Methods and internet of things systems for optimizing metro operation scheduling in smart city

1. A method for optimizing metro operation scheduling in a smart city realized by an Internet of Things (IoT) system for optimizing metro operation scheduling in a smart city, wherein the IoT system for optimizing metro operation scheduling in a smart city includes a user platform, a service platform, a management platform, a sensor network platform, and an object platform, the IoT system being a part of or implemented by a processing device, the user platform being configured as a terminal device and used to obtain an input instruction of a user through the terminal device, query an operation scheduling scheme of a target station in a target period of time, issue a metro operation scheduling scheme query instruction to the service platform, receive the operation scheduling scheme uploaded by the service platform, and feed the operation scheduling scheme back to the user, the service platform being used to provide input and output services for the user, receive data uniformly, process the data uniformly, and send the data uniformly, the management platform being used to overall plans and coordinates connection and collaboration among the user platform, the service platform, the management platform, the sensor network platform, and the object platform, and bring together all information of the IoT system, the sensor network platform being configured as a communication network and a gateway device and used to realize functions of perceptual information sensor communication and control information sensor communication, the object platform being configured as various types of devices to obtain information including an image obtaining device, a metro entry gate device, a weight sensor and used to generate the perceptual information, the method being executed by the management platform, and the method comprises:

obtaining, based on the object platform, passenger flow data of at least one metro station related to the target station by the sensor network platform;

determining predicted passenger flow data of the target station in the target period of time based on the passenger flow data of the at least one metro station through a passenger flow prediction model, the passenger flow prediction model being a graph neural networks model, wherein

an input of the passenger flow prediction model further includes collection time information and target period of time information of the target station, an output of the passenger flow prediction model includes the predicted passenger flow data of the target station in the target period of time, wherein the collection time information and the target period of time information both include time information, weather information, and traffic environment information corresponding to a collection time and the target period of time;

the passenger flow prediction model is obtained through a first training based on a plurality of first training samples with first labels, wherein the first training samples include passenger flow data of a plurality of first sample metro stations during a first sample period of time, the first labels include actual passenger flow data of a second sample metro station during a second sample period of time, wherein the first sample metro station in the first training samples is a metro station related to the second sample metro station, the first labels are obtained by the management platform by querying or detecting the actual passenger flow data of the second sample metro station during the second sample period of time;

the first training includes: inputting the plurality of first training samples with first labels into an initial passenger flow prediction model, constructing a first loss function based on the first labels and output results of the initial passenger flow prediction model, updating parameters of the initial passenger flow prediction model based on the first loss function; and obtaining the passenger flow prediction model until the first loss function of the initial passenger flow prediction model meeting a first preset condition, wherein the first preset condition includes a convergence of the first loss function, a loss function value being smaller than a preset value, and a number of iterations reaching a first threshold;

determining a predicted passenger flow distribution of the target station in the target period of time at least based on the passenger flow data of the target station and the target period of time through a passenger flow distribution prediction model, wherein the passenger flow distribution includes a predicted passenger flow distribution interval and an interval probability, and the passenger flow distribution prediction model is a graph neural networks model;

determining the operation scheduling scheme of the target station in the target period of time based on the predicted passenger flow data, the operation scheduling scheme including at least a metro departure interval;

uploading the operation scheduling scheme to the service platform;

transmitting, based on the service platform, the operation scheduling scheme to the user platform; and

feeding, based on the user platform, the operation scheduling scheme back to the user.

2. The method of claim 1 , further comprising:

transmitting the operation scheduling scheme to the service platform by the management platform; and

transmitting the operation scheduling scheme to the user platform by the service platform.

3. The method of claim 1 , wherein

the sensor network platform includes a general database and a plurality of sub-platforms, the sub-platforms are configured with gateway devices, and the sub-platforms upload data obtained based on the object platform to the general database.

4. The method of claim 1 , wherein

the target station is a metro station with the passenger flow data greater than a preset passenger flow threshold, and

the at least one metro station related to the target station includes the target station itself, another station on a same metro line as the target station, or a station on a different line with a distance being smaller than a distance threshold from the target station.

5. The method of claim 1 , wherein the operation scheduling scheme further includes a metro departure time; and

the determining the operation scheduling scheme of the target station in the target period of time based on the predicted passenger flow data includes:

determining the metro departure interval in the target period of time through a reinforcement learning model based on the predicted passenger flow data; and

determining the metro departure time based on the metro departure interval.

6. The method of claim 5 , wherein a reward function in the reinforcement learning model is 0 when the metro departure interval is smaller than a preset time threshold.

7. The method of claim 6 , wherein the reinforcement learning model further includes: determining, based on a predicted passenger flow distribution, a target count of passengers, the reward function being related to the target count of passengers.

8. The method of claim 5 , wherein

the reinforcement learning model includes an environment module and an optimal action determination module, wherein;

an input of the environment module includes environmental state information, an output of the environmental module includes an optional action set, the environment module includes an optional action determination sub-module, a state determination sub-module, and a reward determination sub-module, wherein

the optional action determination sub-module is used to determine the optional action set based on the environmental state information, the state determination sub-module is used to determine the environmental state information, and the reward determination sub-module is used to determine the reward function, wherein the optional action set refers to a set of actions that can be executed by the management platform, and the set of actions that can be executed by the management platform include the metro departure interval;

the optimal action determination module is a graph neural networks model, wherein

an input of the optimal action determination module further includes the environmental state information and the optional action set, an output of the optimal action determination module includes an optimal optional action;

the optimal action determination module is obtained through a third training using a reinforcement learning manner based on a plurality of third training samples with third labels, wherein the reinforcement learning manner includes Q-Learning, state-action-reward-state-action (Sarsa), deep Q-network (DQN), Policy-Gradients, Actor-Critic, the third training samples include historical passenger flow data of the target station, the third labels include an optimal departure interval corresponding to historical environmental state information, wherein the third training samples are obtained based on historical data, and the third labels are obtained through the reinforcement learning manner.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein when reading the computer instructions in the storage medium, a computer implements the method for optimizing metro operation scheduling in a smart city according to claim 1 .

10. The method of claim 1 , wherein

an input of the passenger flow distribution prediction model further includes the collection time information and the target period of time information of the target station, an output of the passenger flow distribution prediction model includes the predicted passenger flow distribution of the target station in the target period of time;

the passenger flow distribution prediction model is obtained through a second training based on a plurality of second training samples with second labels, wherein the second training samples include passenger flow data of a fifth sample metro station in a fifth sample period of time and a sixth sample period of time, the second labels include a passenger flow distribution of the fifth sample metro station in the sixth sample period of time, wherein the fifth sample metro station is a same station as the target station, the fifth sample period of time is earlier than the sixth sample period of time, and the second labels are obtained by the management platform by querying or detecting an actual passenger flow distribution of the fifth sample metro station in the sixth sample period of time;

the second training includes: inputting the plurality of second training samples with second labels into an initial passenger flow distribution prediction model, constructing a second loss function based on the second labels and output results of the initial passenger flow distribution prediction model updating parameters of the initial passenger flow distribution prediction model based on the second loss function; and obtaining the passenger flow distribution prediction model until the second loss function of the initial passenger flow prediction distribution model meeting a second preset condition, wherein the second preset condition includes a convergence of the second loss function, a loss function value being smaller than a preset value, and a number of iterations reaching a second threshold.

11. An Internet of Things system for optimizing metro operation scheduling in a smart city comprising a user platform, a service platform, a management platform, a sensor network platform, and an object platform, the Internet of Things system being a part of or implemented by a processing device, the user platform being configured as a terminal device and used to obtain an input instruction of a user through the terminal device, query an operation scheduling scheme of a target station in a target period of time, issue a metro operation scheduling scheme query instruction to the service platform, receive the operation scheduling scheme uploaded by the service platform, and feed the operation scheduling scheme back to the user, the service platform being used to provide input and output services for the user, receive data uniformly, process the data uniformly, and send the data uniformly, the management platform being used to overall plans and coordinates connection and collaboration among the user platform, the service platform, the management platform, the sensor network platform, and the object platform and bring together all information of the Internet of Things system, the sensor network platform being configured as a communication network and a gateway device and used to realize functions of perceptual information sensor communication and control information sensor communication, the object platform being configured as various types of devices to obtain information including an image obtaining device, a metro entry gate device, a weight sensor and used to generate the perceptual information, wherein the management platform is further configured to:

obtain, based on the object platform, passenger flow data of at least one metro station related to the target station by the sensor network platform;

determine predicted passenger flow data of the target station in the target period of time based on the passenger flow data of the at least one metro station through a passenger flow prediction model, the passenger flow prediction model being a graph neural networks model, wherein

an input of the passenger flow prediction model further includes collection time information and target period of time information of the target station, an output of the passenger flow prediction model includes the predicted passenger flow data of the target station in the target period of time, wherein the collection time information and the target period of time information both include time information, weather information, and traffic environment information corresponding to a collection time and the target period of time;

the passenger flow prediction model is obtained through a first training based on a plurality of first training samples with first labels, wherein the first training samples include passenger flow data of a plurality of first sample metro stations during a first sample period of time, and the first labels include actual passenger flow data of a second sample metro station during a second sample period of time, wherein the first sample metro station in the first training samples is a metro station related to the second sample metro station, the first labels are obtained by the management platform by querying or detecting the actual passenger flow data of the second sample metro station during the second sample period of time;

the first training includes: inputting the plurality of first training samples with first labels into an initial passenger flow prediction model, constructing a first loss function based on the first labels and output results of the initial passenger flow prediction model, updating parameters of the initial passenger flow prediction model based on the first loss function; and obtaining the passenger flow prediction model until the first loss function of the initial passenger flow prediction model meeting a first preset condition, wherein the first preset condition includes a convergence of the first loss function, a loss function value being smaller than a preset value, and a number of iterations reaching a first threshold;

determine a predicted passenger flow distribution of the target station in the target period of time at least based on the passenger flow data of the target station and the target period of time through a passenger flow distribution prediction model, wherein the passenger flow distribution includes a predicted passenger flow distribution interval and an interval probability, and the passenger flow distribution prediction model is a graph neural networks model;

determine an operation scheduling scheme of the target station in the target period of time based on the predicted passenger flow data, the operation scheduling scheme including at least a metro departure interval;

uploading the operation scheduling scheme to the service platform;

transmitting, based on the service platform, the operation scheduling scheme to the user platform; and

feeding, based on the user platform, the operation scheduling scheme back to the user.

12. The Internet of Things system of claim 11 , wherein

the management platform is configured to transmit the operation scheduling scheme to the service platform; and

the service platform is configured to transmit the operation scheduling scheme to the user platform.

13. The Internet of Things system of claim 11 , wherein

the sensor network platform includes a general database and a plurality of sub-platforms, the sub-platforms are configured with gateway devices, and the sub-platforms upload data obtained based on the object platform to the general database.

14. The Internet of Things system of claim 11 , wherein the target station is a metro station with the passenger flow data greater than a preset passenger flow threshold, and

the at least one metro station related to the target station includes the target station itself, another station on a same metro line as the target station, or a station on a different line with a distance being smaller than a distance threshold from the target station.

15. The Internet of Things system of claim 11 , wherein

the operation scheduling scheme also includes a metro departure time; and

the management platform is further configured to:

determine the metro departure interval in the target period of time through a reinforcement learning model based on the predicted passenger flow data; and

determine the metro departure time based on the metro departure interval.

16. The Internet of Things system of claim 15 , wherein a reward function in the reinforcement learning model is 0 when the metro departure interval is smaller than a preset time threshold.

17. The Internet of Things system of claim 15 , wherein

the reinforcement learning model includes an environment module and an optimal action determination module, wherein;

an input of the environment module includes environmental state information, an output of the environmental module includes an optional action set, the environment module includes an optional action determination sub-module, a state determination sub-module, and a reward determination sub-module, wherein

the optional action determination sub-module is used to determine the optional action set based on the environmental state information, the state determination sub-module is used to determine the environmental state information, and the reward determination sub-module is used to determine the reward function, wherein the optional action set refers to a set of actions that are executed by the management platform, and the set of actions that are executed by the management platform include the metro departure interval;

the optimal action determination module is a graph neural networks model, wherein

an input of the optimal action determination module further includes the environmental state information and the optional action set, an output of the optimal action determination module includes an optimal optional action;

the optimal action determination module is obtained through a third training using a reinforcement learning manner based on a plurality of third training samples with third labels, wherein the reinforcement learning manner includes Q-Learning, state-action-reward-state-action (Sarsa), deep Q-network (DQN), Policy-Gradients, Actor-Critic, the third training samples include historical passenger flow data of the target station, the third labels include an optimal departure interval corresponding to historical environmental state information, wherein the third training samples are obtained based on historical data, and the third labels are obtained through the reinforcement learning manner.

18. The Internet of Things system of claim 11 , wherein

an input of the passenger flow distribution prediction model further includes the collection time information and the target period of time information of the target station, an output of the passenger flow distribution prediction model includes the predicted passenger flow distribution of the target station in the target period of time;

the passenger flow distribution prediction model is obtained through a second training based on a plurality of second training samples with second labels, wherein the second training samples include passenger flow data of a fifth sample metro station in a fifth sample period of time and a sixth sample period of time, the second labels include a passenger flow distribution of the fifth sample metro station in the sixth sample period of time, wherein the fifth sample metro station is a same station as the target station, the fifth sample period of time is earlier than the sixth sample period of time, and the second labels are obtained by the management platform by querying or detecting an actual passenger flow distribution of the fifth sample metro station in the sixth sample period of time;

the second training includes: inputting the plurality of second training samples with second labels into an initial passenger flow distribution prediction model, constructing a second loss function based on the second labels and output results of the initial passenger flow distribution prediction model updating parameters of the initial passenger flow distribution prediction model based on the second loss function; and obtaining the passenger flow distribution prediction model until the second loss function of the initial passenger flow prediction distribution model meeting a second preset condition, wherein the second preset condition includes a convergence of the second loss function, a loss function value being smaller than a preset value, and a number of iterations reaching a second threshold.