CN-121973662-A - Charging system and scheduling method thereof

CN121973662ACN 121973662 ACN121973662 ACN 121973662ACN-121973662-A

Abstract

The application discloses a charging system and a dispatching method thereof, and relates to the technical field of charging. The charging system comprises at least two power modules, at least one charging interface, a main controller and a power distribution device, wherein the method comprises the steps of obtaining state data of the charging system, wherein the state data are used for representing the current states of the power modules, the charging interfaces and the power distribution device, generating environment vectors based on the state data, a topology constraint matrix and an effective charging strategy, wherein the topology constraint matrix is obtained based on the switching topology of the power distribution device, inputting the environment vectors into a pre-trained scheduling model to obtain action vectors, and the action vectors are used for indicating the main controller to schedule the power modules. Therefore, by sensing the full state of the charging system, combining hardware topology constraint and fusing a multidimensional optimization target, dynamic and intelligent power distribution is realized, and the efficiency, adaptability and equipment life of the charging system are improved.

Inventors

NI JIANSHU
WEI JIANRONG
LIU QIANG
YUAN QINGMIN
ZHOU QIANG

Assignees

领充新能源科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260304

Claims (10)

1. A method of scheduling a charging system, the charging system comprising at least two power modules, at least one charging interface, a master controller, and a power distribution device, the method comprising: acquiring state data of the charging system, wherein the state data is used for representing the current states of the power modules, the charging interfaces and the power distribution device; generating an environment vector based on the state data, a topology constraint matrix and an effective charging strategy, wherein the topology constraint matrix is obtained based on a switching topology of the power distribution device; and inputting the environment vector into a pre-trained scheduling model to obtain an action vector, wherein the action vector is used for indicating the main controller to schedule each power module.
2. The method of claim 1, wherein the status data includes at least an open-close status of each controllable switch in the power distribution device, an available status of each power module, an accumulated operating time, and a required power of each charging interface.
3. The scheduling method of a charging system according to claim 1, wherein after inputting the environmental vector into a pre-trained scheduling model to obtain an action vector, the method further comprises: determining whether the motion vector meets a physical constraint condition based on the switch topology, wherein the physical constraint condition at least comprises mutual exclusion of a controllable switch called in the power distribution device; And sending the motion vector to the main controller under the condition that the motion vector meets the physical constraint condition.
4. A scheduling method for a charging system according to claim 3, wherein after the master controller executes the motion vector, the method further comprises: updating the environment vector, and determining each preset rewarding factor and each preset punishment factor; Determining a reward signal based on each of the preset reward factors and each of the preset penalty factors and using a preset reward function; And optimizing the trained scheduling model by using an empirical playback mechanism based on the updated environment vector and the reward signal.
5. The method of claim 4, wherein the predetermined reward factors include a total output power reward factor, a demand satisfaction reward factor, and a power module lifetime balance reward factor, and the predetermined penalty factors include a violation constraint penalty factor and an action frequent oscillation penalty factor; The preset reward function is: R_t=α*F_power+β*F_satisfaction+γ*F_balance-δ*P_violation-ε*P_oscillation Wherein R_t is the reward signal, F_power is the total output power reward factor, F_ satisfaction is the demand satisfaction reward factor, F_balance is the module life balance reward factor, P_ violation is the violation constraint penalty factor, P_ oscillation is the action frequent oscillation penalty factor, and alpha, beta, gamma, delta and epsilon are corresponding weight coefficients.
6. The scheduling method of a charging system according to any one of claims 1 to 5, characterized in that before acquiring the state data of the charging system, the method further comprises: Determining whether the charging system meets preset conditions, and acquiring state data of the charging system under the condition that the charging system meets the preset conditions; the preset conditions at least comprise a preset scheduling period or a state change of any charging interface.
7. The scheduling method of a charging system according to any one of claims 1 to 5, wherein the effective charging policy is determined from preset charging policies based on a current scenario.
8. The scheduling method of a charging system according to claim 1, wherein the training step of the scheduling model includes: Acquiring sample data, wherein the sample data comprises sample state data and a sample topology constraint matrix corresponding to at least one analog charging system; generating at least one sample environment vector for any simulated charging system based on sample state data, a sample topology constraint matrix and a simulated effective charging strategy corresponding to the simulated charging system; And training a scheduling model by taking a sample environment vector corresponding to each simulated charging system as input and taking an action vector as output, and optimizing the scheduling model by taking maximization of long-term accumulation rewards as a target to obtain the trained scheduling model.
9. An edge processing apparatus comprising a memory and a processor, the memory having stored thereon a computer program, characterized in that the processor, when executing the computer program, implements the scheduling method of the charging system according to any one of claims 1 to 8.
10. A charging system comprising at least two power modules, at least one charging interface, a master controller, and a power distribution device, wherein the master controller comprises the edge processing apparatus of claim 9; the main controller is used for acquiring the motion vector output by the edge processing equipment and generating a scheduling instruction based on the motion vector; Each power module is used for outputting power based on the scheduling instruction; The power distribution device is used for controlling the opening or closing of the controllable switch based on the scheduling instruction so as to distribute the output power of each power module to each charging interface.

Description

Charging system and scheduling method thereof Technical Field The application relates to the technical field of charging, in particular to a charging system and a scheduling method thereof. Background With the popularization of new energy automobiles, a modern charging pile is designed to meet the requirements of high-power quick charging and cost control, and a shared power pool with a non-full-connection topology is generally adopted. The structure forms complex path dependence and hardware constraint among the power module, the direct current contactor and the charging gun head, and brings serious challenges to real-time power scheduling of the charging system. However, existing scheduling strategies mostly employ static scheduling algorithms based on fixed rules or simple polling, which have significant drawbacks in dealing with the above-mentioned complex hardware constraints. Firstly, the existing policy resource scheduling is stiff, and the occupied hardware path cannot be dynamically avoided, so that a part of idle power modules cannot be called due to connection blockage, and the waste of actual available power is caused. Secondly, the existing strategy has poor dynamic adaptability, and is difficult to effectively respond to dynamic scenes such as random plug of vehicles, real-time fluctuation of battery demands and the like, so that the system operation efficiency is low. In addition, the existing strategies generally only pay attention to the satisfaction of the instant power demand, and lack comprehensive consideration of multi-dimensional targets such as power module load balancing, equipment life extension, differentiated services and the like. Disclosure of Invention The application mainly aims to provide a charging system and a scheduling method thereof, which are used for realizing dynamic and intelligent power distribution by sensing the full state of the charging system, combining hardware topology constraint and fusing a multi-dimensional optimization target, thereby improving the efficiency, adaptability and equipment service life of the charging system. In order to achieve the above object, the present application provides a scheduling method of a charging system, the charging system including at least two power modules, at least one charging interface, a main controller, and a power distribution device, the method including: acquiring state data of the charging system, wherein the state data is used for representing the current states of the power modules, the charging interfaces and the power distribution device; generating an environment vector based on the state data, a topology constraint matrix and an effective charging strategy, wherein the topology constraint matrix is obtained based on a switching topology of the power distribution device; and inputting the environment vector into a pre-trained scheduling model to obtain an action vector, wherein the action vector is used for indicating the main controller to schedule each power module. Optionally, the state data at least includes an open-close state of each controllable switch in the power distribution device, an available state of each power module, an accumulated working time length, and a required power of each charging interface. Optionally, after the environmental vector is input into a pre-trained scheduling model to obtain an action vector, the method further comprises determining whether the action vector meets a physical constraint condition based on the switch topology, wherein the physical constraint condition at least comprises mutual exclusion of a controllable switch called in the power distribution device, and sending the action vector to the main controller when the action vector meets the physical constraint condition. Optionally, after the master controller executes the motion vector, the method further comprises updating the environment vector, determining each preset rewarding factor and each preset punishment factor, determining a rewarding signal based on each preset rewarding factor and each preset punishment factor and by using a preset rewarding function, and optimizing the trained scheduling model by using an empirical playback mechanism based on the updated environment vector and the rewarding signal. Optionally, the preset reward factors include a total output power reward factor, a demand satisfaction reward factor and a power module lifetime balance reward factor, the preset penalty factors include a violation constraint penalty factor and an action frequent oscillation penalty factor, and the preset reward function is: R_t=α*F_power+β*F_satisfaction+γ*F_balance-δ*P_violation-ε*P_oscillation Wherein R_t is the reward signal, F_power is the total output power reward factor, F_ satisfaction is the demand satisfaction reward factor, F_balance is the module life balance reward factor, P_ violation is the violation constraint penalty factor, P_ oscillation is the action frequent oscillation penalty f