Multi-agent collaborative dynamic safety regulation and control system for power system

Abstract

The invention belongs to the technical field of electric power, and particularly relates to a multi-agent collaborative dynamic safety regulation and control system for an electric power system, which realizes a task redistribution mechanism based on load and communication capability weighted evaluation by introducing a state monitoring module, an abnormal response module, a capability evaluation module and a priority adjustment module. Therefore, the remaining agents can quickly undertake the task of the fault node according to the own processing margin, and the smooth execution of the high-energy-consumption task is preferentially guaranteed. The mechanism not only improves the response speed of the system, but also enhances the overall operation stability and the resource utilization efficiency. Therefore, compared with the prior art, the method has the advantages that higher task migration success rate and shorter recovery time are shown in the fault scene, and the safety and robustness of the power system under the unsteady state condition are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of electric power technology, and in particular relates to a multi-agent collaborative dynamic safety control system for an electric power system. Background Art

[0002] In current power systems, multi-agent collaborative control has been widely applied in key areas such as grid dispatching, load distribution, and fault recovery. Traditional multi-agent systems typically employ centralized or fixed-rule task allocation mechanisms. When an agent fails, the system often relies on pre-set backup nodes or static priority strategies for task transfer. However, this approach suffers from slow response times to sudden failures or load fluctuations, and struggles to balance the real-time load status and communication capabilities of the remaining agents, easily leading to task migration failures or localized system overloads.

[0003] A common approach in existing technologies is to periodically collect status information from each agent and make task redistribution decisions at a central node. While this approach offers a certain degree of fault tolerance, it relies on the computing power of the central node and the stability of the communication link. In the event of node failure or significant network latency, this can easily lead to scheduling delays or even task loss. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-agent collaborative dynamic safety control system for power systems, to achieve rapid and reasonable redistribution of tasks of remaining agents, to avoid system performance degradation or task execution interruption, and to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a multi-agent collaborative power system dynamic safety control system, comprising:

[0006] A status monitoring module for real-time monitoring of the operating status of each agent and recording the load rate of the agent and the stability of the communication link;

[0007] When any agent is detected to be abnormal, an abnormal response module is used to obtain the task list of the abnormal agent and the current resource occupancy parameters;

[0008] A capability assessment module that calculates the remaining processing capacity of each remaining agent based on the task list, where the remaining processing capacity is weighted by the load rate and the communication bandwidth;

[0009] According to the remaining processing capacity, the task priorities of the remaining agents are dynamically adjusted, and the priority adjustment module is preferentially assigned to high energy consumption tasks.

[0010] Preferably, real-time monitoring of the operating status of each agent includes:

[0011] Collect the CPU usage, memory usage and task queue length of each agent to generate the operation load index;

[0012] Based on the operating load index, calculating the fluctuation value of the agent in the continuous time window;

[0013] Build a communication quality score by receiving signal strength and response delay data sent by neighboring agents;

[0014] Combining the fluctuation value and communication quality score, it is determined whether the current agent is in an abnormal edge state. If it exceeds the set threshold, it is marked as an unstable node.

[0015] Preferably, obtaining the task list and current resource occupancy parameters of the abnormal agent includes:

[0016] After receiving the unstable node mark, a task snapshot request is sent to the node to obtain the set of tasks being executed and waiting to be executed;

[0017] Based on the task set, query the data volume and estimated remaining execution time corresponding to each task;

[0018] Summarize the data volume and estimated remaining execution time of the task and calculate the total pending load of the abnormal agent;

[0019] Combined with the total load to be processed and the current available memory capacity of the agent, determine whether to trigger the emergency migration priority and arrange the task migration order according to the data volume.

[0020] Preferably, calculating the remaining processing capacity of each remaining agent based on the task list includes:

[0021] Get the current load status and available communication bandwidth of each remaining agent;

[0022] Calculating an available processing margin according to the current load state, and generating a weighted processing capacity index by combining the available processing margin with available communication bandwidth;

[0023] The weighted processing capability indicators of all remaining agents are normalized to obtain a task acceptance ratio allocation benchmark for each agent.

[0024] Preferably, according to the remaining processing capacity, the task priorities of the remaining agents are dynamically adjusted to give priority to high energy consumption tasks, including:

[0025] Sort the weighted processing capacity indicators of the remaining agents and select the agent with the highest processing margin as the main allocation node;

[0026] Obtain the historical energy consumption records of each task in the abnormal agent and set the task energy consumption level;

[0027] Insert the task queue of the main allocation node from high to low according to the task energy consumption level, replacing the low energy consumption tasks;

[0028] Broadcast the updated task list to the remaining agents and synchronously update the local task scheduling flags.

[0029] Preferably, a task allocation module is further included, which is used to construct a task migration matrix, split the tasks of abnormal agents according to priority and map them to the remaining agents; specifically, the following steps are included:

[0030] Sort the tasks according to their energy consumption levels to form an ordered task queue, with the highest energy consumption tasks at the front;

[0031] Based on the task acceptance ratio of each agent, determine the upper limit of tasks that each agent can accept;

[0032] Allocate the tasks in the ordered task queue to the remaining agents in turn, and map the tasks to the corresponding agent execution list without exceeding the upper limit;

[0033] Generate a task transfer matrix, where the matrix elements represent the assignment relationship between tasks and agents.

[0034] Preferably, a time coordination module is further included, which is used to determine the execution time window of each task in the remaining agents; specifically, the following steps are included:

[0035] Get the estimated completion time set in the current task queue of the remaining agents, which is used to indicate the end time of the last task of each agent;

[0036] Based on the task transfer matrix, extract a new task sequence assigned to each agent and calculate the estimated execution time of each task;

[0037] Set a start time window for each new task, and the start time should not be earlier than the end time of the last task plus the minimum interval time;

[0038] The starting time window is sequentially superimposed with the starting time of subsequent tasks to generate a complete task time schedule.

[0039] Preferably, a topology optimization module is further included, which is used to update the communication topology of the power system and establish redundant paths between the remaining intelligent agents and the task target node; specifically, the following steps are included:

[0040] Get the current communication path set between the remaining agents and the task target node;

[0041] Based on the communication quality score, select available nodes with scores higher than a set threshold as candidate relay points;

[0042] Selecting two nodes in different directions from the candidate relay points to construct a second communication path, thereby forming an alternative transmission channel parallel to the original path;

[0043] The original path and the second communication path are activated at the same time, and the distribution of data packets on the two paths is controlled by the traffic distribution ratio.

[0044] Preferably, a feasibility verification module is further included to verify the feasibility of the task migration matrix. If the total processing capacity of the remaining agents is lower than a preset threshold, the backup node is triggered to start. Specifically, the following steps are included:

[0045] Based on the task migration matrix, the total amount of data assigned to each agent is counted, and the overall load demand is calculated in combination with the task energy consumption level;

[0046] Obtain the sum of the weighted processing capabilities of all remaining agents and compare it with the overall load demand to determine whether there is a resource gap;

[0047] If there is a resource gap, a resource gap signal is generated to trigger a standby node access request;

[0048] Activate the recently unused standby node according to the resource gap signal and add it to the task acceptance list.

[0049] Preferably, the system further includes an execution monitoring module for synchronously executing the task migration plan, continuously monitoring the load changes of the remaining agents, and dynamically correcting the task allocation ratio; specifically, the system includes the following steps:

[0050] According to the task migration matrix, each task is distributed to the corresponding agent to start execution, and the initial allocation timestamp is recorded;

[0051] During the execution process, the current load status and communication quality information of each agent are periodically collected;

[0052] Calculating the deviation between the current load state and the pre-allocated reference, and activating an adjustment mechanism if the deviation exceeds a set threshold;

[0053] Based on the deviation, tasks that exceed expectations are reallocated to other agents with lower loads, and the task mapping relationship is updated.

[0054] Technical effects and advantages of the present invention: The multi-agent collaborative power system dynamic safety control system proposed in the present invention has the following advantages over the existing technology:

[0055] The present invention realizes a task redistribution mechanism based on weighted evaluation of load and communication capabilities by introducing a status monitoring module, an abnormal response module, a capability assessment module, and a priority adjustment module, so that the remaining intelligent agents can quickly take over the tasks of the faulty nodes according to their own processing margins, and give priority to ensuring the smooth execution of high-energy-consuming tasks. This mechanism not only improves the response speed of the system, but also enhances the overall operational stability and resource utilization efficiency. Therefore, compared with the existing technology, the present invention shows a higher task migration success rate and a shorter recovery time in fault scenarios, effectively improving the safety and robustness of the power system under non-steady-state conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A block diagram of the multi-agent collaborative power system dynamic safety control system of the present invention;

[0057] Figure 2 This is the load analysis diagram of the intelligent agent of the present invention;

[0058] Figure 3 This is the matrix allocation diagram of the present invention. DETAILED DESCRIPTION

[0059] 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 embodiments described are only part of the embodiments of the present invention, not all of the embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0060] The present invention provides Figure 1 The multi-agent collaborative power system dynamic security control system shown in the figure includes the following modules:

[0061] The status monitoring module is used to monitor the operating status of each agent in real time and record the load rate and communication link stability of the agent. It includes the following steps:

[0062] Collect CPU usage of each agent , memory usage and task queue length , generate operating load indicators , where i represents the i-th agent; this formula combines the core resources of the agent (CPU and memory) and the scale of the task it is processing to form a comprehensive operating load assessment value.

[0063] Based on operating load indicators , calculate the fluctuation value Δ of the agent in the continuous time window , Δ It is the variation of the agent's operating load over a period of time, and is used to measure whether its operating state is stable.

[0064] Build a communication quality score by receiving signal strength and response delay data sent by neighboring agents , where j represents the agent adjacent to i; For receiving signal strength indication, is the round-trip time delay; this formula quantifies link quality by dividing the received signal strength indicator (RSSI) by the round-trip time delay (RTT), giving higher weight to high signal strength and low latency. The +1 is to prevent division by zero errors.

[0065] Combined fluctuation value Δ and communication quality score , to determine whether the current agent is in an abnormal edge state, if Δ > or < , it is marked as an unstable node; is the operating load fluctuation threshold; The lower threshold of the communication quality score is set. Two independent judgment conditions are set to judge whether the agent is on the edge of abnormality from the two dimensions of operation stability and communication stability.

[0066] The abnormal response module is used to obtain the task list and current resource occupancy parameters of the abnormal agent when any agent is detected to be abnormal. It includes the following steps:

[0067] After receiving the unstable node mark, send a task snapshot request to the node to obtain the set of task identifiers that are being executed and waiting to be executed. ; When a node is marked as unstable, actively request its current task list to prepare for task migration.

[0068] Based on task ID collection , query the amount of data corresponding to each task and the estimated remaining execution time ; Indicates the size of the data that the task needs to process, It reflects the expected completion time of the task, and the two together determine the migration priority of the task.

[0069] Aggregated data volume and the estimated remaining execution time , calculate the total pending load of abnormal agents ; Taking into account the task data volume and time consumption, it represents the overall workload of the abnormal node and is used to determine whether it can be taken over by other nodes.

[0070] Combined total pending load The agent's current available memory , determine whether to trigger the emergency migration priority, if > , then arrange the task migration order from large to small according to the data volume.

[0071] The capacity evaluation module is used to calculate the remaining processing capacity of each remaining agent based on the task list, wherein the remaining processing capacity is obtained by weighting the load rate and the communication bandwidth; and comprises the following steps:

[0072] Get the current load rate of each remaining agent and available communication bandwidth , the two together determine the agent's remaining capacity. Calculate its available processing headroom ; Indicates the degree to which the agent can continue to accept new tasks. Higher values ​​indicate greater ability to take on additional tasks.

[0073] Combined with available processing headroom and communication bandwidth , generating weighted indicators , where α is the processing weight factor and β is the communication weight factor, so that the system can dynamically adjust the task allocation strategy according to actual needs.

[0074] Weighted index for all remaining agents Perform normalization to obtain the task acceptance ratio of each agent. Perform normalization so that the sum is 1, forming the task acceptance ratio of each node.

[0075] The priority adjustment module is used to dynamically adjust the task priorities of the remaining agents based on the remaining processing capacity, giving priority to high-energy-consuming tasks; and comprises the following steps:

[0076] Weighted index for the remaining agents Sort and select the agent with the highest processing margin as the main allocation node; prioritize assigning tasks to nodes with the most sufficient processing resources to reduce overall scheduling time.

[0077] Obtain the historical energy consumption records of each task in the abnormal agent and set the task energy consumption level ,in is the historical average completion time of task t; A larger value indicates that the task consumes more resources per unit time, that is, "high energy consumption".

[0078] The task queue of the main allocation node is divided into task energy consumption levels Insert from high to low, replace the low energy consumption tasks, and keep the total load of the queue no more than Keep the master node load below its available capacity while ensuring that high-value tasks are prioritized.

[0079] Broadcast the updated task list to the remaining agents and update the local task scheduling flags synchronously to ensure that high energy consumption tasks are executed first. Figure 1 To avoid conflict.

[0080] The task allocation module is used to build a task transfer matrix, split the tasks of abnormal agents according to priority, and map them to the remaining agents. It includes the following steps:

[0081] According to the energy consumption level of the task Sort tasks to form an ordered task queue ,in Indicates the task with the highest energy consumption; high energy consumption tasks are processed first to reduce the overall scheduling complexity.

[0082] Based on the task allocation ratio of each agent , determine the upper limit of tasks that each agent can accept ; Indicates the total amount of tasks that the node should undertake, which is allocated based on its weighted capacity to ensure balanced distribution of tasks.

[0083] Assign the tasks in the task queue Q to the remaining agents in turn, satisfying Under the premise of , the tasks are mapped to the corresponding agent execution list; they are assigned item by item to ensure that the sum of the data volume of each task does not exceed the carrying capacity of the node.

[0084] Generate a task transfer matrix M[i][j], where M[i][j] = 1 indicates that task i is assigned to agent j, and 0 otherwise. This matrix is ​​used to guide task scheduling in the subsequent execution phase. Recording how tasks are assigned in matrix form facilitates subsequent scheduling and monitoring.

[0085] The time coordination module is used to determine the execution time window of each task in the remaining agents; it includes the following steps:

[0086] Get the estimated completion time set of the remaining agents in the current task queue , used to indicate the end time of each agent’s last task;

[0087] Based on the task migration matrix M[i][j], extract the new task sequence assigned to each agent and calculate the estimated execution time of each task ; Set the start time window for each new task , where δ is the minimum interval time to ensure that there is no overlap with the original task;

[0088] Set the start time window The start time of each task is superimposed on the start time of subsequent tasks to generate a complete task schedule, which is used to coordinate the orderly execution of each task on the agent.

[0089] The topology optimization module is used to update the communication topology of the power system and establish redundant paths between the remaining agents and the task target node. It includes the following steps:

[0090] Get the current communication path set between the remaining agents and the task target node , where i represents the agent number and j represents the target node number; based on the communication quality score , select available nodes with scores higher than the set threshold as candidate relay points;

[0091] Select two nodes v and w in different directions from the candidate relay points to build the second path , forming an alternative transmission channel parallel to the original path; is the first alternative path from an agent to the goal node; is the second alternative path, which also goes from the same agent to the same target node, but passes through another set of possible relay nodes.

[0092] The original path and the second path are activated simultaneously, and the distribution of data packets on the two paths is controlled by the traffic distribution ratios α and (1-α) to improve transmission reliability.

[0093] The feasibility verification module is used to verify the feasibility of the task migration matrix. If the total processing capacity of the remaining agents is lower than a preset threshold, the backup node is triggered to start. The module includes the following steps:

[0094] Based on the task migration matrix M[i][j], the total amount of data assigned to each agent is counted. , and combined with the task energy consumption level Calculate overall load requirements ;

[0095] Get the sum of weighted metrics of all remaining agents and with the overall load demand Compare and calculate the difference ;

[0096] If the difference ΔQ>0, it is determined that the current resources are insufficient, and a resource gap signal is generated to trigger a standby node access request;

[0097] According to the resource gap signal, activate the recently unused standby node k, add it to the task acceptance list, and re-evaluate whether the overall carrying capacity of the system meets the requirements. .

[0098] The execution monitoring module is used to synchronously execute the task migration plan, continuously monitor the load changes of the remaining agents, and dynamically adjust the task allocation ratio; it includes the following steps:

[0099] According to the task migration matrix M[i][j], each task is distributed to the corresponding agent to start execution, and the initial allocation timestamp is recorded ; Periodically collect the current load rate of each agent during execution and communication quality score ; Calculate the current load rate and pre-allocation benchmark Deviation ,like >ε (set threshold), the adjustment mechanism is activated; based on the deviation , reallocate the tasks that exceed expectations to other agents with lower loads, update the task mapping relationship, and keep the total task volume unchanged.

[0100] In order to more clearly illustrate the operation method of the above-mentioned multi-agent coordinated power system dynamic security control system, the following will be described in conjunction with specific embodiments, as follows:

[0101] 1. Background Setting

[0102] Five agents (Agent 1 through Agent 5) are deployed in a regional power grid, responsible for distributed load control, equipment status monitoring, and emergency response. At some point, Agent 3 suddenly fails. The system needs to quickly redistribute its tasks to other healthy nodes, ensuring that high-energy-consuming tasks are prioritized.

[0103] 2. System Process Execution Instructions

[0104] 1. Condition monitoring module operation

[0105] The collected data are as follows:

[0106] Agent #timg# #timg# #timg# Agent1 0.4 0.3 2 Agent2 0.6 0.5 3 Agent3 0.9 0.8 5 Agent4 0.3 0.2 2 Agent5 0.5 0.4 3

[0107] Operation load index: ;

[0108] Calculation results:

[0109] Agent #timg# Agent1 (0.4+0.3)*2=1.4 Agent2 (0.6+0.5)*3=3.3 Agent3 (0.9+0.8)*5=8.5 Agent4 (0.3+0.2)*2=1.0 Agent5 (0.5+0.4)*3=2.7

[0110] Fluctuation value Judgment: Assume ΔW threshold =3.0;Agent3 =8.5 is significantly greater than , marked as unstable nodes.

[0111] Communication quality score : Assume that the communication data between Agent3 and other nodes are as follows:

[0112] j #timg# #timg# 1 80 10 2 60 20 4 30 30 5 40 25

[0113] Calculation formula: ;

[0114] Calculation results:

[0115] j #timg# 1 80 / 11=7.27 2 60 / 21=2.86 4 30 / 31=0.97 5 40 / 26=1.54

[0116] Set the communication quality lower limit =2.0, found , further confirming that Agent3 is on the edge of abnormality.

[0117] Conclusion: Agent3 is judged as an unstable node and enters the abnormal response stage.

[0118] 2. Abnormal response module starts

[0119] Get the task list of Agent3: ;

[0120] Query the data volume and estimated remaining time for each task:

[0121] Task #timg# #timg# t1 500 20 t2 300 15 t3 400 10

[0122] Calculate the total pending load:

[0123] =(500*20)+(300*15)+(400*10)=10000+4500+4000=18500.

[0124] Agent3's current available memory M_free = 15000.

[0125] Compare > →Trigger the emergency migration mechanism, press The order of tasks from largest to smallest is: t1>t3>t2.

[0126] 3. Capacity Assessment Module Operation

[0127] Get the load rate and bandwidth of the remaining agents:

[0128] Agent #timg# #timg# Agent1 0.3 100 Agent2 0.5 80 Agent4 0.2 120 Agent5 0.4 90

[0129] Calculating available processing headroom : ;

[0130] Agent #timg# Agent1 0.7 Agent2 0.5 Agent4 0.8 Agent5 0.6

[0131] Weighted indicators , let α=0.5, β=0.5:

[0132] Agent #timg# Agent1 0.50.7+0.5100=0.35+50=50.35 Agent2 0.50.5+0.580=0.25+40=40.25 Agent4 0.50.8+0.5120=0.4+60=60.4 Agent5 0.50.6+0.590=0.3+45=45.3

[0133] After normalization, the acceptance ratio benchmark of each node is obtained:

[0134] Agent Undertaking ratio Agent1 50.35 / 196.3=0.257 Agent2 40.25 / 196.3=0.205 Agent4 60.4 / 196.3=0.308 Agent5 45.3 / 196.3=0.230

[0135] 4. Priority adjustment module operation

[0136] Sort tasks by energy consumption level:

[0137] Historical average completion time as follows:

[0138] Task #timg# t1 25 t2 18 t3 12

[0139] calculate :

[0140] Task #timg# t1 500 / 25=20 t2 300 / 18=16.7 t3 400 / 12=33.3

[0141] After sorting: t3>t1>t2.

[0142] Agent4 with the highest processing margin is selected as the main allocation node and high-energy consumption tasks are inserted.

[0143] 5. Task allocation module runs

[0144] Ordered task queue , according to the acceptance ratio :

[0145] Agent #timg# Agent1 50.35*18500=931475 Agent2 40.25*18500=744625 Agent4 60.4*18500=1117400 Agent5 45.3*18500=838050

[0146] Assign tasks in sequence:

[0147] t3: D = 400 → assigned to Agent 4;

[0148] t1: D = 500 → assigned to Agent 1;

[0149] t2: D = 300 → assigned to Agent 5;

[0150] Generate the task migration matrix M[i][j] as follows:

[0151] Task\Agent Agent1 Agent2 Agent4 Agent5 t1 1 0 0 0 t2 0 0 0 1 t3 0 0 1 0

[0152] 6. Time coordination module operation

[0153] Agent4 current task completion time set ={12:00,12:30}→max =12:30;

[0154] Minimum interval δ = 5 minutes;

[0155] New task start time window =12:35, subsequent tasks will be postponed.

[0156] 7. Topology optimization module operation

[0157] Original path →5 is unavailable, build a redundant path ;

[0158] Select two directional nodes v=Agent1 and w=Agent4 to form an alternative path and improve transmission reliability.

[0159] 8. Feasibility verification module operation

[0160] =400+500+300=1200;

[0161] Overall load requirements:

[0162] Q_total=Σ(D_i_j*E_t)=40033.3+50020+300*16.7=13320+10000+5010=28330;

[0163] =196.3;

[0164] =28330-196.3=28133.7>0;

[0165] Trigger the standby node k to access and re-evaluate the carrying capacity.

[0166] 9. Execute the monitoring module

[0167] Record initial allocation timestamp =12:00;

[0168] Periodically collect load changes, if the deviation If it exceeds ε=0.1, the tasks are reallocated.

[0169] like Figure 2 As shown, the following conclusions can be drawn based on the data analysis results:

[0170] Agent load analysis:

[0171] Agent3's load index ( =8.5) significantly exceeded the threshold (3.0), which was 2-8 times that of other agents, verifying its determination as an unstable node;

[0172] The loads of the other agents are all within the safe range, among which Agent4 has the lowest load ( =1.0), which has the greatest potential for acceptance.

[0173] Communication quality assessment:

[0174] The communication quality score between Agent3 and Agent4 ( =0.97) is lower than the threshold (2.0), there is a communication link risk, and the communication quality with Agent1 is the best ( =7.27), which can be used as a priority migration path.

[0175] Task allocation plan:

[0176] The high-energy-consuming task t3 (energy consumption level 33.3) is assigned to Agent 4, which has the largest processing margin;

[0177] The task allocation matrix shows that load balancing is achieved to avoid overloading of a single node;

[0178] Emergency relocation mechanism The dual criteria of size (t1>t3>t2) and energy level (t3>t1>t2) optimize the execution order;

[0179] System optimization effect:

[0180] Through weighted indicators The calculation achieves a balance between computing resources (α=0.5) and bandwidth resources (β=0.5);

[0181] Topology optimization established dual redundant paths for Agent 1 and Agent 4, improving the system's fault tolerance.

[0182] The ε=0.1 deviation threshold set by the monitoring module can ensure the timeliness of dynamic adjustment.

[0183] like Figure 3 As shown, Figure 3The vertical axis (Y-axis) shows the queue of tasks to be migrated, arranged in descending order of energy consumption priority: t3 → t1 → t2 (energy consumption level: 33.3 > 20 > 16.7). The horizontal axis (X-axis) shows the available agent nodes, arranged in descending order of acceptance ratio: Agent4 → Agent1 → Agent5 → Agent2 (acceptance ratio: 30.8% > 25.7% > 23.0% > 20.5%).

[0184] Hollow square: indicates the task allocation location.

[0185] t3 is at the top of the matrix (Y=0), reflecting the highest priority.

[0186] t3 allocation: occupies the top position of the Agent4 column, and the hollow square indicates that this node has the largest processing margin ( =0.8), matching the high energy consumption task requirements (E_t=33.3).

[0187] t1 allocation: Located in the middle of the Agent1 column, the hollow squares reflect suboptimal node selection ( =0.7) Balance bandwidth resources ( =100);

[0188] t2 allocation: Located at the bottom of the Agent5 column, it indicates that the remaining tasks are allocated according to the proportion of tasks to be undertaken to avoid overloading Agent2 ( =0.5 has reached medium load).

[0189] Through this complete process, the system successfully achieved rapid response to abnormal intelligent agents, task reallocation, resource rebalancing, and communication link optimization, effectively resolving the slow response and low resource utilization issues inherent in traditional approaches. Ultimately, this ensured the continuous operation of the power system under unsteady conditions and prioritized execution of critical tasks.

[0190] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-agent collaborative power system dynamic safety control system, characterized by: include: A status monitoring module for real-time monitoring of the operating status of each agent and recording the load rate of the agent and the stability of the communication link; When any agent is detected to be abnormal, an abnormal response module is used to obtain the task list of the abnormal agent and the current resource occupancy parameters; A capability assessment module that calculates the remaining processing capacity of each remaining agent based on the task list, where the remaining processing capacity is weighted by the load rate and the communication bandwidth; According to the remaining processing capacity, the task priorities of the remaining agents are dynamically adjusted, and the priority adjustment module is preferentially assigned to high energy consumption tasks.

2. The multi-agent collaborative power system dynamic safety control system according to claim 1, characterized in that: Real-time monitoring of the operating status of each intelligent agent, including: Collect the CPU usage, memory usage and task queue length of each agent to generate the operation load index; Based on the operating load index, calculating the fluctuation value of the agent in the continuous time window; Build a communication quality score by receiving signal strength and response delay data sent by neighboring agents; Combining the fluctuation value and communication quality score, it is determined whether the current agent is in an abnormal edge state. If it exceeds the set threshold, it is marked as an unstable node.

3. The multi-agent collaborative power system dynamic safety control system according to claim 2 is characterized by: Get the task list and current resource usage parameters of the abnormal agent, including: After receiving the unstable node mark, a task snapshot request is sent to the node to obtain the set of tasks being executed and waiting to be executed; Based on the task set, query the data volume and estimated remaining execution time corresponding to each task; Summarize the data volume and estimated remaining execution time of the task and calculate the total pending load of the abnormal agent; Combined with the total load to be processed and the current available memory capacity of the agent, determine whether to trigger the emergency migration priority and arrange the task migration order according to the data volume.

4. The multi-agent collaborative power system dynamic safety control system according to claim 1, characterized in that: Calculate the remaining processing capacity of each remaining agent based on the task list, including: Get the current load status and available communication bandwidth of each remaining agent; Calculating an available processing margin according to the current load state, and generating a weighted processing capacity index by combining the available processing margin with available communication bandwidth; The weighted processing capability indicators of all remaining agents are normalized to obtain a task acceptance ratio allocation benchmark for each agent.

5. The multi-agent collaborative power system dynamic safety control system according to claim 4, characterized in that: According to the remaining processing capacity, the task priorities of the remaining agents are dynamically adjusted to give priority to high energy consumption tasks, including: Sort the weighted processing capacity indicators of the remaining agents and select the agent with the highest processing margin as the main allocation node; Obtain the historical energy consumption records of each task in the abnormal agent and set the task energy consumption level; Insert the task queue of the main allocation node from high to low according to the task energy consumption level, replacing the low energy consumption tasks; Broadcast the updated task list to the remaining agents and synchronously update the local task scheduling flags.

6. The multi-agent collaborative power system dynamic safety control system according to claim 5, characterized in that: The system also includes a task allocation module for constructing a task transfer matrix, splitting the tasks of abnormal agents according to priority and mapping them to the remaining agents. The specific steps include: Sort the tasks according to their energy consumption levels to form an ordered task queue, with the highest energy consumption tasks at the front; Based on the task acceptance ratio of each agent, determine the upper limit of tasks that each agent can accept; Allocate the tasks in the ordered task queue to the remaining agents in turn, and map the tasks to the corresponding agent execution list without exceeding the upper limit; Generate a task transfer matrix, where the matrix elements represent the assignment relationship between tasks and agents.

7. The multi-agent collaborative power system dynamic safety control system according to claim 1, characterized in that: It also includes a time coordination module for determining the execution time window of each task in the remaining agents; specifically, it includes the following steps: Get the estimated completion time set in the current task queue of the remaining agents, which is used to indicate the end time of the last task of each agent; Based on the task transfer matrix, extract a new task sequence assigned to each agent and calculate the estimated execution time of each task; Set a start time window for each new task, and the start time should not be earlier than the end time of the last task plus the minimum interval time; The starting time window is sequentially superimposed with the starting time of subsequent tasks to generate a complete task time schedule.

8. The multi-agent collaborative power system dynamic safety control system according to claim 2, characterized in that: It also includes a topology optimization module for updating the communication topology of the power system and establishing redundant paths between the remaining intelligent agents and the task target node; specifically, it includes the following steps: Get the current communication path set between the remaining agents and the task target node; Based on the communication quality score, select available nodes with scores higher than a set threshold as candidate relay points; Selecting two nodes in different directions from the candidate relay points to construct a second communication path, thereby forming an alternative transmission channel parallel to the original path; The original path and the second communication path are activated at the same time, and the distribution of data packets on the two paths is controlled by the traffic distribution ratio.

9. The multi-agent collaborative power system dynamic safety control system according to claim 5, characterized in that: The system also includes a feasibility verification module for verifying the feasibility of the task migration matrix. If the total processing capacity of the remaining agents is lower than a preset threshold, the standby node is triggered to start. Specifically, the system includes the following steps: Based on the task migration matrix, the total amount of data assigned to each agent is counted, and the overall load demand is calculated in combination with the task energy consumption level; Obtain the sum of the weighted processing capabilities of all remaining agents and compare it with the overall load demand to determine whether there is a resource gap; If there is a resource gap, a resource gap signal is generated to trigger a standby node access request; Activate the recently unused standby node according to the resource gap signal and add it to the task acceptance list.

10. The multi-agent collaborative power system dynamic safety control system according to claim 6, characterized in that: It also includes an execution monitoring module for synchronously executing the task migration plan, continuously monitoring the load changes of the remaining agents, and dynamically correcting the task allocation ratio; The specific steps include: According to the task migration matrix, each task is distributed to the corresponding agent to start execution, and the initial allocation timestamp is recorded; During the execution process, the current load status and communication quality information of each agent are periodically collected; Calculating the deviation between the current load state and the pre-allocated reference, and activating an adjustment mechanism if the deviation exceeds a set threshold; Based on the deviation, tasks that exceed expectations are reallocated to other agents with lower loads, and the task mapping relationship is updated.

Images