Subway Train Delay Adjustment Method Considering Passenger Flow Influence and Recycling of Regenerative Braking Energy

Abstract

The present invention discloses a method for adjusting subway train delays considering passenger flow influence and utilization of regenerative braking energy. The steps are as follows: Considering the restrictive effect of passenger flow factors on the upper limit of stop time adjustment after a train delay occurs, a passenger flow prediction model based on LSTM is established; Considering the absorption and utilization of regenerative braking energy during the adjustment process of the delayed train, a calculation model for DC traction network energy consumption is established; A stop time model is established by simulating the process of passengers getting on and off the train; Then, combining the above sub-models, a subway train delay adjustment model is established, with the lowest change in substation energy consumption during the adjustment process as the optimization goal, and a solution process for this model is designed using the particle swarm optimization algorithm CDPSO based on center-discrete learning; Finally, simulation verification is carried out using actual subway line data. The method of the present invention is scientific, reliable, and efficient, providing data and theoretical support for subway train delay adjustment, and having high application value and application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal control in the subway operation system, and particularly relates to a method for adjusting subway train delays considering passenger flow influence and the utilization of regenerative braking energy. Background Art

[0002] With the rapid development of the subway operation system, its scale is constantly expanding, and the density of the subway network is also constantly expanding. As a result, the headway between trains is constantly shrinking, and the running density of trains increases accordingly, which invisibly brings great pressure to the subway operation department. A small mistake may lead to train delays. If the train delays are not adjusted in time, it will have a great impact on the subway operation system. Therefore, measures should be taken to adjust the operation of the delayed trains to make them resume on-time operation as soon as possible.

[0003] When the degree of train delay is relatively light, the arrival and departure times of the delayed train at subsequent stations can be adjusted through the subway train ATS system to resume normal operation. When a train is delayed, first, the normal travel needs of platform passengers should be ensured, and then, on this basis, consider how to adjust to minimize the increase in energy consumption during the adjustment process. Summary of the Invention

[0004] The purpose of the present invention is to provide a scientific, reliable, and efficient method for adjusting subway train delays considering passenger flow influence and the utilization of regenerative braking energy.

[0005] The technical solution for achieving the purpose of the present invention is: a method for adjusting subway train delays considering passenger flow influence and the utilization of regenerative braking energy, including the following steps:

[0006] Step 1, establish a passenger flow prediction model based on LSTM: First, preprocess the passenger flow data, then train the constructed LSTM network layer model, and finally use the trained LSTM network layer model to predict the passenger flow data;

[0007] Step 2, respectively model the subway train, traction substation, and traction network, and finally construct a time-varying DC traction network energy consumption calculation model for calculating the substation energy consumption within a period of time;

[0008] Step 3, after obtaining the predicted passenger flow data in Step 1, establish a stop time model by simulating the process of passengers getting on and off the train;

[0009] Step 4, establish a train delay adjustment model, including the adjustment amounts of the delayed train number, the name of the delayed platform, the delay duration, the running time of the subsequent section, and the platform stop time;

[0010] Step 5: Take the lowest change in substation energy consumption during the adjustment process as the optimization goal, use the particle swarm optimization algorithm based on center-discrete learning to solve each modulation quantity in the train delay adjustment model, and finally use the actual subway line data for simulation verification.

[0011] Compared with the prior art, the present invention has the following remarkable advantages: (1) Considering the constraint of dynamic passenger flow on the upper limit of stop time adjustment and the absorption and utilization of regenerative braking energy during the train delay adjustment process, a time-varying DC traction network energy consumption calculation model is constructed to calculate the substation energy consumption within a certain period of time; (2) Use the particle swarm optimization algorithm based on center-discrete learning to design the solution process for the delay train adjustment model, solve the problem that the traditional particle swarm optimization algorithm is prone to falling into local optimum, use a linearly decreasing inertia weight, provide data and theoretical support for subway train delay adjustment, and have high application value and application prospects. Brief Description of the Drawings

[0012] Figure 1 It is the schematic diagram of the subway train delay adjustment method considering the influence of passenger flow and the utilization of regenerative braking energy of the present invention. Detailed Embodiment

[0013] Combined with Figure 1 , a subway train delay adjustment method considering the influence of passenger flow and the utilization of regenerative braking energy of the present invention includes the following steps:

[0014] Step 1: Establish an LSTM-based passenger flow prediction model: First, preprocess the passenger flow data, then train the constructed LSTM network layer model, and finally use the trained LSTM network layer model to predict the passenger flow data;

[0015] Step 2: Model the subway train, traction substation, and traction network respectively, and finally construct a time-varying DC traction network energy consumption calculation model to calculate the substation energy consumption within a certain period of time;

[0016] Step 3: After obtaining the predicted passenger flow data in Step 1, establish a stop time model by simulating the process of passengers getting on and off the train;

[0017] Step 4: Establish a train delay adjustment model, including the adjustment quantities of the delayed train number, the name of the delayed platform, the delay duration, the running time of the subsequent section, and the platform stop time;

[0018] Step 5: Take the lowest change in substation energy consumption during the adjustment process as the optimization goal, use the particle swarm optimization algorithm based on center-discrete learning to solve each modulation quantity in the train delay adjustment model, and finally use the actual subway line data for simulation verification.

[0019] Further, the step LSTM passenger flow prediction model described in step 1 includes: preprocessing the passenger flow data, then training the constructed LSTM network layer model, then outputting the model, and finally predicting the passenger flow data.

[0020] Further, the time-varying DC traction network energy consumption calculation model described in step 2 is specifically as follows:

[0021] (2.1) By modeling trains, traction substations, traction networks, etc., a time-varying electrical network model is finally constructed;

[0022] (2.2) Take the measured data of a time period as input and import it into the DC traction network energy consumption calculation model for calculation to obtain the energy consumption value result. Compare it with the actual energy consumption value. If the deviation is greater than the set threshold, adjust the relevant parameters until the deviation is within the set threshold range.

[0023] Further, after obtaining the predicted passenger flow data in step 1 in step 3, a stop time model is established by simulating the process of passengers getting on and off the train, specifically as follows:

[0024] (3.1) Time model for the getting-off period

[0025] The time required for passengers to get off is approximately linearly related to the number of passengers getting off. The mathematical model is as follows:

[0026] T a (y a ) = A1 * y a + A2

[0027] In the formula, y a is the number of passengers among all passengers getting off who get off during the getting-off period; T a (y a ) is the duration consumed during the getting-off period; A1 and A2 are the model parameters for the getting-off period.

[0028] (3.2) Time model for the mixed getting-on and getting-off period

[0029] At this time, the passengers getting on the train are mixed with the passengers getting off the train, and the duration is 4 seconds.

[0030] (3) Time model for the getting-on period

[0031] When all the passengers on the train have got off, the mixed getting-on and getting-off period ends, and the passengers on the platform line up in two teams to get on the train; at this time, since there are still empty seats on the train, the speed of passengers getting on the train is relatively fast. After all the seats are occupied, the speed will slow down, and the last passengers getting on the train will speed up because of the train's door closing reminder. Therefore, the getting-on period is divided into three time segments;

[0032] According to the above analysis, the following piecewise model is obtained:

[0033]

[0034] In the formula, f(x) is the boarding time required when the number of boarding passengers reaches x during the boarding period; N is the number of passengers who hope to get a seat in the front at each door; P is the average number of people at the door when the passenger density in the carriage reaches the critical density; C1, C2, C3, C4, D1, D2, E1, E2, and E3 are all time model parameters during the boarding period.

[0035] Furthermore, the establishment of the train delay adjustment model described in step 4 includes the adjustment amounts of the delayed train number, the name of the delayed platform, the delay duration, the running time of the subsequent section, and the platform stop time, which are specifically as follows:

[0036] (4.1) Judge the actual arrival time and the planned arrival time of the train to determine whether they are the same, that is, whether the train is delayed. If it is delayed, calculate the difference to obtain the delay duration t delay , and record the number of the delayed train, the name of the delayed platform, and the delay duration;

[0037] (4.2) Judge whether the delay duration t delay is greater than the train tracking redundancy time T trackmin . If t delay ≤T trackmin , the delay of the current train will not affect the subsequent running trains. On the contrary, if t delay >T trackmin , the operation of the following train will be affected, resulting in a chain delay. Obtain the delay information of the following train according to the safe tracking interval between trains, and record the delayed train number, the delayed station, and the delay time;

[0038] (4.3) According to the delay train adjustment plan obtained from the above judgment, obtain the sequence of time adjustment amounts {Δt1, Δt2,..., Δt n}.

[0039] Furthermore, taking the minimum change in substation energy consumption during the adjustment process as the optimization goal, using the particle swarm algorithm based on center-dispersion learning to solve each modulation amount in the train delay adjustment model, and finally using the actual subway line data for simulation verification, which is specifically as follows:

[0040] (5.1) The particle coding method uses real number coding, and the coding object is the adjustment plan after the train is delayed;

[0041] (5.2) Determine the dimension of the particle according to the train delay duration, the upper limit of the adjustment amount of the train running time in the subsequent section, and the upper limit of the adjustment amount of the train stopping time at the subsequent platform. The adjustment values of the train running section and platform in the subsequent section are stored in the particle;

[0042] (5.3) Determine the number of particle swarms, the number of iterations, the inertia weight ω, and the elite learning control factor c;

[0043] (5.4) Set the fitness function: Take the lowest change in substation energy consumption during the adjustment process as the fitness value function, which is obtained from the time-varying power flow calculation of the power grid. The specific form is as follows:

[0044]

[0045] In the above formula, E sub_total is the total energy consumption of the substation in the line; is the energy consumption of the jth substation, j = 1, 2,..., N sub , N sub is the number of substations in the line; is the output power of the jth substation at time t; Δt is the simulation step size; is the terminal voltage of the jth substation at time t; is the branch current of the jth substation at time t; is the node voltage of the jth substation at the catenary end at time t; is the node voltage of the jth substation at the running rail end at time t; is the node injection current of the jth substation at the catenary end at time t; is the node injection current of the jth substation at the running rail end at time t;

[0046] (5.5) Initialize relevant parameters;

[0047] (5.6) Initialize the position of the particle and calculate the fitness value of the particle;

[0048] (5.7) Linearly update the inertia weight ω;

[0049] (5.8) Determine the global optimal particle, and alternately use the following formulas to update the particle velocity and position:

[0050]

[0051]

[0052] In the above formula, ω is the inertia weight of the particle, c is the elite learning control factor, and r is a decimal number that satisfies the random uniform distribution between [0, 1]; is the arithmetic mean of the local historical optimal positions of the top L individuals in the population, γ i (j) is the number of the individual that particle i needs to learn in the j-th dimension.

[0053] (5.9) Calculate the fitness value of the particle, and determine whether the particle converges or meets the number of times. If it meets, go to step (10); if not, return to step (5.6);

[0054] (5.10) Output the optimal adjustment method after train delay: the adjustment amount of the running time in the subsequent section and the adjustment amount of the platform stop time.

[0055] Furthermore, the coding object in step (5.1) is the adjustment plan after train delay. Specifically: the adjustment strategy of the train uses real number coding, and the adjustment amount of the running time in the subsequent section and the adjustment amount of the platform stop duration after the delayed train are used as real number coding values. Each dimension of the particle represents the time adjustment amount of the subsequent section or platform.

[0056] Furthermore, in step (5.2), the dimension of the particle is determined according to the train delay duration, the upper limit of the adjustment amount of the running time in the subsequent section of the train, and the upper limit of the adjustment amount of the platform stop time in the subsequent section of the train, as follows:

[0057]

[0058] where h is the dimension of the particle, t delay is the delay duration, T Reds 、T Redr are the running redundancy time and the stop redundancy time in the subsequent section of the train respectively.

[0059] Furthermore, in step (5.3), the number of the particle swarm, the number of iterations, the inertia weight ω, and the elite learning control factor c are determined. Specifically: the number of the particle swarm is set to 25, the number of iterations is set to 50, the inertia weight ω is set to 0.9, and the elite learning control factor c is set to 2.

[0060] Furthermore, the calculation steps of the particle fitness value in step (5.9) include:

[0061] (a) Take the i-th particle in the population and decode the train adjustment strategy corresponding to this individual, and the initial value of i is 0;

[0062] (b) Import the running adjustment method decoded from the particle into the train delay adjustment model;

[0063] (c) Conduct train delay adjustment simulation and calculate the fitness value of the individual according to the running method;

[0064] (d) Save the individual fitness value of the particle;

[0065] (e) Determine whether the current particle is the last particle in the group: If so, the calculation ends; otherwise, i = i + 1, and jump to (a).

[0066] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0067] Embodiment

[0068] Combined with Figure 1 , the subway train delay adjustment method of the present invention considering passenger flow influence and utilization of regenerative braking energy includes the following steps:

[0069] Step 1, establish a basic data module:

[0070] Including line data, platform data, substation data, planned operation schedule, interval operation power data, predicted platform passenger flow data, train departure interval, etc.

[0071] Step 2, establish a stop time model, specifically as follows:

[0072] (1) Time model for the getting-off period

[0073] The time required for passengers to get off is approximately linearly related to the number of passengers getting off. The mathematical model is as follows:

[0074] T a (y a ) = A1 * y a + A2

[0075] In the formula, y a is the number of passengers getting off during the getting-off period among all passengers getting off; T a (y a ) is the duration consumed during the getting-off period; A1 and A2 are model parameters for the getting-off period.

[0076] (2) Time model for the mixed getting-on and getting-off period

[0077] At this time, passengers getting on and getting off are mixed together, and the duration is about 4s.

[0078] (3) Time model for the getting-on period

[0079] When all passengers on the train have got off, the mixed getting-on and getting-off period ends, and passengers on the platform line up in two teams and quickly get on the train. At this time, since there are still empty seats on the train, the speed of passengers getting on is relatively fast. After all the seats are occupied, the speed will slow down, and the last passengers getting on will speed up due to the train's door closing reminder. Therefore, this period can be divided into three sub-periods.

[0080] Based on the above analysis, a piecewise model as shown in the following formula can be obtained:

[0081]

[0082] In the above formula, f(x) is the boarding time required when the number of boarding passengers reaches x during the boarding period; N is the number of passengers who hope to get a seat in the front row at each door; P is the average number of people at the door when the passenger density in the carriage reaches a certain critical density; C1, C2, C3, C4, D1, D2, E1, E2, and E3 are all parameters of the boarding period model.

[0083] Step 3: Establish a time-varying DC traction network energy consumption calculation model, which is specifically as follows:

[0084] (1) By modeling the train, traction substation, traction network, etc., a time-varying electrical network model is finally constructed;

[0085] (2) Import the measured data of a certain time period as input into the program for calculation to obtain the energy consumption value result, and compare it with the actual energy consumption value. If the deviation is large, adjust the relevant parameters to reduce the deviation of the calculation result.

[0086] Step 5: Establish a delayed train adjustment model, which is specifically as follows:

[0087] (1) Judge the actual arrival time and the planned arrival time of the train to determine whether they are the same, that is, whether the train is delayed. If it is delayed, calculate the delay duration t delay , and record the number of the delayed train, the name of the delayed platform, and the delay duration.

[0088] (2) Judge whether the delay duration t delay is greater than the train tracking redundancy time T trackmin . If t delay <T trackmin , the delay of the current train will not affect the subsequent running trains. Conversely, if t delay >T trackmin , the operation of the following train will be affected, resulting in a chain delay. According to the safe tracking interval between trains, the delay information of the following train can be obtained, and the number of the delayed train, the delayed station, and the delay time are recorded;

[0089] (3) Obtain a sequence of time adjustment amounts {Δt1, Δt2,..., Δt n} according to the delayed train adjustment plan obtained from the above judgment.

[0090] Step 6: Take the lowest change in substation energy consumption during the adjustment process as the optimization goal, use the particle swarm optimization algorithm based on center-dispersion learning to design the solution process for the delayed train adjustment model, and finally use the actual subway line data for simulation verification, as follows:

[0091] (1) Use real number coding for the particle coding method, and the coding object is the adjustment plan after train delay;

[0092] (2) Determine the dimension of the particle according to the train delay duration, the upper limit of the adjustment amount of the running time of the train in the subsequent sections, and the upper limit of the adjustment amount of the stopping time of the train at the subsequent platforms. The adjustment values of the train in the subsequent running sections and platforms are stored in the particle;

[0093] (3) Determine the particle swarm population size, the number of iterations, the inertia weight ω, and the elite learning control factor c;

[0094] (4) Set the fitness function: Take the lowest change in substation energy consumption during the adjustment process as the fitness value function, which is obtained from the time-varying power network flow calculation. The specific form is as follows:

[0095]

[0096] In the above formula, E sub_total is the total energy consumption of the substations in the line; is the energy consumption of the j-th substation, j = 1, 2,..., N sub and N sub is the number of substations in the line; is the output power of the j-th substation at time t; Δt is the simulation time step; is the terminal voltage of the j-th substation at time t; is the branch current of the j-th substation at time t; is the node voltage of the j-th substation at the catenary end at time t; is the node voltage of the j-th substation at the running rail end at time t; is the node injection current of the j-th substation at the catenary end at time t; is the node injection current of the j-th substation at the running rail end at time t.

[0097] (5) Initialize the relevant parameters;

[0098] (6) Initialize the position of the particle and calculate the fitness value of the particle;

[0099] (7) Update the inertia weight ω linearly;

[0100] (8) Determine the global optimal particle, and alternately use the following formulas to update the particle velocity and position:

[0101]

[0102]

[0103] In the above formula, ω is the inertial weight of the particle. c is the "elite learning control factor", and r is a decimal number that satisfies a random uniform distribution between [0, 1]. is the arithmetic mean of the local historical optimal positions of the top L individuals in the population, and γ i (j) is the number of the individual that particle i needs to learn in the j-th dimension.

[0104] (9) Calculate the fitness value of the particle, and determine whether the particle converges or whether the number of times is satisfied. If it is satisfied, go to step (10); if not, return to step (6);

[0105] (10) Output the optimal adjustment method after train delay: the adjustment amount of the running time in the subsequent section and the adjustment amount of the platform stop time.

[0106] The present invention uses a particle swarm optimization algorithm based on center-dispersed learning to design a solution process for the delay train adjustment model, solves the problem that the traditional particle swarm optimization algorithm is prone to falling into local optimum, uses a linearly decreasing inertial weight, provides data and theoretical support for subway train delay adjustment, and has high application value and application prospects.

Claims

1. A subway train delay adjustment method considering passenger flow influence and regenerative braking energy utilization, characterized in that, It includes the following steps: Step 1, establish a passenger flow prediction model based on LSTM: First, preprocess the passenger flow data, then train the constructed LSTM network layer model, and finally use the trained LSTM network layer model to predict the passenger flow data; Step 2, model the subway train, traction substation, and traction network respectively, and finally build a time-varying DC traction network energy consumption calculation model for calculating the substation energy consumption within a period of time; Step 3, after obtaining the predicted passenger flow data in Step 1, establish a stop time model by simulating the process of passengers getting on and off the train; Step 4, establish a train delay adjustment model, including the adjustment amounts of the delayed train number, the name of the delayed platform, the delay duration, the running time of the subsequent section, and the platform stop time; Step 5, take the lowest change amount of substation energy consumption during the adjustment process as the optimization goal, use the particle swarm algorithm based on center-discrete learning to solve each modulation amount in the train delay adjustment model, and finally use the actual subway line data for simulation verification, specifically as follows: (5.1) The particle coding method uses real number coding, and the coding object is the adjustment plan after train delay; (5.2) Determine the dimension of the particle according to the train delay duration, the upper limit of the adjustment amount of the running time of the subsequent section passed by the train, and the upper limit of the adjustment amount of the platform stop time passed by the train. What is stored in the particle is the adjustment value of the subsequent running section and platform of the train; (5.3) Determine the particle swarm population number, the number of iterations, the inertia weight ω, and the elite learning control factor c; (5.4) Set the fitness function: Take the lowest change amount of substation energy consumption during the adjustment process as the fitness value function, which is obtained by the time-varying electric network power flow calculation, and the specific form is as follows: In the above formula, E sub_total is the total energy consumption of the substations in the line; is the energy consumption of the j-th substation, where j = 1, 2,..., N sub , N sub is the number of substations in the line; is the output power of the j-th substation at time t; Δt is the simulation step size; is the terminal voltage of the j-th substation at time t; is the branch current of the j-th substation at time t; is the node voltage of the j-th substation at the catenary end at time t; is the node voltage of the j-th substation at the running rail end at time t; is the node injection current of the j-th substation at the catenary end at time t; is the node injection current of the j-th substation at the running rail end at time t; (5.5) Initialize the relevant parameters; (5.6) Initialize the position of the particle and calculate the fitness value of the particle; (5.7) Update the inertia weight ω linearly; (5.8) Determine the global optimal particle, and alternately use the following formulas to update the particle velocity and position: In the above formula, ω is the inertia weight of the particle, c is the elite learning control factor, and r is a decimal number that satisfies a random uniform distribution between [0, 1]; is the arithmetic mean of the local historical optimal positions of the top L individuals in the population, and γ i (j) is the number corresponding to the individual that particle i needs to learn in the j-th dimension; V i j (t) is the velocity of particle i in the j-th dimension at time t; V i j (t + 1) is the velocity of particle i in the j-th dimension at time t + 1; is the position of particle i in the j-th dimension at time t; is the position of particle i in the j-th dimension at time t + 1; is the current global optimal position of the group, and γ i (j) is the number corresponding to the individual that particle i needs to learn in the j-th dimension, and γ i (j) ∈ [1, N], and it can be known that N is the total number of individuals that particle i needs to learn in the j-th dimension; (5.9) Calculate the particle fitness value, and judge whether the particle converges or whether the number of times is satisfied. If it is satisfied, enter Step (10). If it is not satisfied, return to Step (5.6); (5.10) Output the optimal adjustment method after train delay: the adjustment amount of the running time of the subsequent section and the adjustment amount of the platform stop time.

2. The subway train delay adjustment method considering passenger flow influence and utilization of regenerative braking energy according to claim 1, characterized in that The time-varying DC traction network energy consumption calculation model described in Step 2 is specifically as follows: Import the measured data of a period of time as input into the DC traction network energy consumption calculation model for calculation to obtain the energy consumption value result, compare it with the actual energy consumption value. If the deviation is greater than the set threshold, adjust the relevant parameters until the deviation is within the set threshold range.

3. The subway train delay adjustment method considering passenger flow influence and utilization of regenerative braking energy according to claim 1, characterized in that, After obtaining the predicted passenger flow data in Step 1 in Step 3, establish a stop time model by simulating the process of passengers getting on and off the train, specifically as follows: (3.1) Time model for the getting-off period The time required for passengers to get off is approximately linearly related to the number of passengers getting off. The mathematical model is as follows: T a (y a ) = A1 * y a + A2 where y a is the number of passengers getting off during the alighting period among all passengers getting off; T a (y a ) is the time consumed during the alighting period; A1 and A2 are model parameters for the alighting period; (3.2) Time model for the mixed getting-on and getting-off period At this time, the passengers getting on the train are mixed with the passengers getting off the train, and the duration is 4 seconds; (3.3) Time model for the getting-on period When all the passengers on the train have got off, the up-and-down mixed operation period ends, and the passengers on the platform line up in two teams to board the train. At this time, since there are still empty seats on the train, the boarding speed of passengers is relatively fast. After all the seats are occupied, the speed will slow down. The last passengers to board the train will speed up due to the door closing reminder of the train. Therefore, the boarding period is divided into three time segments. Based on the above analysis, the following piecewise model is obtained: In the formula, f(x) is the boarding time required when the number of boarding passengers reaches x during the boarding period; N is the number of passengers who hope to get a seat in front of each door; P is the average number of people at the door when the passenger density in the carriage reaches the critical density; C1, C2, C3, C4, D1, D2, E1, E2, and E3 are all time model parameters of the boarding period.

4. The subway train delay adjustment method considering passenger flow influence and regenerative braking energy utilization according to claim 1, characterized in that, The establishment of the train delay adjustment model described in step 4 includes the train number of the delayed train, the name of the delayed platform, the duration of the delay, the adjustment amount of the running time of the subsequent section, and the adjustment amount of the platform stop time, which are specifically as follows: (4.1) Determine the actual arrival time of the train and the planned arrival time to check if they are the same, i.e., whether the train is delayed. If it is delayed, calculate the delay duration t delay , and record the number of the delayed train, the name of the delayed platform, and the delay duration; (4.2) Determine the delay duration t delay Is it greater than the train tracking redundancy time T trackmin , if t delay ≤T trackmin , the delay of the current train will not affect the subsequent trains in operation. On the contrary, if t delay >T trackmin , the operation of the following train will be affected, resulting in a chain delay. Obtain the delay information of the following train based on the safe tracking interval between trains, and record the delayed train number, delayed station and delay time; (4.3) The train delay adjustment plan obtained based on the above judgment yields a sequence of time adjustment amounts {Δt1, Δt2,..., Δt n}.

5. The subway train delay adjustment method considering passenger flow influence and utilization of regenerative braking energy according to claim 1, characterized in that The coding object in step (5.1) is the adjustment plan after the train is delayed, specifically: The adjustment strategy of the train uses real number coding. The adjustment amount of the running time of the section passed by the delayed train and the adjustment amount of the platform stop duration are used as real number coding values. Each dimension of the particle represents the time adjustment amount of the subsequent section or platform.

6. The subway train delay adjustment method considering passenger flow influence and regenerative braking energy utilization according to claim 1, characterized in that In step (5.2), the dimensions of the particles are determined according to the train delay duration, the upper limit of the adjustment amount of the running time of the section passed by the train subsequently, and the upper limit of the adjustment amount of the platform stop time of the train subsequently, which are specifically as follows: where h is the dimension of the particle and t delay is the delay duration, and T Reds , T Redr are the running redundancy time and the stopping redundancy time of the subsequent sections of the train, respectively.

7. The subway train delay adjustment method considering passenger flow influence and utilization of regenerative braking energy according to claim 1, characterized in that In step (5.3), the number of the particle swarm population, the number of iterations, the inertia weight ω, and the elite learning control factor c are determined. Specifically: the number of the particle swarm is set to 25, the number of iterations is set to 50, the inertia weight ω is set to 0.9, and the elite learning control factor c is set to 2.

8. The subway train delay adjustment method considering passenger flow influence and regenerative braking energy utilization according to claim 1, characterized in that, The calculation steps of the particle fitness value in step (5.9) include: (a) Take the i-th particle in the population and decode the train adjustment strategy corresponding to this individual, where the initial value of i is 0; (b) Import the operation adjustment method decoded from the particle into the train delay adjustment model; (c) Conduct a train delay adjustment simulation and calculate the fitness value of the individual according to the operation method; (d) Save the individual fitness value of the particle; (e) Determine whether the current particle is the last particle in the population: if so, the calculation ends; otherwise, i = i + 1, and jump to (a).

Images