Substation operation and maintenance method and device based on multiple agents, equipment and medium

Abstract

Embodiments of the present disclosure disclose a power substation operation and maintenance method and device based on multi-agent, equipment and medium. A specific embodiment of the method comprises: in response to determining that the volume of the power substation is higher than a target volume, obtaining a real-time operation data set and an agent coordination graph; performing graph partitioning on the agent coordination graph to obtain at least one agent group set; for each agent group set, performing a generation step: for each agent group, generating group decision information; displaying the group decision information set on a decision page; generating an agent decision graph; performing agent screening of the coordination agent set to obtain a coordination agent subset and agent decision information; generating operation and maintenance information; and controlling the master agent to instruct each coordination agent to perform equipment operation and maintenance processing on the target power substation according to the operation and maintenance information. This embodiment can achieve accurate generation of corresponding operation and maintenance information of the power substation while reducing the computing resources of the agent, thereby ensuring the normal operation of the power substation.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and specifically to a substation operation and maintenance method, apparatus, equipment, and medium based on multi-agent systems. Background Technology

[0002] Currently, with the continuous development of the power industry, timely maintenance of substations can effectively ensure timely power supply and avoid power outages. The typical approach to substation maintenance is as follows: First, based on reinforcement learning, the various agents corresponding to the substation are instructed to make overall decision-making outputs to obtain the corresponding operation and maintenance information. Then, the maintenance operations corresponding to this information are executed.

[0003] However, when using the above method, the following technical problems often arise:

[0004] When the substation is large, the number of intelligent agents set up in the substation is large, which leads to a large space for joint actions of intelligent agents and an exponential growth of linkage action controls when making overall decision outputs. This results in a large amount of server resources being used up, leading to the exhaustion of memory and computing resources.

[0005] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure propose a substation operation and maintenance method, apparatus, equipment, and medium based on multi-agent systems to solve one or more of the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide a substation operation and maintenance method based on multiple agents, including: in response to determining that the target substation's corresponding substation size is higher than the target size, acquiring a real-time operation dataset and an agent coordination graph corresponding to the target substation, wherein the agent coordination graph corresponds to: a master agent and a set of coordinating agents; using the master agent, partitioning the agent coordination graph to obtain at least one agent group, wherein each agent group is partitioned based on target partitioning features, and the agents in the agent group make collaborative decisions; for each agent group, performing the following generation steps: for each agent group, generating the aforementioned... Group decision information corresponding to the intelligent agent group; displaying the group decision information set on the decision page corresponding to the main intelligent agent; generating an intelligent agent decision graph corresponding to the intelligent agent group based on the candidate decision information determined on the decision page, the intelligent agent decision graph supporting various visualization operations; performing intelligent agent filtering of the coordinated intelligent agent set based on at least one intelligent agent decision graph to obtain intelligent agent decision information corresponding to the coordinated intelligent agent subset and the remaining coordinated intelligent agent set; generating operation and maintenance information corresponding to the target substation based on the intelligent agent decision information and the coordinated intelligent agent subset; controlling the main intelligent agent to instruct each coordinated intelligent agent to perform equipment operation and maintenance processing on the target substation based on the operation and maintenance information.

[0009] Secondly, some embodiments of this disclosure provide a substation operation and maintenance device based on a multi-agent system, comprising: an acquisition unit configured to, in response to determining that the target substation's size is higher than the target size, acquire a real-time operation dataset and an agent coordination graph corresponding to the target substation, wherein the agent coordination graph corresponds to: a master agent and a set of coordinating agents; a partitioning unit configured to, using the master agent, partition the agent coordination graph to obtain at least one agent group, wherein each agent group is partitioned based on a target partitioning feature, and the agents in the agent group make collaborative decisions; and an execution unit configured to, for each agent group, execute the following generation steps: for each agent group, generate the aforementioned... The system includes: group decision information corresponding to the intelligent agent group; displaying the group decision information set on the decision page corresponding to the main intelligent agent; generating an intelligent agent decision graph corresponding to the intelligent agent group based on the candidate decision information determined on the decision page, wherein the intelligent agent decision graph supports various visualization operations; a filtering unit configured to perform intelligent agent filtering of the coordinated intelligent agent group based on at least one intelligent agent decision graph, obtaining intelligent agent decision information corresponding to the coordinated intelligent agent subset and the remaining coordinated intelligent agent group; a generation unit configured to generate operation and maintenance information corresponding to the target substation based on the intelligent agent decision information and the coordinated intelligent agent subset; and a control unit configured to control the main intelligent agent to instruct each coordinated intelligent agent to perform equipment operation and maintenance processing on the target substation based on the operation and maintenance information.

[0010] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, such that when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any implementation of the first aspect.

[0011] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method as described in any implementation of the first aspect.

[0012] The above embodiments of this disclosure have the following beneficial effects: Through the multi-agent-based substation operation and maintenance method of some embodiments of this disclosure, accurate generation of substation operation and maintenance information can be achieved while reducing the computing resources of the agents, ensuring the normal operation of the substation. Specifically, the reason for the exhaustion of computing resources during the generation of relevant operation and maintenance information is that: when the substation is large, the number of agents set for the substation is large, resulting in a large space for joint actions of the agents and an exponential increase in the linkage action controls, leading to a large amount of server resources being used, resulting in the exhaustion of memory and computing resources. Based on this, the multi-agent-based substation operation and maintenance method of some embodiments of this disclosure firstly, in response to determining that the substation size of the target substation is higher than the target size, obtains the real-time operation dataset and agent coordination graph corresponding to the target substation. The agent coordination graph includes: a master agent and a set of coordinating agents. Here, after determining that the substation size is higher than the target size, by setting up an agent coordination graph, the coordination relationship between various substations can be displayed, so that when performing graph partitioning later, closely related agents can be grouped together for collaborative decision-making as much as possible. The master agent here is responsible for global decision-making and core function integration. The coordinating agent is responsible for executing related tasks among the various agents in the complex subsystem. By setting up the master agent and the coordinating agent set, efficient execution of various events corresponding to the target substation can be achieved. Then, using the aforementioned master agent, the aforementioned agent coordination graph is partitioned to obtain at least one agent group. Each agent group is partitioned based on target partitioning features, and the agents in each agent group make collaborative decisions. Here, by using graph partitioning, the agent groups that make collaborative decisions under different target requirements (i.e., target partitioning features) can be determined. By partitioning based on different target partitioning features, it is not necessary to make overall decisions for all agents; only local decisions of the coordinating agents under the corresponding partitioning features are needed. This can greatly reduce the number of agents involved in overall decision-making, avoid the occurrence of large joint action spaces for agents, large amounts of server resources, and exhaustion of memory and computing resources. Next, for each agent group, the following generation steps are performed: First, for each agent group, based on the corresponding real-time running data set, accurate group decision information can be generated. Second, the group decision information set is displayed on the decision page corresponding to the main agent, allowing the operation and maintenance object to confirm the group decision information on the decision page, facilitating subsequent generation of overall decisions under the target segmentation features. Third, based on the candidate decision information determined on the decision page, an agent decision graph corresponding to the agent group is generated, supporting various visualization operations.Here, by generating an agent decision graph, the operation and maintenance object can clearly understand the coordination relationships and decision-making status among various agents, enabling the operation and maintenance object to adaptively adjust the individual intelligent decisions in the agent decision graph and enhance the accuracy of the information in the agent decision graph. Next, based on at least one agent decision graph, agent filtering of the aforementioned coordinated agent set is performed to reduce the number of decisions made by agents, avoiding further waste of server resources, and accurately obtaining the agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set. Furthermore, based on the aforementioned agent decision information and the aforementioned coordinated agent subset, the operation and maintenance information corresponding to the aforementioned target substation is accurately generated with minimal waste of computing resources. Finally, based on the aforementioned operation and maintenance information, the master agent is controlled to instruct the various coordinated agents to perform equipment operation and maintenance processing on the aforementioned target substation, ensuring the normal operation of the target substation and avoiding fault problems. In summary, by performing diverse graph partitioning of the agent coordination graph according to target segmentation characteristics, it is possible to initially and accurately identify agents that do not require subsequent decision-making and to determine the individual decision information of the agents that do make decisions. Based on this, operation and maintenance information can be accurately generated with fewer intelligent agents and fewer joint action controls to maintain the normal operation of substations. Attached Figure Description

[0013] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0014] Figure 1 This is a flowchart of some embodiments of the multi-agent-based substation operation and maintenance method according to this disclosure;

[0015] Figure 2 This is a structural schematic diagram of some embodiments of the substation operation and maintenance device based on multi-agent systems according to the present disclosure;

[0016] Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0017] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0018] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0019] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0020] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0021] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0022] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] refer to Figure 1 The diagram illustrates a flow 100 of some embodiments of a multi-agent-based substation operation and maintenance method according to the present disclosure. This multi-agent-based substation operation and maintenance method includes the following steps:

[0024] Step 101: In response to determining that the target substation's size is greater than the target size, obtain the real-time operation dataset and agent coordination diagram corresponding to the target substation.

[0025] In some embodiments, in response to determining that the size of the corresponding substation is greater than the target substation size, the executing entity (e.g., electronic equipment) of the above-mentioned multi-agent-based substation operation and maintenance method can obtain the real-time operation dataset and agent coordination diagram corresponding to the target substation. The target substation can be the substation to be maintained. Substation size refers to its physical scale, carrying capacity, investment intensity, and number of agents. Substation size can be measured by five dimensions: land area, equipment capacity, investment amount, spatial layout, and number of agents. Substation sizes vary significantly depending on voltage level and design type (above-ground / underground / box-type). For example, substation size can be determined by a preset association table. The association table can characterize the relationship between the number of agents, agent complexity, physical scale, carrying capacity, and substation size information. Substation size can be a value between 1 and 100. The more agents, the more complex the agent integration, the larger the physical scale, and the stronger the carrying capacity, the higher the corresponding substation size value. The target size can be a set value to measure the size of the substation. In other words, for target substations with a size larger than the target size, a selection decision needs to be made regarding the intelligent agents. For target substations with a size no larger than the target size, no selection decision is required. The real-time operational dataset can be the operational data of each power device in the target substation at the current time. The intelligent agent coordination graph can be a graph representing the coordination relationships between intelligent agents. Coordination relationships can be relationships between intelligent agents that involve mutual coordination or collaborative control during task execution (e.g., calling each other's data). For example, in the "smart customer service" scenario for the target intelligent platform, three intelligent agents—intent understanding, information extraction, and knowledge base retrieval—work in parallel, with the coordinating intelligent agent integrating the results to achieve a second-level response. There is a relationship of mutual coordination between the three intelligent agents. Nodes in the intelligent agent coordination graph are intelligent agent nodes, and the corresponding edges represent the coordination relationships between intelligent agents. Coordination relationships can include, but are not limited to, at least one of the following: collaborative execution relationship, collaborative control relationship, parallel execution relationship, sequential scheduling relationship, and master-slave control relationship. The node content corresponding to an intelligent agent node can be the identifier of the intelligent agent. The edge content indicates the specific type of coordination relationship. Here, within the target substation, the primary intelligent agent and the coordinating intelligent agent achieve autonomous decision-making and efficient operation through hierarchical collaboration. The primary intelligent agent acts as the "brain" of the substation, responsible for global data integration, advanced analysis, and decision-making. The coordinating intelligent agent is responsible for command distribution, multi-device collaboration, emergency response, and the execution of localized services. The primary intelligent agent aligns with the overall control and decision-making of various services within each area of ​​the target substation. The coordinating intelligent agent is responsible for the control and decision-making of localized services within a specific area. The agent coordination diagram can be generated based on the tasks performed and the data invoked by each intelligent agent. The agent coordination diagram can be maintained periodically.

[0026] Step 102: Using the aforementioned main agent, perform graph partitioning on the aforementioned agent coordination graph to obtain at least one agent set.

[0027] In some embodiments, the aforementioned executing entity can utilize the aforementioned master agent to partition the agent coordination graph, obtaining at least one agent set. Each agent set is partitioned based on target partitioning features, and the agents within each agent set perform collaborative decision-making. Target partitioning features can be feature objects used to partition the agent coordination graph. For example, target partitioning features can be deployment area location features. That is, by partitioning the agent coordination graph according to the deployment area location of the processing area corresponding to the agent, agent sets under each deployment area can be obtained. Another example is that target partitioning features can be fault features. That is, by partitioning the agent coordination graph according to the correlation between fault detection of the power equipment corresponding to the agent, agent sets under each fault collaborative detection can be obtained. Here, the agents within the agent set perform fault detection together (i.e., fault collaborative detection). Yet another example is that target partitioning features can be economic benefit features. That is, by partitioning the agent coordination graph according to the economic benefits of the processing area corresponding to the agent, agent sets under each deployment area can be obtained. In practice, target partitioning features can be pre-set or set on the decision page. Each agent group has corresponding target partitioning features. Within each agent group, there is a master agent. Because of the coordination relationship between the master agent and the coordinating agents, each agent group has a master agent to instruct the agents in the group to make collaborative decisions. This collaborative decision-making can be based on reinforcement learning and mean-field theory, where the agents in the group work together to make action decisions in the current environment, aiming to output a locally optimal decision for the corresponding region or environment.

[0028] As an example, the aforementioned executing entity can issue a graph partitioning instruction to the master agent, so that the master agent can partition the aforementioned agent coordination graph based on the feature performance of each coordinating agent under the target partitioning features, thereby obtaining at least one agent set. The graph partitioning instruction can be an instruction for partitioning the agent coordination graph.

[0029] In some optional implementations of certain embodiments, after step 102, the steps further include:

[0030] The first step involves responding to the agent processing operations performed by the maintenance object on the corresponding agent coordination graph on the decision page. Based on these operations, adjustments are made to each agent node and edge in the coordination graph to obtain the adjusted coordination graph. The decision page supports clicking on each agent node in the coordination graph. After clicking, a pop-up window appears, displaying at least one of the following: associated device, associated agent node, environmental status, execution action set, and confirmation information regarding participation in maintenance. The maintenance object can be the entity making maintenance decisions for the target substation. The agent processing operation can be the operation performed by the maintenance object on the decision page after adjusting the coordination graph. The associated device can be any power equipment associated with the agent node. The associated agent node can be at least one agent node with a coordination relationship to the agent node. The environmental status can be the current environmental condition of the agent node. The execution action set can be the complete set of action decisions that the agent node can execute. The confirmation information regarding participation in maintenance indicates the agent's participation in the generation of subsequent maintenance information. The confirmation message indicating non-participation in operations and maintenance (O&M) signifies that the agent will not participate in the generation of subsequent O&M information. The content of the pop-up window supports dynamic adjustment of the O&M object. Changes to the pop-up window's content may also synchronously change the relevant content of the corresponding agent node in the agent coordination graph. For example, if the agent confirms non-participation in subsequent O&M in the pop-up window, the corresponding agent node in the agent coordination graph will be set to a gray node, indicating non-participation in future O&M.

[0031] The second step is to determine the adjusted agent coordination diagram as the above agent coordination diagram, wherein each agent node in the above agent coordination diagram is a node that has been confirmed to participate in operation and maintenance.

[0032] In some optional implementations of certain embodiments, the aforementioned executing entity may utilize the aforementioned master agent to partition the aforementioned agent coordination graph to obtain at least one agent set, including the following steps:

[0033] The first step is to obtain the operation and maintenance (O&M) direction information corresponding to the target substation entered on the decision-making page. This O&M direction information can represent textual information indicating the O&M direction for the target substation. For example, the O&M direction information could be the desired outcome of reducing the failure rate and improving economic efficiency of the target substation without affecting normal power supply. This O&M direction information can be entered by the O&M entity in the corresponding input field on the decision-making page.

[0034] The second step involves decomposing the aforementioned operational direction information into features, yielding at least one target segmentation feature and at least one corresponding feature segmentation direction. Each target segmentation feature has a corresponding feature segmentation direction. The feature segmentation direction can be the desired development direction of the corresponding feature content under the target segmentation feature. For example, for a target segmentation feature of failure rate, the corresponding feature segmentation direction could be the direction of reducing the overall failure rate. For a target segmentation feature of economic efficiency, the corresponding feature segmentation direction could be the direction of improving overall economic efficiency.

[0035] As an example, the aforementioned execution entity can extract keywords from the operation and maintenance direction information to obtain at least one target partitioning feature and at least one corresponding feature partitioning direction.

[0036] The third step is to perform the following partitioning steps for each target feature:

[0037] Sub-step 1: Determine the target feature division direction corresponding to the above target division features.

[0038] Sub-step 2 involves using the aforementioned main agent to partition the agent coordination graph according to the target partitioning features, resulting in an initial agent sub-graph set. In this initial sub-graph set, the agents within each agent exhibit similar behavior under the target partitioning features. For example, if the target partitioning feature is location West Sydney, the agents in the corresponding initial sub-graph are adjacent agents operating within the same module function.

[0039] As an example, the aforementioned executing entity can determine the performance of each agent in the agent coordination graph under the target partitioning features. Then, based on the performance, agents with similar performance are grouped into one category to partition the agent coordination graph, resulting in an initial agent subgraph set.

[0040] Sub-step 3: Using the aforementioned main agent, determine the initial decision adjustability of each agent in the initial agent sub-graph set under the target feature partitioning direction. Here, the initial decision adjustability information can be adjustment information determined by the main agent regarding whether the decisions corresponding to each agent are adjustable based on the target feature partitioning direction. Each agent has corresponding decision adjustability to be further determined. In practice, the initial decision adjustability can be one of the following: adjustment information representing adjustable decisions, or adjustment information representing unadjustable decisions.

[0041] As an example, the main agent can determine the adjustability of the initial decision of each agent in the above initial agent sub-graphet under the target feature division direction based on historical adjustment experience.

[0042] Sub-step 4 involves displaying the obtained initial agent sub-graph set on the page redirected by the graph partitioning control on the decision page, in order to confirm and adjust the adjustability of each partitioning graph under the target partitioning features and the corresponding initial decisions.

[0043] Sub-step 5: Obtain the adjusted agent subgraph and corresponding decision adjustability set based on the initial agent subgraph. There is a one-to-one correspondence between the decision adjustability in the decision adjustability set and the agents in the agent subgraph. There is also a one-to-one correspondence between the decision adjustability groups in the decision adjustability set and the decision adjustability groups in the decision adjustability set. The decision adjustability set and agent subgraph can be determined by the decision object on the decision page.

[0044] Sub-step 6: Determine the intelligent agent set corresponding to the above-mentioned target partitioning feature by identifying the intelligent agent set corresponding to the above-mentioned intelligent agent sub-graph set.

[0045] Step 103: For each set of agents, perform the following generation steps:

[0046] Step 1031: For each intelligent agent group, generate group decision information corresponding to the intelligent agent group based on the real-time running data group corresponding to the intelligent agent group.

[0047] In some embodiments, the execution entity can generate group decision information for each agent group based on the real-time operation data group corresponding to the agent group. The real-time operation data group can be a subset of the real-time operation data of each power device corresponding to the agent group. For example, the agent group includes: a transformer fault detection agent, a transformer-associated device fault detection agent A, and a transformer-associated device fault detection agent B. The power device corresponding to the transformer fault detection agent is a transformer. The power device corresponding to transformer-associated device fault detection agent A is transformer-associated device A. The power device corresponding to transformer-associated device fault detection agent B is transformer-associated device B. The real-time operation data group can include: real-time operation data corresponding to the transformer, real-time operation data corresponding to transformer-associated device A, and real-time operation data corresponding to transformer-associated device B. The group decision information can be the decision result after coordinated decision-making by each agent in the agent group. The decision result can be the optimal action to be executed by each agent in the agent group under the current environment. By having each agent execute its corresponding optimal action, the optimal feature effect can be achieved under the target segmentation features. In practice, the reward and penalty functions corresponding to a group of intelligent agents can be set based on the contribution of the real-time running data group to the target segmentation features. That is, different contribution levels can correspond to different reward and penalty functions. For example, group decision information can be the decision action on how to maintain a transformer, transformer-related equipment A, and transformer-related equipment B.

[0048] As an example, the aforementioned executing entity can use the real-time running data group as the dataset in the reinforcement learning process. Through mean field theory, it can achieve coordinated decision-making among the agents in the agent group and obtain the optimal decision actions of each agent under the target partitioning features, which serve as the group decision information.

[0049] In some optional implementations of certain embodiments, generating group decision information corresponding to the intelligent agent group based on the real-time running data group corresponding to the intelligent agent group includes:

[0050] The first step is to set the reward and penalty functions for each intelligent agent group based on the feature partitioning directions. Each feature partitioning direction has a unique corresponding reward and penalty function. In practice, the reward and penalty function can be a reward function based on the task objective. The task objective can be a directional objective that maximizes the feature partitioning direction. The reward and penalty functions can be set by relevant technical experts.

[0051] The second step is to generate group decision information corresponding to the intelligent agent group based on the decision adjustability group and real-time operation data group corresponding to the intelligent agent group, using the reward and punishment function.

[0052] As an example, the aforementioned executing entity can generate group decision information corresponding to the intelligent agent group based on the decision adjustability group and real-time running data group corresponding to the intelligent agent group, using the aforementioned reward and punishment function, through the average field effect and reinforcement learning reward and punishment mechanism.

[0053] Step 1032: Display the group decision information set on the decision page corresponding to the main intelligent agent.

[0054] In some embodiments, the aforementioned executing entity can display the group decision information set on the decision page corresponding to the main intelligent agent. The decision page can display the decision actions required by each intelligent agent during the operation and maintenance process of the target substation. In addition, the decision page can also display information on whether the decision actions are adjustable. Here, by displaying the group decision information set on the decision page, the operation and maintenance object can adaptively adjust the group decision information. There is a one-to-one correspondence between the group decision information in the group decision information set and the intelligent agent groups in the intelligent agent group set.

[0055] Here, after the operation and maintenance object processes the group decision information on the decision page, it will be transmitted to the main intelligent agent so that the main intelligent agent can transmit the confirmed group decision information to the corresponding coordinating intelligent agent in the subsequent process of generating operation and maintenance information.

[0056] Step 1033: Generate an agent decision graph corresponding to the above agent set based on the candidate decision information determined on the above decision page.

[0057] In some embodiments, the aforementioned executing entity can generate an agent decision graph corresponding to the agent set based on the candidate decision information determined on the aforementioned decision page. The candidate decision information can be the result of the operation and maintenance object adjusting the decision information of the agents on the decision page. That is, the decision page displays the decision information corresponding to each agent. The decision information corresponding to each agent can be extracted from the group decision information set. The agent decision graph can be a schematic diagram representing the execution action of each agent in the agent set to maximize the expected feature corresponding to the target partitioning feature in the current environmental state at the current time.

[0058] As an example, the aforementioned implementing entity can add the obtained candidate decision information set to the agent coordination graph to obtain the agent decision graph.

[0059] Step 104: Based on at least one agent decision graph, perform agent screening of the above-mentioned coordinated agent set to obtain agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set.

[0060] In some embodiments, the aforementioned executing entity can perform agent screening of the aforementioned coordinated agent set based on at least one agent decision graph, to obtain agent decision information corresponding to a subset of coordinated agents and the remaining coordinated agent set. Agent screening may involve selecting agents from the coordinated agent set that do not require further decision-making. Each coordinated agent in the subset of coordinated agents may be a coordinated agent whose decision information needs further determination. The remaining coordinated agents may be coordinated agents whose decision information has been determined and will not undergo further changes. Agent decision information may be a set of determined decision information corresponding to each remaining coordinated agent in the remaining coordinated agent set.

[0061] As an example, the aforementioned executing entity can display at least one agent decision graph on the decision page, allowing the operation and maintenance object to filter agents on the decision graph and obtain agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set.

[0062] In some optional implementations of certain embodiments, the execution entity performs agent screening of the coordinated agent set based on at least one agent decision graph to obtain agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set, including the following steps:

[0063] The first step is to determine the feature weight corresponding to each of the at least one target partitioning feature. The feature weight characterizes the importance of the target partitioning feature in the overall decision-making process of all agents. In practice, a higher feature weight indicates that the agent's decision-making is more inclined to align with the target partitioning feature's objective, thereby maximizing the objective. The feature weights can be customized on the feature processing page. The feature processing page supports the definition and weight setting of at least one target partitioning feature.

[0064] The second step is to perform the first processing step for each agent in the above agent coordination graph:

[0065] Sub-step 1 involves determining at least one individual decision information corresponding to the aforementioned agent in the at least one agent decision graph. This individual decision information is extracted from the group decision information. Individual decision information can be the decision information that the agent itself needs to make within the group decision information. The individual decision information in the at least one individual decision information is the agent's own decision information made under the target segmentation features.

[0066] Sub-step 2: Determine at least one feature weight corresponding to at least one individual decision information mentioned above.

[0067] Sub-step 3 involves selecting the two highest-weighted individual decision pieces from the at least one set of individual decision information, based on at least one feature weight, and designating them as the primary decision information and secondary decision information, respectively. The primary decision information can be the individual decision information with the highest corresponding weight among all individual decision information sets. The secondary decision information can be the individual decision information with the second highest corresponding weight among all individual decision information sets. The feature weighting can be achieved by accumulating the feature weights of identical individual decision information sets.

[0068] As an example, firstly, at least one individual decision information is deduplicated to obtain a set of deduplicated individual decision information. Then, for each deduplicated individual decision information, the first step is to determine the feature weight set corresponding to that individual decision information. The feature weights in the feature weight set are then summed to obtain the summed weight. Finally, the individual decision information with the highest summed weight is selected from the set of deduplicated individual decision information as the primary decision information. The individual decision information with the second highest summed weight is selected from the set of deduplicated individual decision information as the secondary decision information.

[0069] The third step is to add the information of each main decision to the agent coordination graph to obtain the first addition graph.

[0070] As an example, the aforementioned executing entity can use the main decision information as information tags to add to the agent coordination graph, thus obtaining the first added graph.

[0071] The fourth step involves using the aforementioned main intelligent agent to perform coordination checks on the various main decision-making information pieces in the first added graph, obtaining the detection information. The coordination check can determine whether there are decision conflicts among the various main decision-making information pieces. The detection information can characterize whether decision conflicts exist.

[0072] As an example, the aforementioned executing entity can perform coordination detection on the various main decision information in the first addition graph based on the lightweight large language model deployed by the main intelligent agent, and obtain detection information.

[0073] Fifth, in response to the aforementioned detection information indicating the existence of at least one conflict decision-making region, the set of coordinating agents corresponding to the at least one conflict decision-making region is determined as at least one coordinating agent. The conflict decision-making region information can be region information of a graph region where decision-making conflicts exist. For example, the conflict decision-making region information can be region identification information. The set of coordinating agents corresponding to at least one conflict decision-making region can be a collection of the subsets of coordinating agents corresponding to each conflict decision-making region.

[0074] Step 6: Remove at least one of the above-mentioned coordinating agents from the set of coordinating agents to obtain the set of the remaining coordinating agents.

[0075] The seventh step is to determine the master decision information set corresponding to the remaining set of coordinating agents as the agent decision information.

[0076] Here, "steps one through seven" (as one of the inventors of this disclosure) addresses another technical problem: "how to screen agents to reduce the waste of decision-making resources when there are many agents and a large amount of state space information in subsequent decision-making processes." Based on this, firstly, this disclosure uses the feature weights corresponding to the agents to screen out the primary and secondary decision-making information for each agent. By determining the primary and secondary decision-making information, subsequent judgments based on the coordination between decision information can determine which agents need to make further decisions and which do not, thereby reducing the number of agents and lowering decision-making computational resources. Furthermore, by determining conflict decision-making region information, coordinating agents with decision-making conflicts can be effectively screened out. Based on this, coordinating agents with decision-making conflicts can make further decisions, ensuring the accuracy of the agent decision graph in overall decision-making.

[0077] Step 105: Based on the above-mentioned intelligent agent decision information and the above-mentioned coordination intelligent agent subset, generate the operation and maintenance information corresponding to the above-mentioned target substation.

[0078] In some embodiments, the aforementioned executing entity can generate operation and maintenance information corresponding to the target substation based on the aforementioned agent decision information and the aforementioned coordination agent subset. The operation and maintenance information can be a set of action executions for each agent corresponding to the target substation to perform subsequent operation and maintenance-related actions.

[0079] As an example, firstly, the aforementioned executing entity can obtain a subset of real-time operational data corresponding to a subset of coordinating agents. Then, using this real-time operational data subset as a dataset for reinforcement learning, and based on mean-field theory, instructing the master agent to make collaborative decisions for each coordinating agent within the subset, thereby obtaining collaborative decision information. Finally, the agent decision information and the collaborative decision information are defined as operational information.

[0080] In some optional implementations of certain embodiments, the aforementioned executing entity can generate the operation and maintenance information corresponding to the target substation based on the aforementioned intelligent agent decision information and the aforementioned coordination intelligent agent subset, including the following steps:

[0081] The first step is to determine the server resources corresponding to each server node in the server cluster corresponding to the above-mentioned agent coordination graph. These server resources can be the computing resources that the server can schedule and use for decision-making.

[0082] The second step is to determine the number of agents and state space information corresponding to each of the at least one decision region information mentioned above. The agent's state space information is the set of all possible states of the agent's environment, used to describe the agent's situation or position at a given moment, and is the basis for the agent's decision-making.

[0083] The third step involves combining or decomposing the decision region information from the at least one set of information based on the server resources, the number of agents corresponding to each decision region, and the state space information. This yields at least one processed decision region. The computational resources required for each agent to make decisions within the processed decision region do not exceed the server resources at the target quantile. The server resources at the target quantile can be the resource values ​​corresponding to each server resource at that target quantile. For example, if the target quantile is 1 / 3, the server resources at that target quantile can be the server resource values ​​at 1 / 3.

[0084] As an example, firstly, for each decision region information, the required decision resources are estimated based on the number of agents and state space information corresponding to the decision region information, thus obtaining estimated decision resources. Then, using at least one estimated decision resource and the resources of each server, the decision region information in the above-mentioned at least one decision region information is combined or decomposed to obtain at least one processed decision region information whose required computational resources do not exceed the target quantile. That is, decision region information with higher estimated resources is split into regions, and decision region information with lower estimated resources is combined into regions.

[0085] Fourth, for each processed decision region information, perform the following second processing step:

[0086] Sub-step 1: Determine the target server node corresponding to the decision area information after the above processing.

[0087] As an example, firstly, based on the estimated decision resources corresponding to the processed decision region information, a corresponding server node is allocated to each processed decision region information. The server resource corresponding to each server node divided by the corresponding estimated decision resource is 1.5.

[0088] Sub-step 2 involves expanding the processed decision region information based on the server resources corresponding to the target server node at the current time, resulting in expanded decision region information. The estimated decision resources corresponding to the expanded decision region information are 0.8 times the corresponding service resources. This expansion can be achieved by using the location of the processed decision region information as the center and expanding the decision region information in all directions.

[0089] Sub-step 3: Determine at least one extended coordinating agent and the corresponding remaining coordinating agent subset corresponding to the extended decision region information.

[0090] Sub-step 4: Determine the decision sub-information in the agent decision information corresponding to the remaining subset of coordinating agents.

[0091] Sub-step 5: Using the target server node and based on the decision sub-information, instruct at least one extended coordinating agent and the remaining subset of coordinating agents to coordinate the output of operation and maintenance information, so as to generate sub-operation and maintenance information corresponding to the processed decision area information.

[0092] As an example, the aforementioned execution entity can utilize the target server node, using the decision sub-information as the basis for decision information, and employing normal field effects and reinforcement learning methods to instruct at least one extended coordinating agent and the remaining subset of coordinating agents to coordinate the output of operation and maintenance information, thereby generating sub-operation and maintenance information corresponding to the processed decision area information.

[0093] The fifth step is to generate maintenance information based on at least one sub-maintenance information and the aforementioned agent decision information.

[0094] As an example, the aforementioned execution entity can combine at least one sub-operation and maintenance information with the aforementioned intelligent agent decision information to obtain operation and maintenance information.

[0095] Optionally, the aforementioned executing entity may utilize the aforementioned target server node, based on the aforementioned decision sub-information, to instruct at least one extended coordinating agent and the aforementioned subset of remaining coordinating agents to coordinate the output of operation and maintenance information, thereby generating sub-operation and maintenance information corresponding to the aforementioned processed decision region information, including:

[0096] The first step is to determine at least one piece of auxiliary decision-making information corresponding to at least one extended coordinating agent. Each extended coordinating agent has a unique set of auxiliary decision-making information.

[0097] The second step involves using the target server node to instruct at least one extended coordinating agent and the remaining subset of coordinating agents to coordinate the output of maintenance information based on the aforementioned decision sub-information and at least one auxiliary decision information, in order to generate sub-maintenance information corresponding to the processed decision area information.

[0098] As an example, the aforementioned execution entity can utilize the target server node, using the decision sub-information as the basis for decision information and the at least one auxiliary decision information as optional decision information, and using the normal field effect and reinforcement learning methods, instruct the at least one extended coordinating agent and the subset of the remaining coordinating agents to coordinate the output of operation and maintenance information, so as to generate the sub-operation and maintenance information corresponding to the processed decision area information.

[0099] Here, the content corresponding to "in some optional implementations of some embodiments" serves as another inventive point of this disclosure, solving another technical problem: "how to achieve accurate generation of operation and maintenance information when the resources of each server node in a server cluster are limited." Based on this, this disclosure achieves dynamic combination or decomposition of decision area information by comparing the server resources corresponding to each server node with the estimated decision resources corresponding to the decision area information. This results in the coordinated decision-making of various agents within the decision area information processed by each server node. Through this parallel execution method, the efficiency of coordinated decision-making by various agents is greatly improved while fully utilizing server resources, and operation and maintenance information can be accurately generated.

[0100] Step 106: Based on the above operation and maintenance information, control the main intelligent agent to instruct each coordinating intelligent agent to perform equipment operation and maintenance on the target substation.

[0101] In some embodiments, the aforementioned execution entity may, based on the aforementioned operation and maintenance information, control the aforementioned master intelligent agent to instruct various coordinating intelligent agents to perform equipment operation and maintenance processing on the aforementioned target substation.

[0102] As an example, the aforementioned executing entity can generate multiple control commands corresponding to the operation and maintenance information. These multiple control commands are then sent to the master intelligent agent, which, based on these commands, controls the automated machinery and devices corresponding to each coordinating intelligent agent to perform equipment operation and maintenance on the target substation.

[0103] In some optional implementations of certain embodiments, after step 106, the steps further include:

[0104] The first step, in response to the determination that the size of the corresponding substation to the target substation is not greater than the target size, is to determine whether the resources of the server cluster corresponding to the target substation meet the requirements for coordination output by each agent in the agent coordination diagram to obtain operation and maintenance information, based on the resources of the server cluster corresponding to the target substation. The resources corresponding to the server cluster can be the highest available resources among the various server nodes.

[0105] The second step, in response to the determination that the condition is not met, involves selecting at least one agent from the aforementioned agent coordination graph whose importance level meets the target importance condition. The resources required for the coordination output of this at least one agent are less than the resources required for the aforementioned server cluster. The importance level of an agent characterizes the importance of its decision. That is, the higher the importance level, the more important and crucial the decision output by the agent is considered. The target importance condition can be an agent with an importance level higher than the target importance level. The target importance level can be a pre-set value. The importance level can be a value between 0 and 1; the higher the value, the more important the decision of the corresponding agent.

[0106] The third step involves generating operation and maintenance information for at least one of the aforementioned intelligent agents based on the real-time operational dataset, using mean field theory and reinforcement learning. This information serves as the operation and maintenance information for the target substation.

[0107] The above embodiments of this disclosure have the following beneficial effects: Through the multi-agent-based substation operation and maintenance method of some embodiments of this disclosure, accurate generation of substation operation and maintenance information can be achieved while reducing the computing resources of the agents, ensuring the normal operation of the substation. Specifically, the reason for the exhaustion of computing resources during the generation of relevant operation and maintenance information is that: when the substation is large, the number of agents set for the substation is large, resulting in a large space for joint actions of the agents and an exponential increase in the linkage action controls, leading to a large amount of server resources being used, resulting in the exhaustion of memory and computing resources. Based on this, the multi-agent-based substation operation and maintenance method of some embodiments of this disclosure firstly, in response to determining that the substation size of the target substation is higher than the target size, obtains the real-time operation dataset and agent coordination graph corresponding to the target substation. The agent coordination graph includes: a master agent and a set of coordinating agents. Here, after determining that the substation size is higher than the target size, by setting up an agent coordination graph, the coordination relationship between various substations can be displayed, so that when performing graph partitioning later, closely related agents can be grouped together for collaborative decision-making as much as possible. The master agent here is responsible for global decision-making and core function integration. The coordinating agent is responsible for executing related tasks among the various agents in the complex subsystem. By setting up the master agent and the coordinating agent set, efficient execution of various events corresponding to the target substation can be achieved. Then, using the aforementioned master agent, the aforementioned agent coordination graph is partitioned to obtain at least one agent group. Each agent group is partitioned based on target partitioning features, and the agents in each agent group make collaborative decisions. Here, by using graph partitioning, the agent groups that make collaborative decisions under different target requirements (i.e., target partitioning features) can be determined. By partitioning based on different target partitioning features, it is not necessary to make overall decisions for all agents; only local decisions of the coordinating agents under the corresponding partitioning features are needed. This can greatly reduce the number of agents involved in overall decision-making, avoid the occurrence of large joint action spaces for agents, large amounts of server resources, and exhaustion of memory and computing resources. Next, for each agent group, the following generation steps are performed: First, for each agent group, based on the corresponding real-time running data set, accurate group decision information can be generated. Second, the group decision information set is displayed on the decision page corresponding to the main agent, allowing the operation and maintenance object to confirm the group decision information on the decision page, facilitating subsequent generation of overall decisions under the target segmentation features. Third, based on the candidate decision information determined on the decision page, an agent decision graph corresponding to the agent group is generated, supporting various visualization operations.Here, by generating an agent decision graph, the operation and maintenance object can clearly understand the coordination relationships and decision-making status among various agents, enabling the operation and maintenance object to adaptively adjust the individual intelligent decisions in the agent decision graph and enhance the accuracy of the information in the agent decision graph. Next, based on at least one agent decision graph, agent filtering of the aforementioned coordinated agent set is performed to reduce the number of decisions made by agents, avoiding further waste of server resources, and accurately obtaining the agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set. Furthermore, based on the aforementioned agent decision information and the aforementioned coordinated agent subset, the operation and maintenance information corresponding to the aforementioned target substation is accurately generated with minimal waste of computing resources. Finally, based on the aforementioned operation and maintenance information, the master agent is controlled to instruct the various coordinated agents to perform equipment operation and maintenance processing on the aforementioned target substation, ensuring the normal operation of the target substation and avoiding fault problems. In summary, by performing diverse graph partitioning of the agent coordination graph according to target segmentation characteristics, it is possible to initially and accurately identify agents that do not require subsequent decision-making and to determine the individual decision information of the agents that do make decisions. Based on this, operation and maintenance information can be accurately generated with fewer intelligent agents and fewer joint action controls to maintain the normal operation of substations.

[0108] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a substation operation and maintenance device based on multi-agent systems. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, this multi-agent-based substation operation and maintenance device can be specifically applied to various electronic devices.

[0109] like Figure 2As shown, a substation operation and maintenance device 200 based on multi-agent systems includes: an acquisition unit 201, a partitioning unit 202, an execution unit 203, a filtering unit 204, a generation unit 205, and a control unit 206. The acquisition unit 201 is configured to, in response to determining that the target substation's size is greater than the target substation's size, acquire the real-time operation dataset and agent coordination graph corresponding to the target substation. The agent coordination graph corresponds to: a master agent and a set of coordinating agents. The partitioning unit 202 is configured to, using the master agent, partition the agent coordination graph to obtain at least one agent group. Each agent group is partitioned based on target partitioning features, and the agents in each agent group make collaborative decisions. The execution unit 203 is configured to, for each agent group, perform the following generation steps: for each agent group, generate group decision information corresponding to the agent group based on the real-time operation data set corresponding to the agent group; and then... The information set is displayed on the decision page corresponding to the main agent; based on the candidate decision information determined on the decision page, an agent decision graph corresponding to the agent set is generated, and the agent decision graph supports various visualization operations; the filtering unit 204 is configured to perform agent filtering of the coordinated agent set based on at least one agent decision graph to obtain agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set; the generation unit 205 is configured to generate operation and maintenance information corresponding to the target substation based on the agent decision information and the coordinated agent subset; the control unit 206 is configured to control the main agent to instruct each coordinated agent to perform equipment operation and maintenance processing on the target substation based on the operation and maintenance information.

[0110] It is understandable that the units described in the multi-agent-based substation operation and maintenance device 200 are similar to those in the reference system. Figure 1 The steps described in the method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the multi-agent-based substation operation and maintenance device 200 and the units contained therein, and will not be repeated here.

[0111] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device (e.g., an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0112] like Figure 3As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0113] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0114] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0115] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0116] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0117] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: in response to determining that the target substation's corresponding substation size is higher than the target size, acquire the real-time operation dataset and agent coordination graph corresponding to the target substation, wherein the agent coordination graph corresponds to: a master agent and a set of coordinating agents; using the master agent, partition the agent coordination graph to obtain at least one agent group, each agent group being partitioned based on target partitioning features, and the agents in each agent group making collaborative decisions; for each agent group, perform the following generation steps: for each agent group, based on the real-time operation data corresponding to the agent group... The system generates group decision information corresponding to the aforementioned intelligent agent group; displays the group decision information set on the decision page corresponding to the aforementioned master intelligent agent; generates an intelligent agent decision graph corresponding to the aforementioned intelligent agent group based on the candidate decision information determined on the aforementioned decision page, the intelligent agent decision graph supporting various visualization operations; performs intelligent agent filtering of the aforementioned coordinated intelligent agent set based on at least one intelligent agent decision graph, obtaining intelligent agent decision information corresponding to the coordinated intelligent agent subset and the remaining coordinated intelligent agent set; generates operation and maintenance information corresponding to the aforementioned target substation based on the aforementioned intelligent agent decision information and the aforementioned coordinated intelligent agent subset; and controls the aforementioned master intelligent agent to instruct each coordinated intelligent agent to perform equipment operation and maintenance processing on the aforementioned target substation based on the aforementioned operation and maintenance information.

[0118] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0119] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0120] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor can be described as including an acquisition unit, a partitioning unit, an execution unit, a filtering unit, a generation unit, and a control unit. The names of these units do not necessarily limit the specific unit; for example, a control unit can also be described as "a unit that, based on the aforementioned maintenance information, controls the main intelligent agent to instruct various coordinating intelligent agents to perform equipment maintenance processing on the target substation."

[0121] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0122] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A substation operation and maintenance method based on multi-agent systems, comprising: In response to determining that the size of the corresponding substation of the target substation is higher than the target size, the real-time operation dataset and agent coordination diagram of the target substation are obtained. The agent coordination diagram corresponds to the existence of: a master agent and a set of coordinating agents. In response to the agent processing operation performed by the operation and maintenance object on the corresponding agent coordination graph on the decision page, the agent nodes and agent edges in the agent coordination graph are adjusted according to the agent processing operation to obtain the adjusted agent coordination graph. The decision page supports the click operation of each agent node in the agent coordination graph. After clicking the agent node, a pop-up window corresponding to the agent node appears. The pop-up window displays at least one of the following: associated device, associated agent node, environment status, execution action set, and confirmation information on whether to participate in operation and maintenance. The adjusted agent coordination graph is defined as the agent coordination graph, wherein each agent node in the agent coordination graph is a node that has been confirmed to participate in operation and maintenance. Using the main agent, the agent coordination graph is partitioned to obtain at least one agent group. Each agent group is partitioned based on the target partitioning features, and the agents in the agent group make collaborative decisions. For each set of agents, perform the following generation steps: For each group of intelligent agents, group decision information corresponding to the intelligent agent group is generated based on the real-time running data group corresponding to the intelligent agent group. The group decision information set is displayed on the decision page corresponding to the main intelligent agent; Based on the candidate decision information determined on the decision page, an agent decision graph corresponding to the agent set is generated, and the agent decision graph supports various visualization operations. Based on at least one agent decision graph, agent screening of the coordinated agent set is performed to obtain agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set; Based on the decision information of the intelligent agent and the subset of the coordinating intelligent agents, the operation and maintenance information corresponding to the target substation is generated; Based on the operation and maintenance information, the main intelligent agent is controlled to instruct various coordinating intelligent agents to perform equipment operation and maintenance on the target substation.

2. The method according to claim 1, wherein, The process of using the main agent to partition the agent coordination graph to obtain at least one agent set includes: Obtain the operation and maintenance direction information corresponding to the target substation entered on the decision page; The operation and maintenance direction information is decomposed into features to obtain at least one target partitioning feature and at least one corresponding feature partitioning direction; For each target partitioning feature, perform the following partitioning steps: Determine the target feature partitioning direction corresponding to the target partitioning feature; Using the main agent, the agent coordination graph is partitioned according to the target partitioning features to obtain an initial agent sub-graph set; Using the main agent, determine the adjustability of the initial decision for each agent in the initial agent sub-graph set under the target feature partitioning direction; The obtained initial intelligent agent sub-graph set is displayed on the page jump-to by the graph partitioning control on the decision page, so as to confirm and adjust the adjustability of each partitioning graph under the target partitioning feature and the corresponding initial decision. Obtain the adjusted agent subgraph set and the corresponding decision adjustability set for the initial agent subgraph set; The intelligent agent set corresponding to the intelligent agent sub-graph is used to determine the intelligent agent set corresponding to the target partitioning feature.

3. The method according to claim 2, wherein, The step of generating group decision information corresponding to the intelligent agent group based on the real-time operation data group corresponding to the intelligent agent group includes: Based on the feature division direction corresponding to the agent group, set the reward and punishment function corresponding to the agent group; Based on the decision adjustability group and real-time operation data group corresponding to the intelligent agent group, the group decision information corresponding to the intelligent agent group is generated using the reward and punishment function.

4. The method according to claim 1, wherein, The method further includes: In response to determining that the size of the corresponding substation of the target substation is not higher than the target size, based on the resources of the server cluster corresponding to the target substation, it is determined whether the resources meet the requirements for the coordination output of each intelligent agent in the intelligent agent coordination diagram to obtain operation and maintenance information. In response to the determination that the condition is not met, at least one agent whose importance meets the target importance condition is selected from the agent coordination graph, wherein the resources required for the coordination output of the at least one agent are less than the resources corresponding to the server cluster. Based on the real-time running dataset, mean field theory and reinforcement learning are used to generate operation and maintenance information for the execution of at least one intelligent agent, which serves as the operation and maintenance information for the target substation.

5. A substation operation and maintenance device based on multi-agent systems, comprising: The acquisition unit is configured to, in response to determining that the target substation has a larger substation size than the target size, acquire the real-time operation dataset and agent coordination diagram corresponding to the target substation. The agent coordination diagram corresponds to the existence of: a master agent and a set of coordinating agents. The device further includes: responding to the agent processing operation performed by the maintenance object on the agent coordination graph corresponding to the decision page, adjusting each agent node and agent edge in the agent coordination graph according to the agent processing operation to obtain an adjusted agent coordination graph, wherein the decision page supports click operations on each agent node in the agent coordination graph, and after clicking on an agent node, a pop-up window corresponding to the agent node appears, displaying at least one of the following: associated device, associated agent node, environmental status, execution action set, and confirmation information on whether to participate in maintenance; and determining the adjusted agent coordination graph as the agent coordination graph, wherein each agent node in the agent coordination graph is a node that has confirmed whether to participate in maintenance; The partitioning unit is configured to use the main agent to partition the agent coordination graph to obtain at least one agent group. Each agent group is partitioned based on the target partitioning features, and the agents in the agent group make collaborative decisions. The execution unit is configured to perform the following generation steps for each agent group: for each agent group, generate group decision information corresponding to the agent group based on the real-time running data group corresponding to the agent group; display the group decision information set on the decision page corresponding to the main agent; generate an agent decision graph corresponding to the agent group based on the candidate decision information determined on the decision page, wherein the agent decision graph supports various visualization operations; The filtering unit is configured to perform agent filtering of the coordinated agent set based on at least one agent decision graph to obtain agent decision information corresponding to the coordinated agent subset and the remaining coordinated agent set; The generation unit is configured to generate operation and maintenance information corresponding to the target substation based on the decision information of the intelligent agent and the subset of the coordinating intelligent agents. The control unit is configured to control the main intelligent agent to instruct various coordinating intelligent agents to perform equipment operation and maintenance on the target substation based on the operation and maintenance information.

6. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-4.

7. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the method as described in any one of claims 1-4.

Images