations provided by the present invention uses the passenger flow and speed parameter data of the relevant stations to establish a vector error correction model to predict the passenger flow and passenger flow speed of the target station, thereby improving the accuracy of short-term passenger flow prediction and improving the speed of passenger flow. reliability.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims (8)

1. a short-term passenger flow state prediction method for rail transit stations, characterized in that the method comprises: The passenger flow and speed time series {S _it , i=1, 2, 3} are generated according to the obtained passenger flow and passenger flow speed of the target station of rail transit and the adjacent stations upstream and downstream of the target station, wherein the passenger flow and the passenger flow speed are obtained from the video data in the channel; S _it =(q _it , v _it ) ^T , i=1, 2, 3, respectively representing the upstream adjacent site of the target site, the target site and the target The downstream adjacent station of the station, q _it is the passenger flow, and v _it is the passenger flow speed; The first-order difference operation is performed on the passenger flow and speed time series to obtain a stationary time series S' _it = (q' _it , v' _it ) ^T , where S' _it is the ith station in the time interval t. order difference value; Using the historical sample data of the stationary time series to construct a vector autoregressive model, the vector autoregression model is:

Among them, p, q are the lag order of the vector autoregressive process, △q _2t is the first-order difference value of the passenger flow at the target site within the time interval t, and △v _2t is the first-order passenger flow velocity of the target site within the time interval t. The difference value, △q _1(tm) and △q _3(tm ) are the first-order difference values of the passenger flow between the upstream and downstream adjacent stations within the time interval (tm), △v _1(tm) , △v _3(tm) is the first-order difference value of the passenger flow velocity between the _upstream and _{downstream adjacent stations within} the _{time interval} ( _{tm ) ;} , ε _ym , ∈ _ym , c _x , and _cy are parameters to be estimated of the vector autoregressive model; ∈ _xt , ∈ _yt are the error terms of the vector autoregressive model; According to the original passenger flow in the same period, the historical sample data of the passenger flow and the speed time series, check the cointegration relationship between the passenger flow and the speed time series {S _it , i=1, 2, 3}, and get the error correction term λecm _t-1 ; According to the vector autoregression model and the error correction term, a vector error correction model for the passenger flow and passenger flow speed of the target site is established, and the vector error correction model is:

Among them, the error correction coefficients corresponding to λ ₁ and λ ₂ bits, λ ₁ ecm _t-1 and λ ₂ ecm _t-1 are error correction terms; According to the vector error correction model and the video data, the predicted value of the passenger flow and the passenger flow speed of the target site within the time interval t is calculated. 2. method according to claim 1, is characterized in that, described according to described vector error correction model and video data calculating the predicted value of target site passenger flow and passenger flow speed in time interval t, comprises: Obtain actual observations at time intervals (t-1), (t-2), (t-3)...(t-p+1) from the video; Calculate the predicted values Δq _2t and Δv _2t of the first-order difference time series of passenger flow and passenger flow velocity at the target site within the time interval t according to the actual observed value and the vector error correction model; According to the predicted values Δq _2t and Δv _2t of the first-order difference time series of the passenger flow and passenger flow speed of the site, the predicted value of the passenger flow and passenger flow speed of the target site within the time interval t is calculated. 3 . The method according to claim 1 , wherein, before generating the passenger flow and speed time series, the method further comprises: obtaining, according to video data collected by a video detector in the passage of the rail transit station, obtaining 3 . the passenger flow and the passenger flow speed. 4 . The method according to claim 1 , wherein the lag orders p and q in the vector autoregressive model are determined by Bayesian criterion. 5 . 5 . The method according to claim 1 , wherein the cointegration relationship between the original passenger flow and the passenger flow and speed time series {S _it , i=1, 2, 3} is performed by Johansen cointegration. 6 . test method. 6. The method according to any one of claims 1 to 5, wherein the parameters to be estimated: α _xm , β _xm , γ _xm , δ _xm , ε _xm , ∈ _xm , α _ym , β _ym , γ _ym , δ _ym , ε _ym , ∈ _ym , c _x , _cy , and the error correction coefficients λ ₁ , λ ₂ are all estimated by using the least squares method. 7. A device for predicting short-term passenger flow state of a rail transit station, wherein the device comprises: The passenger flow and speed time series generating unit is used for generating passenger flow and speed time series {S _it , i=1,2 ,3}, wherein, the passenger flow and the passenger flow speed are obtained from the video data in the channel; S _it =(q _it , v _it ) ^T , i=1, 2, 3, respectively represent the upstream adjacent site of the target site, The target site and the downstream adjacent sites of the target site, q _it is the passenger flow, and v _it is the passenger flow speed; The difference operation unit is used to perform first-order difference operation on the passenger flow and speed time series to obtain a stationary time series S' _it =(q' _it , v' _it ) ^T , where S' _it is the ith station in The first-order difference value within the time interval t; The first modeling unit is used to construct a vector autoregressive model using the historical sample data of the stationary time series, and the vector autoregression model is:

Among them, p, q are the lag order of the vector autoregressive process, △q _2t is the first-order difference value of the passenger flow at the target site within the time interval t, and △v _2t is the first-order passenger flow velocity of the target site within the time interval t. The difference value, △q _1(tm) and △q _3(tm ) are the first-order difference values of the passenger flow between the upstream and downstream adjacent stations within the time interval (tm), △v _1(tm) , △v _3(tm) is the first-order difference value of the passenger flow velocity between the _upstream and _{downstream adjacent stations within} the _{time interval} ( _{tm ) ;} , ε _ym , ∈ _ym , c _x , and _cy are parameters to be estimated of the vector autoregressive model; ∈ _xt , ∈ _yt are the error terms of the vector autoregressive model; A cointegration relationship testing unit, configured to check the relationship between the passenger flow and the speed time series {S _it , i=1, 2, 3} according to the original passenger flow in the same period, the historical sample data of the passenger flow and the speed time series The cointegration relationship of , the error correction term λecm _t-1 is obtained; The second modeling unit is used to establish a vector error correction model of the passenger flow and passenger flow speed of the target site according to the vector autoregression model and the error correction term, and the vector error correction model is:

Among them, the error correction coefficients corresponding to λ ₁ and λ ₂ bits, λ ₁ ecm _t-1 and λ ₂ ecm _t-1 are error correction terms; The predicted value calculation unit is configured to calculate the predicted value of the passenger flow and the passenger flow speed of the target site within the time interval t according to the vector error correction model and the video data. 8 . The device according to claim 7 , wherein the device further comprises a passenger flow information acquisition unit, configured to acquire the passenger flow according to video data collected by a video detector in the passage of the rail transit station. 9 . and the passenger flow rate.

Images