ATAS-based power distribution system communication optimization scheduling method

Abstract

The invention discloses a power distribution system communication optimization scheduling method based on ATAS, and relates to the technical field of power systems, and the method comprises the following steps: constructing a power distribution area electric equipment information data training set; constructing an ATAS-based intelligent communication optimization scheduling model of the power distribution system; based on the intelligent communication optimization scheduling model of the power distribution system, constructing an intelligent agent to execute scheduling operation, and optimizing the intelligent agent according to a preset multi-agent reward mechanism; according to the power distribution area electric equipment information data training set, in combination with a preset loss function, training the intelligent communication optimization scheduling model of the power distribution system to obtain a trained intelligent communication optimization scheduling model of the power distribution system; and according to the trained intelligent communication optimization scheduling model of the power distribution system, scheduling the input power utilization equipment information of the power distribution area. The method is used for solving the problem that an existing power distribution system scheduling method lacks real-time performance and adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and more specifically, to a communication optimization and dispatching method for a power distribution system based on ATAS. Background Art

[0002] As the demand for electricity continues to grow, the challenges facing the distribution system are also increasing. How to ensure efficient and stable supply of electricity and effectively handle possible faults in the system are the core issues that the distribution dispatching system needs to solve. The communication dispatching method in the distribution system is mainly to ensure efficient collaboration, reliable operation and timely response in all aspects of the distribution network. With the development of smart grids, communication and dispatching technologies are gradually developing towards automation, informatization and intelligence.

[0003] Existing scheduling methods often rely on fixed rules and manual intervention, lacking real-time performance and adaptability.

[0004] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a distribution system communication optimization scheduling method based on ATAS, which constructs an intelligent communication optimization scheduling model for the distribution system, analyzes the system status in real time and makes corresponding adjustments to solve the problem that the existing distribution system scheduling method lacks real-time and adaptability.

[0006] To achieve the above object, the present invention provides the following technical solutions: A distribution system communication optimization scheduling method based on ATAS includes the following steps: constructing a training set of distribution substation power equipment information data; constructing a distribution system intelligent communication optimization scheduling model based on ATAS; constructing an intelligent agent to perform scheduling operations based on the distribution system intelligent communication optimization scheduling model, and optimizing the intelligent agent according to a preset multi-agent reward mechanism; training the distribution system intelligent communication optimization scheduling model according to the distribution substation power equipment information data training set and a preset loss function to obtain a trained distribution system intelligent communication optimization scheduling model; and scheduling the input distribution substation power equipment information according to the trained distribution system intelligent communication optimization scheduling model.

[0007] In a preferred embodiment, the distribution substation area power equipment information data training set includes three sub-training sets; the construction of the distribution substation area power equipment information data training set is specifically: obtaining the power equipment information data set, and marking the power equipment information data set by a preset power equipment location label information set; dividing the power equipment information data set with location labels into three sub-data sets according to location coordinates; and randomly shuffling the order of the power equipment information sub-data sets with location labels as the distribution substation area power equipment information data sub-training set.

[0008] In a preferred embodiment, the ATAS-based distribution system intelligent communication optimization scheduling model includes three sub-agents; the three sub-agents are responsible for the scheduling of three sub-training sets respectively.

[0009] In a preferred embodiment, the intelligent agent is constructed based on the intelligent communication optimization scheduling model of the distribution system to perform scheduling operations, and the intelligent agent is optimized according to the preset multi-agent reward mechanism, specifically: the time for the sub-agent to schedule the corresponding sub-training set is defined as a scheduling cycle, and a scheduling cycle includes multiple scheduling intervals; after each scheduling cycle, the sub-training set data is updated and the sub-data set content is redivided; after the sub-data set content is updated, the intelligent agent performs a new round of scheduling cycle work; within each scheduling interval, the sub-agent performs scheduling operations, and the sub-agents share information and collaborate to optimize; define a state set, and take all network topologies and resource usage states as system states; define an action set, and take all resource scheduling decisions as the next optional scheduling operation of the intelligent agent; after the scheduling operation is executed, according to the system feedback results, based on the preset reward function, obtain the action reward corresponding to the scheduling operation.

[0010] In a preferred embodiment, the training of the ATAS-based intelligent communication optimization scheduling model is specifically as follows: obtaining simulated rewards through a preset reward simulation reference module; when the gap between the action reward and the simulated reward after executing the scheduling operation exceeds the preset simulation threshold, adjusting the scheduling operation; the intelligent agent includes the selection of a first preset probability and the selection of a second preset probability; the selection of the first preset probability is to select a scheduling operation with the second simulated reward value to execute; the selection of the second preset probability is to randomly select a scheduling operation to execute; this round of scheduling operation execution does not participate in the simulation reward comparison link, and the sum of the first preset probability and the second preset probability is one; gradually reduce the first preset probability, and when the first preset probability of any one of the three sub-agents reaches the minimum possible value, end this round of scheduling cycle, perform the sub-training set data update operation, and re-divide the sub-data set.

[0011] In a preferred embodiment, the electrical equipment information data includes equipment type, equipment technical parameters and equipment operating status information; the preset electrical equipment location label information set is specifically obtained by: the location of each device is described by a three-dimensional Cartesian coordinate system; the origin coordinates in the three-dimensional Cartesian coordinate system are the location of the distribution station gateway dispatcher; and each element is constructed as a set of three-dimensional coordinates of each device location.

[0012] In a preferred embodiment, the preset multi-agent reward mechanism is specifically as follows: three agents obtain individual rewards based on the global reward index and their own local performance, and the local performance includes the data processing time in the local area.

[0013] The technical effects and advantages of the ATAS-based distribution system communication optimization scheduling method of the present invention are as follows: 1. The present invention constructs an intelligent communication optimization scheduling model for the distribution system and uses a data-driven approach to analyze the system status in real time and make corresponding adjustments, thereby significantly improving the scheduling efficiency.

[0014] 2. The present invention adopts a multi-agent system so that each subsystem can learn and optimize independently, which improves the overall performance of the system and enhances the fault tolerance of the system. In addition, the multi-agent reward mechanism promotes the collaboration between agents and further optimizes the effect of scheduling decisions. In summary, the present invention has significant advantages in improving the scheduling efficiency, flexibility and stability of the distribution system, and can effectively meet the needs of modern power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flow chart of a communication optimization and scheduling method for a power distribution system based on ATAS.

[0016] Figure 2 Schematic diagram of dividing the electric equipment information sub-dataset with location tags according to the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1, Figure 1 The present invention provides a distribution system communication optimization scheduling method based on ATAS, comprising the following steps: Construct a training set of information data of power equipment in distribution substation area; Construct an intelligent communication optimization and dispatching model for distribution system based on ATAS; Based on the intelligent communication optimization dispatching model of the power distribution system, an intelligent agent is built to perform dispatching operations, and the intelligent agent is optimized according to the preset multi-agent reward mechanism; According to the training set of information data of power consumption equipment in the distribution substation area, combined with the preset loss function, the distribution system intelligent communication optimization scheduling model is trained to obtain the trained distribution system intelligent communication optimization scheduling model; According to the trained distribution system intelligent communication optimization scheduling model, the input distribution substation area power equipment information is scheduled.

[0019] In this embodiment, the training set of information data of power equipment in the distribution station area includes three sub-training sets; The construction of the training set of information data of power equipment in the distribution station area is specifically as follows: Acquire an electric device information data set, and mark the electric device information data set using a preset electric device location tag information set; The electric equipment information data set with location tags is divided into three sub-data sets according to the location coordinates; The power equipment information sub-dataset with location labels is randomly shuffled and used as the power equipment information sub-data training set of the distribution station area.

[0020] In this embodiment, the ATAS-based distribution system intelligent communication optimization scheduling model includes three sub-agents; The three sub-agents are respectively responsible for the scheduling of the three sub-training sets.

[0021] In this embodiment, the intelligent agent is constructed based on the distribution system intelligent communication optimization scheduling model to perform scheduling operations, and the intelligent agent is optimized according to the preset multi-agent reward mechanism, specifically: Define the time for sub-agent scheduling corresponding to the sub-training set as a scheduling cycle, and a scheduling cycle contains multiple scheduling intervals; After each scheduling cycle, the sub-training set data is updated and the sub-data set content is re-divided; After the sub-dataset content is updated, the agent performs a new round of scheduling cycle work; In each scheduling interval, the sub-agents perform scheduling operations, and the sub-agents share information and collaborate on optimization; Define a state set and take all network topology and resource usage states as system states; Define a set of actions and take all resource scheduling decisions as the next optional scheduling operation of the agent; After the scheduling operation is executed, the action reward of the corresponding scheduling operation is obtained based on the system feedback result and the preset reward function.

[0022] In this embodiment, the preset reward function is defined as:

[0023] In the formula, They are the reward functions corresponding to the three sub-agents, i represents the number of actions executed by the system, N represents the total number of times, represents the overall effective throughput evaluation function of the system, represents the sub-flow throughput evaluation function, c represents the number of sub-channels, represents the action delay evaluation function, represents the action efficiency evaluation function, Represents the indicator function, corresponding to three sub-training sets.

[0024] It should be noted that the agent's choice of the next action depends on the previously selected action and its corresponding reward.

[0025] In this embodiment, the training of the ATAS-based intelligent communication optimization scheduling model is specifically as follows: Get simulated rewards through the preset reward simulation reference module; When the difference between the action reward and the simulation reward after the scheduling operation is executed exceeds the preset simulation threshold, the scheduling operation is adjusted; The agent includes a choice of a first preset probability and a choice of a second preset probability; The selection of the first preset probability is to select the scheduling operation with the second simulation reward value for execution; The second preset probability is selected to randomly select a scheduling operation to execute; This round of scheduling operation execution does not participate in the simulated reward comparison link, and the sum of the first preset probability and the second preset probability is one; The first preset probability is gradually reduced. When the first preset probability of any of the three sub-agents reaches the minimum possible value, this round of scheduling cycle is ended, the sub-training set data update operation is performed, and the sub-data set is re-divided.

[0026] In this embodiment, the preset reward simulation reference module is defined as:

[0027] In the formula, Simulate the value of system rewards. , represents the reward simulation function, Indicates the action simulation delay, Indicates the efficiency of motion simulation, Indicates the action simulation indication value, Indicates the current system status; The specific formula of the preset simulation threshold is:

[0028] In the formula, express , Any function in represents the upper threshold, represents the lower threshold, represents the 2-norm, and N represents the total number of times.

[0029] In this embodiment, the electrical equipment information data includes equipment type, equipment technical parameters and equipment operation status information; The specific method for obtaining the preset electrical equipment location tag information set is as follows: The location of each device is described by a three-dimensional Cartesian coordinate system; The origin coordinates in the three-dimensional Cartesian coordinate system are the location of the distribution station area gateway dispatcher; Construct a set in which each element is the three-dimensional coordinate of each device position.

[0030] In this embodiment, the preset multi-agent reward mechanism is specifically as follows: The three agents receive individual rewards based on the global reward index and their own local performance, which includes the data processing time in the local area.

[0031] In this embodiment, the preset loss function is defined as:

[0032] The gradient of the preset loss function is defined as:

[0033] In the formula, is the loss weight, is the model stability evaluation function, n is the number of sub-agent updates, Simulate the value of system rewards. is the action reward function.

[0034] This embodiment constructs an intelligent communication optimization and dispatching model for the distribution system, and uses a data-driven approach to analyze the system status in real time and make corresponding adjustments, thereby significantly improving the dispatching efficiency. This embodiment uses a multi-agent system so that each subsystem can learn and optimize independently, which improves the overall performance of the system while enhancing the fault tolerance of the system. In addition, the multi-agent reward mechanism promotes collaboration between agents and further optimizes the effect of dispatching decisions. In summary, the present invention has significant advantages in improving the dispatching efficiency, flexibility and stability of the distribution system, and can effectively meet the needs of modern power systems.

[0035] Embodiment 2 comprises the following steps: S1, construct a training set of information data of power equipment in the distribution area; S2, build an intelligent communication optimization dispatch model for distribution system based on ATAS; S3, construct an intelligent agent to perform scheduling operations based on the intelligent communication optimization scheduling model of the distribution system, and optimize the intelligent agent according to the preset multi-agent reward mechanism; S4, according to the training set of information data of power consumption equipment in the distribution area, combined with the preset loss function, the distribution system intelligent communication optimization scheduling model is trained to obtain the trained distribution system intelligent communication optimization scheduling model; S5, according to the trained distribution system intelligent communication optimization scheduling model, the input distribution substation area power equipment information is scheduled.

[0036] S1, construct a training set of information data of power equipment in distribution station area.

[0037] Obtain the power consumption equipment information data set, and mark the power consumption equipment information data set with the preset power consumption equipment location label information set, recorded as ,and , represents the data information of the kth electrical equipment collected, and , Indicates that the kth electrical equipment is in A collection of data information at each sampling moment; , K represents the total number of sampling points of electrical equipment information, , R represents the total sampling time; the electrical equipment information data includes equipment type, equipment technical parameters and equipment operating status information.

[0038] Construct the electrical equipment location label information set L, denoted as ,in represents the coordinate position of the kth electrical device. The position of each device is described by a three-dimensional Cartesian coordinate system. The origin coordinate (0, 0, 0) in the three-dimensional Cartesian coordinate system is the position of the gateway dispatcher in the substation. is the distance from the sampling point to the scheduler.

[0039] The electrical equipment information dataset with location tags According to the location coordinates, it is divided into three sub-datasets, denoted as , the classification method is as follows: The coordinate direction of the electrical equipment is ,and ,remember ; Subset ; Subset ); Subset ; The electrical equipment information dataset with location tags The order is randomly shuffled and used as sub-training sets TR1, TR2, and TR3.

[0040] S2, build an intelligent communication optimization and dispatching model for distribution system based on ATAS.

[0041] The ATAS-based distribution system intelligent communication optimization scheduling model consists of three sub-agents: They are responsible for the scheduling of sub-training sets TR1, TR2, and TR3 respectively.

[0042] S3, based on the intelligent communication optimization scheduling model of the distribution system, builds an intelligent agent to perform scheduling operations, and optimizes the intelligent agent according to the preset multi-agent reward mechanism.

[0043] Define the time for sub-agent scheduling corresponding to the sub-training set as a scheduling cycle SC, and a scheduling cycle contains multiple scheduling intervals .

[0044] Scheduling cycle SC: After each scheduling cycle, the sub-training set data is updated and the sub-data set content is re-divided. The sub-data set content is divided in the following way:

[0045] Where n represents the number of scheduling cycles, Indicates the nth scheduling cycle Numeric value.

[0046] After the sub-dataset content is updated, the agent will start a new round of scheduling cycle.

[0047] Scheduling interval SI: In each scheduling interval, the sub-agent performs scheduling operations, and each sub-agent performs information sharing and collaborative optimization strategies.

[0048] Define the state set Z: , taking all network topology and resource usage status as system status, is the system state at time step t; Define action set B: , taking all resource scheduling decisions as the next optional scheduling actions of the agent.

[0049] The agent's choice of the next action depends on the previously selected action and its corresponding reward. After the scheduling operation is executed, the action is rewarded based on the system feedback result. The reward function is defined as:

[0050] In the formula, They are the reward functions corresponding to the three sub-agents, i represents the number of actions executed by the system, N represents the total number of times, represents the overall effective throughput evaluation function of the system, represents the sub-flow throughput evaluation function, c represents the number of sub-channels, represents the action delay evaluation function, represents the action efficiency evaluation function, Represents the indicator function, corresponding to the sub-training sets TR1, TR2 and TR3 respectively.

[0051] Multi-agent reward mechanism: The team as a whole shares the reward mechanism. The scheduling results of the three agents jointly affect the global reward index. The three agents receive individual rewards based on this index and their own local performance (including data processing time in this area).

[0052] Define the reward simulation reference module:

[0053] In the formula, Simulate the value of system rewards. , represents the reward simulation function, Indicates the action simulation delay, Indicates the efficiency of motion simulation, Indicates the action simulation indication value, Indicates the current system status.

[0054] When the difference between the action reward and the simulation reward under the scheduling operation exceeds the simulation threshold, the operation is adjusted. The probability of selecting the scheduling action with the second reward value is 1- A scheduling action is randomly selected with probability to execute, and , the execution of this round of actions does not participate in the reward simulation reference comparison link; At the beginning of the scheduling cycle, , gradually decreasing value;

[0055] In the formula, is the number of trained scheduling actions, express The minimum value that can be taken; The simulation threshold:

[0056] In the formula, express , Any function in represents the upper threshold, represents the lower threshold, represents the 2-norm, and N represents the total number of times.

[0057] When any of the three sub-agents When , this round of scheduling cycle SC ends; Update operation, re-dividing the sub-dataset.

[0058] S4, according to the training set of power equipment information data in the distribution station area, combined with the preset loss function, the distribution system intelligent communication optimization scheduling model is trained to obtain the trained distribution system intelligent communication optimization scheduling model.

[0059] S5, according to the trained distribution system intelligent communication optimization scheduling model, the input distribution substation area power equipment information is scheduled.

[0060] Define the loss function:

[0061] Define the gradient of the loss function:

[0062] In the formula, is the loss weight, is the model stability evaluation function, n is the number of sub-agent updates, Simulate the value of system rewards. is the action reward function.

[0063] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0064] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0065] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0066] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0067] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0068] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A communication optimization scheduling method for a power distribution system based on ATAS, characterized in that: The following steps are involved: Construct a training set of information data of power equipment in distribution substation area; Construct an intelligent communication optimization and dispatching model for distribution system based on ATAS; Based on the intelligent communication optimization dispatching model of the power distribution system, an intelligent agent is built to perform dispatching operations, and the intelligent agent is optimized according to the preset multi-agent reward mechanism; According to the training set of information data of power consumption equipment in the distribution substation area, combined with the preset loss function, the distribution system intelligent communication optimization scheduling model is trained to obtain the trained distribution system intelligent communication optimization scheduling model; According to the trained distribution system intelligent communication optimization scheduling model, the input distribution substation area power equipment information is scheduled.

2. According to the ATAS-based distribution system communication optimization scheduling method of claim 1, it is characterized in that: The training set of power distribution area electrical equipment information data includes three sub-training sets; The construction of the training set of information data of power equipment in the distribution station area is specifically as follows: Acquire an electric device information data set, and mark the electric device information data set using a preset electric device location tag information set; The electric equipment information data set with location tags is divided into three sub-data sets according to the location coordinates; The power equipment information sub-dataset with location labels is randomly shuffled and used as the power equipment information sub-data training set of the distribution station area.

3. According to the ATAS-based distribution system communication optimization scheduling method of claim 2, it is characterized in that: The ATAS-based distribution system intelligent communication optimization scheduling model includes three sub-agents; The three sub-agents are respectively responsible for the scheduling of the three sub-training sets.

4. According to the ATAS-based distribution system communication optimization scheduling method of claim 3, it is characterized in that: The intelligent communication optimization scheduling model based on the power distribution system is used to build an intelligent agent to perform scheduling operations, and the intelligent agent is optimized according to the preset multi-agent reward mechanism, specifically: Define the time for sub-agent scheduling corresponding to the sub-training set as a scheduling cycle, and a scheduling cycle contains multiple scheduling intervals; After each scheduling cycle, the sub-training set data is updated and the sub-data set content is re-divided; After the sub-dataset content is updated, the agent performs a new round of scheduling cycle work; In each scheduling interval, the sub-agents perform scheduling operations, and the sub-agents share information and collaborate on optimization; Define a state set and take all network topology and resource usage states as system states; Define a set of actions and take all resource scheduling decisions as the next optional scheduling operation of the agent; After the scheduling operation is executed, the action reward of the corresponding scheduling operation is obtained based on the system feedback result and the preset reward function.

5. According to claim 4, a distribution system communication optimization scheduling method based on ATAS is characterized in that: The training of the intelligent communication optimization dispatching model of the power distribution system is specifically as follows: Get simulated rewards through the preset reward simulation reference module; When the difference between the action reward and the simulation reward after the scheduling operation is executed exceeds the preset simulation threshold, the scheduling operation is adjusted; The agent includes a choice of a first preset probability and a choice of a second preset probability; The selection of the first preset probability is to select the scheduling operation with the second simulation reward value for execution; The second preset probability is selected to randomly select a scheduling operation to execute; This round of scheduling operation execution does not participate in the simulated reward comparison link, and the sum of the first preset probability and the second preset probability is one; The first preset probability is gradually reduced. When the first preset probability of any of the three sub-agents reaches the minimum possible value, this round of scheduling cycle is ended, the sub-training set data update operation is performed, and the sub-data set is re-divided.

6. According to claim 5, a distribution system communication optimization scheduling method based on ATAS is characterized in that: The electrical equipment information data includes equipment type, equipment technical parameters and equipment operation status information; The specific method for obtaining the preset electrical equipment location tag information set is as follows: The location of each device is described by a three-dimensional Cartesian coordinate system; The coordinates of the origin in the three-dimensional Cartesian coordinate system are the location of the distribution station area gateway dispatcher; Construct a set in which each element is the three-dimensional coordinate of each device position.

7. The method for communication optimization and dispatching of a power distribution system based on ATAS according to claim 6, characterized in that: The preset multi-agent reward mechanism is specifically as follows: The three agents receive individual rewards based on the global reward index and their own local performance, which includes the data processing time in the local area.

8. The method for communication optimization and dispatching of a power distribution system based on ATAS according to claim 7, characterized in that: The preset reward function is defined as: In the formula, They are the reward functions corresponding to the three sub-agents, i represents the number of actions executed by the system, N represents the total number of times, represents the overall effective throughput evaluation function of the system, represents the sub-flow throughput evaluation function, c represents the number of sub-channels, represents the action delay evaluation function, represents the action efficiency evaluation function, Represents the indicator function, corresponding to three sub-training sets.

9. The method for communication optimization and dispatching of a power distribution system based on ATAS according to claim 8, characterized in that: The preset reward simulation reference module is defined as: In the formula, Simulate the value of system rewards. , represents the reward simulation function, Indicates the action simulation delay, Indicates the efficiency of motion simulation, Indicates the action simulation indication value, Indicates the current system status; The specific formula of the preset simulation threshold is: In the formula, express , Any function in represents the upper threshold, represents the lower threshold, represents the 2-norm, and N represents the total number of times.

10. The method for communication optimization and dispatching of a power distribution system based on ATAS according to claim 9, characterized in that: The preset loss function is defined as: The gradient of the preset loss function is defined as: In the formula, is the loss weight, is the model stability evaluation function, n is the number of sub-agent updates, Simulate the value of system rewards. is the action reward function.

Images