CN112070325B

CN112070325B - Road network train optimization method, device, equipment and storage medium under abnormal event

Info

Publication number: CN112070325B
Application number: CN202011257253.9A
Authority: CN
Inventors: 许心越; 许旺土; 刘军; 秦勇; 李建民; 王铭铭; 马慧茹
Original assignee: Xiamen University; Beijing Jiaotong University
Current assignee: Xiamen University; Beijing Jiaotong University
Priority date: 2020-11-12
Filing date: 2020-11-12
Publication date: 2021-02-26
Anticipated expiration: 2040-11-12
Also published as: CN112070325A

Abstract

The invention provides a road network train optimization method, a road network train optimization device, road network train optimization equipment and a storage medium under an abnormal event, and relates to a high-speed railway road network train adjustment optimization technology, wherein the method extracts a time-space characteristic set of a train to a station late point from historical train operation data; and constructing and training an extreme learning machine late prediction model based on the space-time feature set in combination with a particle swarm optimization algorithm, and predicting the train arrival late time by using the extreme learning machine late prediction model. And constructing an interval passing capacity optimization model under the abnormal event with the aim of minimizing the train delay time and the average running time, and calculating and outputting the interval passing capacity of the train. And establishing a train dispatching optimization model aiming at the minimum time of the trains at the later points and the minimum number of the trains at the later points, and calculating and outputting a train dispatching optimization strategy. The method provides a scheduling decision basis for a train dispatcher, and is beneficial to improving the railway running organization efficiency and guaranteeing the running safety under the condition of high-speed railway network formation.

Description

Road network train optimization method, device, equipment and storage medium under abnormal event

Technical Field

The invention relates to a method and a system for adjusting and optimizing a high-speed railway road network train, in particular to a road network train optimizing method, a road network train optimizing device, road network train optimizing equipment and a storage medium under an abnormal event.

Background

In recent years, high-speed railways in China are rapidly developed. However, the high-speed train may encounter various random and emergency situations during the operation process, such as bad weather, equipment failure, operator error, etc.; the probability of random interference on train operation is increased along with the extension of the train operation distance, and the influence caused by abnormal events is further enlarged through a network effect, and the influence is possibly brought to the operation of adjacent stations, lines and even regional networks, so that the occurrence of train late points and associated late points is caused, and the difficulty of train operation adjustment is increased; even, the situation of line capability failure or complete failure is very easy to occur under the disturbance of an abnormal event, and the situation further evolves to large-area road network functional paralysis, so that the road network capability loss is serious.

Disclosure of Invention

In order to solve the above problem, an embodiment of the present invention provides a method, an apparatus, a device, and a storage medium for optimizing a road network train in an abnormal event, where the method includes:

extracting a time-space characteristic set of a train to a station late point from historical train operation data;

establishing and training an extreme learning machine late prediction model based on the space-time feature set in combination with a particle swarm optimization algorithm, wherein the extreme learning machine late prediction model is used for predicting train-to-station late;

and predicting the late time of the train arriving at the station by using the extreme learning machine late prediction model.

Further optionally, the method further comprises:

constructing an interval passing capacity optimization model under an abnormal event with the aim of minimizing train delay time and average running time;

and calculating and outputting the interval passing capacity of the train according to the interval passing capacity optimization model.

Further optionally, the method further comprises:

establishing a train dispatching optimization model aiming at the minimum time of the train at a later point and the minimum number of trains at the later point;

and calculating and outputting a train dispatching optimization strategy according to the train dispatching optimization model.

Further optionally, calculating train interval operation and station operation time data by acquiring the actual operation data and the planned operation data;

further optionally, the establishing and training of the extreme learning machine late prediction model by combining the particle swarm optimization algorithm comprises:

s131: generating a population, the number of particles in the population

Number of iterations

Each particle having its own position

And velocity

The position of each particle in the population is equivalent to the number of neurons in a hidden layer of an extreme learning machine;

s132: inputting a processed late point space-time feature set and a Particle position generated by a Particle Swarm Optimization (PSO) by using an Extreme Learning Machine (ELM) algorithm, calculating a fitness function of the PSO, updating the position and the speed of the Particle according to the fitness, and continuously searching;

s133: the position of the first particle is used as the optimal value of the positions of all the particles in the first iteration, and then the corresponding fitness of the first particle is used as the optimal fitness in the current iteration step;

s134: calculating the fitness of each particle, and comparing the optimal fitness of each particle

Global optimum fitness in current iteration step

Iteratively updating the optimum fitness and the location of its particles, if

Then search for the firstlThe global optimum fitness for +1 particle is

(ii) a If it is not

，

. After the iteration is finished, the current optimal fitness is the global optimal fitness

，

。

S135: and entering a loop, updating the positions and the speeds of all the particles, then repeating the step S134, and if the maximum iteration number is exceeded, jumping out of the loop.

S136: training an extreme learning machine late prediction model by combining a particle swarm optimization algorithm;

further optionally, constructing an interval transit capability optimization model under the abnormal event with the minimum train delay time and the minimum average running time as objective functions comprises:

the minimum delay time and the minimum average running time of the train are taken as model objective functions;

constructing a first model constraint condition based on the model objective function;

calculating and outputting the interval passing capacity according to the interval passing capacity optimization model as follows: and solving a model objective function under the first model constraint condition to obtain and output an interval passing capacity value of the train under the abnormal event.

Further optionally, the first model constraint condition includes: train operation time constraint, arrival safety interval time constraint, departure safety interval time constraint and train collinear operation constraint in a fault section.

Further optionally, the collinear operation constraint of the trains in the fault section is that when the planned departure time of the train in the fault direction is within the fault time period, the train in the fault direction needs to enter the opposite track to operate, and the train running on the opposite track needs to meet the time constraint of leaving and arriving safety intervals.

Further optionally, the train scheduling optimization model aiming at the minimum train delay time and the minimum train delay number comprises:

taking the minimum total delay time of the trains and the minimum number of delay trains as a model objective function;

constructing a second model constraint condition based on the model objective function, and establishing a double-layer objective planning model;

the calculating and outputting the train dispatching optimization strategy according to the train dispatching optimization model comprises the following steps: and solving a model objective function under the second model constraint condition to obtain and output a train dispatching optimization strategy under an abnormal event.

Further optionally, the second model constraint condition includes: the train departure interval constraint, the train arrival interval constraint, the interval minimum running time constraint, the train departure time constraint, the train arrival time constraint, the train operation time constraint, the overtravel late train constraint and the station capacity constraint.

Based on the same inventive concept, the invention also provides a road network train optimization device under the abnormal event, which comprises:

the train late point prediction module is used for extracting a space-time feature set of a train arrival late point from historical train operation data, constructing an extreme learning machine late point prediction model based on the feature set and a particle swarm optimization algorithm, wherein the extreme learning machine late point prediction model is used for predicting the train arrival late point, and the extreme learning machine late point prediction model is used for predicting the train arrival late point time;

the train passing capacity analysis module is used for establishing a station passing capacity calculation model based on station operation route distribution optimization;

the train interval capacity analysis module is used for constructing an interval passing capacity optimization model under the abnormal event by taking the minimum train delay time and the minimum average running time as objective functions;

the train dispatching optimization module is used for a train dispatching optimization model aiming at the minimum time of the train at a later point and the minimum number of trains at the later point;

and the information issuing module is used for outputting the late prediction time, the station passing capacity, the interval capacity and the train dispatching strategy under the abnormal event.

Further optionally, the apparatus further comprises:

the data management module is used for importing and inquiring various data based on the database and statistically analyzing the time-space distribution characteristics of the arrival and departure delay points of the train;

the risk management module is used for realizing the functions of risk probability analysis of the abnormal events and risk calculation and prediction of the abnormal events;

an embodiment of the present invention further provides an electronic device, including: at least one processor; and a memory coupled to the at least one processor; wherein the memory stores a computer program executable by the at least one processor to implement the method described above.

The present invention provides a computer-readable storage medium for storing a computer program which, when executed, is capable of carrying out the above-mentioned method.

The invention provides a road network train optimization method, a road network train optimization device, road network train optimization equipment and a storage medium under an abnormal event, which are used for solving the technologies of train late prediction, high-speed railway capacity analysis, scheduling strategy generation and the like under the abnormal event, pushing train late information to passengers and managers, meeting the traveling demands and the operation management demands of the passengers, providing scheduling decision basis for train dispatchers, contributing to improving the railway running organization efficiency under the high-speed railway network formation condition and ensuring the running safety.

Drawings

Fig. 1 is a schematic flow chart of a road network train optimization method under an abnormal event according to an embodiment of the present invention;

fig. 2 is a schematic flow chart of a road network train optimization method under an abnormal event according to another embodiment of the present invention;

FIG. 3 is a schematic flow chart of an implementation of step S22 in FIG. 2;

FIG. 4 is a schematic flow chart of an implementation of step S23 in FIG. 2;

FIG. 5 is a schematic flow chart of an implementation of step S25 in FIG. 2;

FIG. 6 is a schematic flow chart of an implementation of step S26 in FIG. 2;

FIG. 7 is a flowchart illustrating an implementation of step S263 in FIG. 5;

fig. 8 is a schematic structural diagram of a road network train optimization device under an abnormal event according to an embodiment of the present invention;

FIG. 9 is a flowchart of late prediction based on an extreme learning machine in an embodiment of the present invention;

FIG. 10 is a flow chart of a multi-objective interval capability optimization model in an embodiment of the present invention;

FIG. 11 is a flow chart of an event-activity diagram based train dispatch optimization model in an embodiment of the present invention;

FIG. 12 is a flow chart of a Bayesian prediction model in an embodiment of the present invention;

FIG. 13 is a train operation diagram after adjustment of a road network train optimization method in an abnormal event according to an embodiment of the present invention;

fig. 14 is an analysis diagram of train numbers at a station late point in the road network train optimization method under an abnormal event according to the embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments. Examples of the embodiments are illustrated in the accompanying drawings, and specific embodiments described in the following embodiments of the invention are provided as illustrative of the embodiments of the invention only and are not intended to be limiting of the invention.

Example 1

The embodiment provides a road network train optimization method under an abnormal event, which comprises the following steps:

s11, extracting a space-time feature set of the train to the station late point from the historical train operation data;

s12, constructing and training a limit learning machine late prediction model based on a space-time feature set and a particle swarm optimization algorithm, wherein the limit learning machine late prediction model is used for predicting train-to-station late;

s13, predicting the train arrival late time by using the extreme learning machine late prediction model.

According to the method and the device, the off-late prediction of the abnormal event is realized, the train late information is pushed to passengers and managers in real time, the travel demands of the passengers and the management demands of operation are met, the scheduling decision basis is provided for train dispatchers, the improvement of the railway running organization efficiency under the high-speed railway network forming condition is facilitated, and the running safety is guaranteed.

Example 2

The embodiment provides a road network train optimization method under an abnormal event, as shown in fig. 2, the method includes:

and S21 obtaining train operation data.

Wherein the train operation data at least comprises: the data processing system comprises one of planned operation data, road network topological structure data, train actual operation data, abnormal event data and train automatic protection system data. Calculating train interval operation and station operation time data by acquiring operation data and planned operation data, wherein the station operation time data is calculated from the existing data;

s22 analyzes the abnormal event risk probability based on the acquired train data, and calculates and predicts the risk of the abnormal event and the initial degree of delay.

In some embodiments, as shown in fig. 2, step S22 may include, but is not limited to, being implemented by the following process:

s2201, an initial train late reasoning model based on a static Bayesian network is constructed based on input variables of the actual operation data and the fault realistic data selection model, and the risk probability of the abnormal event is obtained.

S2202, according to the type of the event (the type of the event is shown in table 1) to which the abnormal event belongs and the severity of the fault (the severity of the fault is related to the duration of the abnormal event, and specific values are shown in table 2 below), the train initial late point inference model predicts the initial late point degree.

TABLE 1 event types

TABLE 2 severity of failure

S2203, obtaining a failure risk value of the train in the section according to the risk probability of the abnormal event and the predicted delay time, and outputting the train number and the initial delay degree of the affected train.

S23, extracting a space-time feature set of the train arriving at the station late point according to the planned operation data, the actual operation data and the initial late point degree data, and constructing and training a late point prediction model of the extreme learning machine based on the space-time feature set and a particle swarm optimization algorithm to output the late point prediction time of the station, wherein the late point refers to all the late point time including the initial late point and possible associated late points.

As shown in fig. 4, S23 may include, but is not limited to, being implemented by the following processes:

s231: generating a population, setting the number of particles in the population

20, number of iterations

Set to 10, each particle has its own position

And velocity

The position of each particle in the population is equivalent to the number of neurons in the hidden layer of an extreme learning machine, namely:

(ii) a In the formula:

is shown as

At the time of the next iterationlThe number of hidden layer neurons of the extreme learning machine represented by the particle.

S232: inputting a well processed late point space-time characteristic set by using an ELM (hidden layer activation function sigmoid function)

And the particle position (number of hidden layer neurons) generated by the PSO, calculating a fitness function of the PSO, updating the position and the speed of the particle according to the fitness, and continuously searching:

in the formula:Nto test feature sets

The number of samples of (a);

the actual output value is the actual input value, namely the arrival late time of the train at the next station;

the predicted value is the time when the neuron number of the hidden layer of the extreme learning machine is equal to the predicted output value, namely the arrival late time of the train at the next station predicted by the model

Predicting the obtained value;

are particles in the population.

S233: and the position of the first particle is used as the optimal value of the positions of all the particles in the first iteration, and then the corresponding fitness of the first particle is used as the optimal fitness in the current iteration step.

S234: calculating the fitness of each particle, and comparing the optimal fitness of each particle

Global optimum fitness in current iteration step

Iteratively updating the optimum fitness and the location of its particles, if

Then search for the firstlThe global optimum fitness for +1 particle is

(ii) a If it is not

，

，

。

S235: and entering a loop, updating the positions and the speeds of all the particles, repeating the step S234, and jumping out of the loop if the maximum iteration number is exceeded.

；

In the formula:

and

the constant of the acceleration is constant and the acceleration is constant,

is a factor of the inertia, and is,

and

is [0,1 ]]A random number within a range;

s236: training an extreme learning machine late prediction model based on parameter adjustment of a particle swarm optimization algorithm.

S24: and constructing an interval passing capacity optimization model under the abnormal events with the aim of minimizing the train delay time and the average running time. And calculating and outputting the section passing capacity of the train according to the section passing capacity optimization model.

In some alternative embodiments, as shown in fig. 5, step S24 may be implemented by, but is not limited to, the following:

mode 1:

s501, taking the minimum train delay time and the minimum average running time as model objective functions;

s502, constructing a first model constraint condition based on the model objective function;

s503, solving the model objective function under the first model constraint condition to obtain and output the interval passing capacity value of the train under the abnormal event.

Mode 2:

the minimum time division of the total delay of the train and the maximum total weight of the operation route are used as two targets;

integrally applying the route and the departure line of the throat area, and constructing space-time untwining constraint between train operation routes in a segmented unlocking mode;

constructing other constraint conditions of the model based on the objective function;

other constraints of the model include: abnormal event constraints, route occupancy uniqueness constraints, throat area route and departure route uniqueness constraints, departure route continuity constraints, track circuit section locking and unlocking constraints, train operation time constraints, throat occupancy compatibility constraints, and departure occupancy compatibility constraints.

S25, establishing a train dispatching optimization model aiming at the minimum time of the train at the late point and the minimum number of the trains at the late point, and calculating and outputting a train dispatching optimization strategy according to the train dispatching optimization model.

In some alternative embodiments, as shown in fig. 4, step S25 may include, but is not limited to, being implemented by the following processes:

constructing a model constraint condition based on the model objective function;

the model constraints include: train operation time constraint, arrival safety interval time constraint, departure safety interval time constraint and train collinear operation constraint in a fault section. And the collinear operation constraint of the trains in the fault interval is that when the planned departure time of the train in the fault direction is in the fault time interval, the train in the fault direction needs to enter the opposite track to operate, and the train running on the opposite track needs to meet the time constraint of the departure and arrival safety interval.

And optimizing and solving a model objective function by a branch cutting algorithm combining a branch-and-bound method and a secant plane method and outputting an interval passing capacity value.

Based on the functions of road network capacity bottleneck identification, station and interval capacity analysis and the like under abnormal events; constructing a road network capacity weight network according to the road network topological structure data, the station capacity and the interval capacity data; and identifying the bottleneck of the road network service capability by using a spectral clustering algorithm.

S26, establishing a train dispatching optimization model based on the late prediction time and the event-activity diagram;

and constructing a scheduling optimization model under the abnormal event based on the late prediction time output in the step S23, the station passing capacity output in the step S24 and the section capacity data output in the step S25, and solving the model to generate a train scheduling strategy.

In some alternative embodiments, as shown in fig. 6, step S26 may include, but is not limited to, being implemented by the following processes:

s261 determining the train number and the section under the abnormal event;

s262, according to planned departure and arrival time of the train, actual departure and arrival time of the train, train number and section under abnormal events, establishing a train event-activity diagram by taking the operation logical relationship among trains as an arc;

s263, constructing a train dispatching optimization model based on the train event-activity diagram;

s264 establishes a double-layer target planning model by taking the minimum total late train time and the minimum number of late trains as target functions.

In some alternative embodiments, as shown in fig. 7, step S263 may include, but is not limited to, being implemented by the following processes:

s2631, constructing a minimum train total delay time function according to planned arrival time, actual arrival time and train total delay time penalty factors;

s2632, constructing a minimum late train quantity function according to planned time, actual time and late train quantity penalty factors of train departure;

s2633 constructs a minimum objective function based on the minimum train total late point time function and the minimum late point train number function.

And S27, outputting the late prediction time, the station passing capacity, the section capacity and the train dispatching result in the abnormal event.

And according to the S initial delay degree, the delay prediction time output by the S23, the obtained road network capacity bottleneck, the station passing capacity and interval capacity values and the generated train scheduling strategy, realizing the functions of real-time distribution information of the train delay of the road network, dynamic display of statistical information of the actual running delay of the train, the road network capacity bottleneck, the station and interval capacity, pushing of the macroscopic scheduling strategy of the train and the like.

Example 3

In a preferred embodiment, there is provided a road network train optimization device in abnormal event, as shown in fig. 8, the device includes:

the train data acquisition module 10 is used for acquiring one of planned operation data, road network topological structure data, train actual operation data, abnormal event data, train ATP data and fault writing data; calculating time data of train interval operation and station operation by acquiring operation data and planned operation data;

a risk analysis module 20 for analyzing the risk probability of the abnormal event based on the acquired train data, and calculating and predicting the risk and the initial late degree of the abnormal event;

the train late point prediction module 30 is used for extracting a space-time feature set of a train arriving at a station late point, and constructing an extreme learning machine late point prediction model through a particle swarm optimization algorithm based on the space-time feature set so as to output late point prediction time;

the train passing capacity analysis module 40 is used for establishing a station passing capacity calculation model based on the later prediction time and station operation route distribution optimization;

the train interval capacity analysis module 50 is used for constructing an interval capacity optimization model under an abnormal event based on the late prediction time, the minimum train delay time and the minimum average running time;

a train dispatching optimization module 60 for establishing a train dispatching optimization model based on the late prediction time and the event-activity diagram;

and the information issuing module 70 is used for outputting the late prediction time, the station passing capacity, the interval capacity and the train dispatching result under the abnormal event.

In a more preferred embodiment, there is provided an electronic device including: at least one processor; and a memory coupled to the at least one processor; wherein the memory stores a computer program executable by the at least one processor to implement the above method.

In a more preferred embodiment, a computer-readable storage medium is provided for storing a computer program which, when executed, is capable of carrying out the above-mentioned method.

Example 4

The embodiment provides another road network train optimization method under an abnormal event, which comprises the following steps:

s1: predicting the late time of the train based on the late prediction model of the extreme learning machine; the specific prediction process is shown in FIG. 9.

S11: acquiring data of a plan operation diagram and an actual operation diagram;

TABLE 3 train operating data

S12: determining a characteristic set influencing the late point of arrival of the train;

determining station sequencing numbers, station numbers, train numbers, scheduled operation time of a previous station to the current station, a distance between the current station and the previous station, scheduled operation time of the current station to the next station, a distance between the current station and the next station, actual operation time of the previous station to the current station and 9 characteristic variables of the current station to a late point as inputs of a late point prediction model;

s13: constructing a limit learning machine late prediction model based on parameter adjustment of a particle swarm optimization algorithm;

s131: generating a population, setting the number of particles in the population

20, number of iterations

Set to 10, each particle has its own position

And velocity

in the formula:

is shown as

At the time of the next iterationlThe number of neuron of hidden layer of extreme learning machine represented by particle;

s132: inputting a processed feature set by using an ELM algorithm (hidden layer activation function sigmoid function)

And the positions (the number of hidden layer neurons) of the particles generated by the PSO, and outputting a weight matrix of the extreme learning machine under the current number of hidden layer neurons, wherein a function for calculating the fitness is as follows:

；

；

in the formula:Nthe number of samples;

is an actual output value, is an actual input value;

for predicting the output value, the predicted value is that the number of neurons in the hidden layer of the extreme learning machine is

Predicting the obtained value;

are particles in a population;

s133: the position of the first particle is used as the optimal value of the positions of all the particles in the first iteration, and the corresponding fitness of the particle is used as the current optimal fitness;

Best fitness with current global situation

Updating the current best fitness and the position of the particle if

Then search for the firstlThe global optimum fitness for +1 particle is

(ii) a If it is not

，

。

；

；

；

In the formula:

and

the constant of the acceleration is constant and the acceleration is constant,

is a factor of the inertia, and is,

and

is [0,1 ]]A random number within a range;

s14: training an extreme learning machine late prediction model based on parameter adjustment of a particle swarm optimization algorithm;

s15: outputting a late prediction result;

the results of calculating the evaluation indexes of the different models using the training set are shown in table 4. As can be seen from table 4, the R-squared of the ELM-PSO (extreme learning machine-particle swarm optimization) is as high as 0.9838, the RMSE is the lowest compared to three models, namely ANN (artificial neural network), Decision Tree (Decision Tree) and Lasso (regression model), and the MAE is the lowest compared to all baseline models (ANN, Decision Tree, Lasso and KNN (K nearest neighbor classification algorithm)), so the ELM-PSO model has better fitting effect. To verify the effectiveness of each model, the test set data is input into the trained model, and as a result, as shown in table 5, it can be seen that the respective evaluation index values of the ELM-PSO in the test set are slightly inferior to those of the training set, but the results are still the best compared with the four baseline models. Therefore, the ELM-PSO prediction model has a good effect on the delay prediction of the high-speed rail late point.

TABLE 4 comparison of results for six models (training set)

TABLE 5 comparison of results for six models (test set)

S2: and establishing a station passing capacity calculation model based on station operation route distribution optimization.

Step1, determining abnormal event characteristics including station abnormal event range definition, scenario division and duration estimation;

step2, constructing abnormal event constraints of different scenes;

step3, constructing space-time untwining constraints of trains between station operation routes according to the time sequence relationship of track circuit subsection unlocking;

step4, establishing a station trafficability calculation model based on station operation route distribution optimization;

and Step5, adopting Python and Gurobi to solve a model and calculating an optimal route allocation scheme under the condition of station passing capacity and capacity application.

S3: and (3) constructing a multi-target interval capacity optimization model under abnormal events, and referring to FIG. 10.

S31: and taking the minimum train delay time and the minimum average running time as model objective functions.

；

；

；

In the formula:

representing the total delay time of the train;

representing the average train running time;

representing a train;

representing a set of trains;

representing a station;

representing a station set;

representing trains

At station

Actual departure time of;

representing trains

At station

The actual arrival time of;

representing trains

At station

The planned departure time of (a);

s32: constructing model constraints, including: train operation time constraints, arrival and departure safety interval time constraints, train co-linear operation between fault zones constraints, and other constraints.

The train operation time is as follows:

；

；

；

；

the arrival and departure safety interval time constraint is as follows:

the collinear operation constraint of the trains in the fault interval is that when the planned departure time of the train in the fault direction is in the fault time interval, the train in the fault direction needs to enter the opposite track to operate, and the train running on the opposite track needs to meet the time constraint of the departure and arrival safety intervals.

The other constraints are:

，

in the formula:

representing trains

In the interval

Minimum run time ofA (c) is added;

representing a train;

representing trains

At station

Minimum station-stop time of;

is a sufficiently large positive integer;

representing trains

From the starting station

To the final station

The maximum run time of;

is the fault start time;

is the fault end time;

representing trains

And

at station

A departure safety interval time;

representing trains

And

at station

Time to safety interval;

representing trains

At station

The planned departure time of (a);

and

train for respectively indicating fault directions

At station

Departure and arrival times of;

representing trains

Whether or not at station

Entering a binary variable of opposite track operation;

representing trains

And a train

Arrival and departure station

Binary variable of sequence of when the train

And prior to the train

Arrival and departure station

The value is 1 when the current value is not zero, otherwise

The value is 1;

and

respectively representing trains

At terminal station

And a departure station

Actual arrival and departure times of;

indicating that type 1 train is at station when fault occurs

Is equal to 1 type train at the station when the fault occurs

Planned departure time of

；

Indicating that type 1 train is at station when fault occurs

Is equal to the actual arrival time of the type 1 train at the station when the fault occurs

Scheduled arrival time of

；

S33: and (3) model solving, wherein MATLAB R2018b is adopted for programming, a branch cutting algorithm which is arranged in the CPLEX solver and combines a branch-and-bound method and a cutting plane method is called to solve the optimized model, and the related parameters of CPLEX are set as default values.

S34: outputting an interval capacity value; the input data for the numerical case is shown in table 6.

Table 6 numerical case input data

Note: TID-train ID, SID-station ID, Direct-train running direction, U-up, D-down, x-stop-free.

The MATLAB R2018b is adopted for programming, and a CPLEX solver is called to solve the optimization model to obtain an optimized optimal time table shown in a table 7;

TABLE 7 optimal time schedule for numerical case output

S4: establishing a train dispatching optimization model based on the event-activity diagram;

the detailed flow chart is shown in FIG. 10;

s41: abnormal events affect train number and section determination;

s42: constructing a train event-activity diagram, namely constructing the train event-activity diagram by taking arrival and departure events or passing events of a train at a station as nodes and taking a running logic relation among trains as an arc according to data such as a railway event influence interval, an influence train number and a train plan running diagram;

s43: establishing a train dispatching optimization model;

s431: and establishing a double-layer target planning model by taking the minimum total late train time and the minimum number of late trains as target functions.

The total late time of the train is minimum:

；

the number of trains at a late spot is minimum:

；

；

defining functions

；

In the formula:

represents the minimum total train late time;

representing a minimum number of train delays;

representing a minimum objective function value;

representing trains

At station

The planned arrival time of (c);

respectively a train total late time penalty factor and a late train quantity penalty factor.

S432: constraint conditions

Restraint of train departure intervals:

in the formula:

representing the minimum departure interval time of the train;

train arrival interval constraint:

in the formula:

representing a train minimum inter-arrival time;

and (3) interval minimum operation time division constraint:

，

in the formula:

indication interval

The minimum operation time division;

is shown as

Train at station

Additional time division is started;

is shown as

Train at station

Additional time divisions of station.

Restraint of train departure time:

；

and (3) train arrival time constraint:

；

and (3) restricting train operation time:

in the formula:

is shown as

Train at station

The standard working time of (2);

restraint of the overtaking late train:

；

station capacity constraint:

；

wherein,

，

indicating station

The number of arrival lines;

representing the moment of the train operation adjusting stage;

indicating the train operation adjustment phase duration.

S44: and outputting a train dispatching strategy under the abnormal event.

The macro level model is initially established, and verified by the jinghu high speed railway (beijing south-junan west) example, the influence of several train delay points caused by setting up the emergency is shown in table 8. The train operation adjustment result obtained by model solution is shown in fig. 12 and 13; it can be seen from fig. 12 and 13 that the late point of each adjusted train number is decreasing, indicating that the event-activity based train operation adjustment model described above is valid.

TABLE 8 initial late situation table

The road network train adjusting and optimizing system under the abnormal event comprises a data management module, an abnormal event risk management module, a train late prediction module under the abnormal event, a high-speed railway capacity analysis module under the abnormal event, a high-speed railway dispatching strategy generation module under the abnormal event and a train operation information display and release module,

the data management module is mainly used for importing and inquiring various data based on a database and statistically analyzing the time-space distribution characteristics of arrival and departure delay points of the train;

the abnormal event risk management module mainly realizes the functions of abnormal event risk probability analysis and abnormal event risk calculation and prediction;

the train late prediction module under the abnormal event mainly predicts the late time of the train by constructing an extreme learning machine late prediction model;

the high-speed railway capacity analysis module under the abnormal event mainly realizes the functions of road network capacity bottleneck identification, station and interval capacity analysis and the like;

the high-speed railway scheduling strategy generating module under the abnormal event is mainly used for constructing a scheduling optimization model under the abnormal event and generating a train scheduling strategy;

the train operation information issuing module is used for issuing the time-space distribution information of the train at the rear of the road network in real time based on the output result of the module.

The main functions of the functional modules of the system are shown in table 9,

main function of each functional module of table 9 system

The road network train adjustment and optimization system under the abnormal event is completed by using the road network train adjustment and optimization method under the abnormal event, and the system comprises the following steps:

step1, the data management module acquires data such as road network topological structure, train actual performance operation diagram, plan operation diagram and abnormal event, the content of the imported data of the data management module is shown in table 10:

table 10 content of data imported by data management module

Calculating to obtain data such as train section operation time, station operation time and the like according to actual performance train operation diagram data and planned operation diagram data, wherein the calculating method comprises the following steps:

train section

Run time = train arrival station

Time D-train slave station

The departure time of (c);

train at station

Working time of (2 = train slave station)

Departure time of the train to the station

The time of day;

analyzing the time-space characteristics of the train at a later point based on actual performance train operation data;

the case statistically analyzes the time of the late point of each station in the section from south to north with red wall of Guangzhou, including south station of Guangzhou, Shaoguan station, east station of Hengyang, West station of Rizhou, east station of Yueyang and north with red wall, and the analysis result is shown in table 11, taking the south station of Guangzhou as an example, within the time span, the trains with the late point time of more than 10min account for about 2.5%, and the number of the trains at the late point is 2589 columns.

TABLE 11 statistics of late arrival of trains

Step2, an abnormal event risk management module, which analyzes the abnormal event risk probability and calculates and predicts the risk of the abnormal event, wherein Step2 specifically comprises the following steps:

step21, selecting input variables of the model based on the train operation actual performance data and the fault record data, and constructing a train initial late reasoning model based on a static Bayesian network to obtain the risk probability of the abnormal event;

step22, obtaining the predicted late time by a Bayes model according to the type of the emergency and the severity of the fault;

step23, obtaining the failure risk value of the train in the section according to the probability of the abnormal event and the influence consequence in the later time, and outputting the train number, the initial later degree and the later prediction time of the influenced train, specifically referring to FIG. 11;

inputting abnormal event data, train planning operation diagram data and actual performance operation data, wherein the abnormal event data comprises: the abnormal event occurrence time, the abnormal event ending time, the abnormal event occurrence position and the abnormal event directly affect the train number, the abnormal event type, the abnormal event duration, the fault type and the fault severity, wherein the fault severity is related to the abnormal event duration, and specific values are shown in table 12:

TABLE 12

The late point degree is evaluated according to the initial late point duration, as shown in table 13:

watch 13

The method comprises the following steps that a worker sets an abnormal event scene, inputs the type, the occurrence time, the occurrence position, the end time and the fault type of an abnormal event, and outputs a predicted value through a Bayesian model;

TABLE 14

As shown in table 14, the variable state with the highest posterior probability is regarded as the prediction result. Taking foreign invasion as an example, the train late point state which is most probably caused after the I-level fault occurs is the I-level late point, and the train late point state which is most probably caused after the II-level fault and the III-level fault occur is the II-level late point. And finally, searching a conditional probability table of a Bayes model according to the type of the emergency and the severity of the fault, selecting the delay degree with the maximum posterior probability as the initial delay state of the train, and setting the delay time as the average value in the delay degree period.

Step3, a train late prediction module under an abnormal event is used for constructing an extreme learning machine late prediction model to predict the late time of the train, and Step3 specifically comprises the following steps:

step 31: acquiring a planned operation diagram, an actual performance operation diagram and initial late degree data output by Step 23;

step 32: determining a characteristic set influencing the late point of arrival of the train;

step 33: constructing a limit learning machine late prediction model based on parameter adjustment of a particle swarm optimization algorithm;

step 34: training an extreme learning machine late prediction model based on parameter adjustment of a particle swarm optimization algorithm;

step 35: and outputting the late prediction result.

Step4, a high-speed railway capacity analysis module in abnormal events realizes the functions of road network capacity bottleneck identification, station and section capacity analysis and the like in abnormal events based on the late prediction time output by Step35, wherein Step4 specifically comprises the following steps:

step 41: constructing a passing capacity calculation model based on station operation route distribution optimization, and solving the model to output a station capacity value;

step 42: constructing a multi-target interval capacity optimization model under an abnormal event, and outputting an interval capacity value;

step 43: constructing a road network capacity weight network according to the road network topological structure, the Step41 station capacity and the Step42 interval capacity data; and identifying the bottleneck of the road network service capability by using a spectral clustering algorithm.

Step5, namely, a high-speed railway dispatching strategy generation module under an abnormal event, constructing a dispatching optimization model under the abnormal event based on the late prediction time output by Step35, the station capacity output by Step41 and the interval capacity output by Step42, solving the model to generate a train dispatching strategy, wherein Step5 specifically comprises the following steps:

s51: abnormal events affect train number and section determination;

s52: constructing a train event-activity diagram;

s53: establishing a train dispatching optimization model;

s54: and outputting a train dispatching strategy under the abnormal event.

The data flow relationship among the function modules described in Step2, Step3, Step4 and Step5 is shown in fig. 14;

and Step6, a train operation information issuing module, which is used for realizing the functions of issuing the time-space distribution information of the late point of the train in the road network in real time, dynamically displaying the capacity of the bottleneck of the road network, the station and the section, pushing the macro scheduling strategy of the train and the like according to the initial late point degree obtained by the Step2, the late point time of the train predicted by the Step3, the capacity bottleneck of the road network obtained by the Step4, the capacity values of the station and the section and the train scheduling strategy generated by the Step 5.

The embodiment of the present invention further provides a computer-readable storage medium, on which computer instructions are stored, and when the computer instructions are executed, the steps of the road network train adjustment and optimization system under the abnormal event are executed. Please refer to the description in the foregoing section, and details are not repeated.

The embodiment of the present invention further provides an electronic device, which includes a memory and a processor, where the memory stores a computer instruction capable of running on the processor, and the processor executes the steps of the road network train adjustment and optimization system under the abnormal event when running the computer instruction. Please refer to the description in the foregoing section, and details are not repeated.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by instructions associated with hardware via a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.

In summary, according to the road network train adjustment optimization method, device and storage medium under the abnormal event, the extreme learning machine-based late prediction model is adopted to predict the train late, the multi-objective optimization model analysis interval capacity is built, the scheduling optimization model is built to generate the scheduling strategy, so that the late can be effectively predicted, and the capacity value and the scheduling strategy are output; the road network train adjusting and optimizing system under the abnormal event, which is built by the road network train adjusting and optimizing method under the abnormal event, can realize the functions of road network risk management, train late prediction, high-speed railway capacity analysis, macroscopic train dispatching strategy generation and the like under the abnormal event; the train operation information publishing module is used for pushing the train late information to passengers and managers in real time so as to meet the travel demands of the passengers and the management demands of operation and provide scheduling decision bases for train dispatchers.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. The present embodiments are therefore to be considered as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein, and any reference signs in the claims are not intended to be construed as limiting the claim concerned. In the claims, the word "comprising" does not exclude the presence of data or steps not listed in a claim.

Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.

Claims

1. A road network train optimization method under an abnormal event is characterized by comprising the following steps:

predicting the late time of the train arriving at the station by using the extreme learning machine late prediction model;

the method for extracting the space-time feature set of the train to the station late point from the historical train operation data comprises the following steps:

selecting input variables of a model based on actual operation data and fault realistic data, and constructing a train initial late reasoning model based on a static Bayesian network to obtain the risk probability of an abnormal event;

according to the type of the event to which the abnormal event belongs and the severity of the fault, predicting by using an initial delay reasoning model of the train to obtain the initial delay;

obtaining a failure risk value of the train in the section according to the risk probability of the abnormal event and the predicted delay time, and outputting the train number and the initial delay degree of the affected train;

and extracting a space-time feature set of the train to the station late point according to the planned operation data, the actual operation data and the initial late point degree data.

2. The method for optimizing a train in a road network under an abnormal event according to claim 1, further comprising:

3. The method for optimizing a train in a road network under an abnormal event according to claim 2, further comprising:

4. The road network train optimization method under the abnormal event according to claim 1, wherein the building and training of the extreme learning machine late prediction model by combining the particle swarm optimization algorithm comprises:

s131: generating a population, the number of particles in the population

Number of iterations

Each particle having its own position

And velocity

s132: inputting the processed late point time-space feature set and the particle position generated by the particle swarm optimization algorithm by using an extreme learning machine algorithm, calculating a fitness function of the particle swarm optimization algorithm, updating the position and the speed of the particle according to the fitness, and continuing searching;

s133: the position of the first particle is used as the optimal value of the positions of all the particles in the first iteration, and the corresponding fitness of the first particle is used as the optimal fitness in the current iteration step;

Global optimum fitness in current iteration step

Iteratively updating the optimum fitness and the location of its particles, if

Then search for the first

The global optimum fitness for each particle is

(ii) a If it is not

，

After the iteration is finished, the current optimal fitness is the global optimal fitness

，

；

S135: entering a loop, updating the positions and the speeds of all the particles, repeating the step S134, and jumping out of the loop if the maximum iteration times is exceeded;

s136: and training an extreme learning machine late prediction model by combining a particle swarm optimization algorithm.

5. The abnormal event down-road network train optimization method according to claim 2, wherein constructing the abnormal event down-interval passing capability optimization model with the objective of minimizing train delay time and minimizing average running time comprises:

calculating and outputting the section passing capacity of the train according to the section passing capacity optimization model as follows: and solving a model objective function under the first model constraint condition to obtain and output an interval passing capacity value under an abnormal event.

6. The road network train optimization method under abnormal events according to claim 5, wherein said first model constraint condition comprises: train operation time constraint, arrival safety interval time constraint, departure safety interval time constraint and train collinear operation constraint in a fault section.

7. The off-normal event road network train optimization method according to claim 6, wherein the collinear operation constraint of the trains in the fault section is that when the planned departure time of the train in the fault direction is within the fault time period, the train in the fault direction needs to enter the opposite track to operate, and the train running on the opposite track needs to meet the departure and arrival safety interval time constraint.

8. The off-normal event road network train optimization method according to claim 3, wherein the establishing of the train scheduling optimization model aiming at the minimum time of the train at the later point and the minimum number of the trains at the later point and the outputting of the scheduling strategy under the off-normal event comprises the following steps:

9. The method for optimizing a train in a road network under abnormal events according to claim 8, wherein said second model constraint condition comprises: the train departure interval constraint, the train arrival interval constraint, the interval minimum running time constraint, the train departure time constraint, the train arrival time constraint, the train operation time constraint, the overtravel late train constraint and the station capacity constraint.

10. An optimization device for a road network train in an abnormal event, the optimization device comprising:

the train late point prediction module is used for extracting a space-time feature set of a train arrival late point from historical train operation data, constructing an extreme learning machine late point prediction model based on the feature set and a particle swarm optimization algorithm, wherein the extreme learning machine late point prediction model is used for predicting the train arrival late point, and the extreme learning machine late point prediction model is used for predicting the train arrival late point time; the method for extracting the space-time feature set of the train to the station late point from the historical train operation data comprises the following steps: selecting input variables of a model based on actual operation data and fault realistic data, and constructing a train initial late reasoning model based on a static Bayesian network to obtain the risk probability of an abnormal event; according to the type of the event to which the abnormal event belongs and the severity of the fault, predicting by using an initial delay reasoning model of the train to obtain the initial delay; obtaining a failure risk value of the train in the section according to the risk probability of the abnormal event and the predicted delay time, and outputting the train number and the initial delay degree of the affected train; extracting a time-space characteristic set of the train arriving at the station late point according to the planned operation data, the actual operation data and the initial late point degree data;

the train station passing capacity analysis module is used for establishing a station passing capacity calculation model based on station operation route distribution optimization;

the train interval passing capacity analysis module is used for constructing an interval capacity optimization model under an abnormal event by taking the minimum train delay time and the minimum average running time as objective functions;

the train dispatching optimization module is used for establishing a train dispatching optimization model which aims at minimizing the time of the train at a later point and minimizing the number of trains at the later point;

and the information issuing module is used for issuing the late prediction time, the station passing capacity, the interval passing capacity and the train dispatching optimization strategy under the abnormal event.

11. An electronic device, comprising:

at least one processor; and

a memory coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to implement the method of any one of claims 1-9.

12. A computer-readable storage medium, characterized in that the computer-readable storage medium is used to store a computer program which, when executed, is capable of implementing the method of any one of claims 1-9.