A Fault Reconstruction Method for Fully Distributed Ship Power Supply Systems Based on Multi-Agents

Abstract

This invention relates to a fault reconstruction method for a fully distributed ship power supply system based on multi-agent technology, belonging to the field of ship power system automation. This method maps generator sets, loads, and circuit breakers as agents, constructing a fully distributed multi-agent system model. Global network information is obtained iteratively through information acquisition algorithms. After a fault occurs, a two-stage heuristic reconstruction is performed: based on the acquired global information, a heuristic algorithm is used for reconstruction calculation. The first stage prioritizes reconstructing the power supply paths of critical equipment to ensure system survivability; the second stage reconstructs the power supply paths of non-critical equipment to maximize system load recovery. The agents possess autonomy and can independently handle local faults. This invention effectively avoids single-point-of-failure problems, improves information processing efficiency and the real-time and automation level of reconstruction, and while ensuring the survivability of critical loads, also considers the operational economy of generator sets, avoiding excessively low single-pole load rates.

Description

Technical Field

[0001] This invention relates to the field of ship power system automation technology, and in particular to a fault reconfiguration method for a fully distributed ship power supply system based on multiple agents. Background Technology

[0002] Reconfiguration methods for ship electrical systems typically rely on human judgment, with operators providing reconfiguration strategies based on their experience and executing them manually. However, with the increasing complexity and scale of ship electrical networks, and the continuous improvement of industrial automation and intelligent algorithm technologies, existing ship electrical system reconfiguration methods can no longer meet the accuracy, real-time performance, and automation requirements of complex power systems for fault reconfiguration. Therefore, there is an urgent need to develop a new, automatically executable fault reconfiguration method for ship power supply systems based on intelligent algorithms. Summary of the Invention

[0003] To address the challenges of unmanned and intelligent fault reconfiguration in ship power systems, a fully distributed fault reconfiguration method for ship power supply systems based on multi-agent systems is proposed.

[0004] The technical solution of this invention is: a fault reconfiguration method for a fully distributed ship power supply system based on multiple agents, which maps the equipment in the ship power supply system to agents, constructs a fully distributed multi-agent architecture, and each agent only interacts with its neighboring agents; the method includes the following steps: S1, Distributed information acquisition: Each agent obtains global network topology information, device status information and global partition information through local iterative calculation; S2, Fault Isolation and Autonomous Handling: When a system fault occurs and the protection device activates to isolate the fault area, the agent in the fault area determines whether it can resolve the fault independently. If it can, it will autonomously perform the isolation operation without coordinating with agents in other areas. S3, Two-stage heuristic refactoring: Based on the acquired global information, a heuristic algorithm is used for refactoring calculations. In the first stage, the power supply paths of critical equipment are refactored first to ensure system survivability. In the second stage, the power supply paths of non-critical equipment are refactored to maximize the recovery of system load.

[0005] Furthermore, the specific steps for obtaining global network topology information in S1 are as follows: Each agent in the system is assigned a number, and an initial structure vector is constructed for each agent. In the initial structure vector, only the element corresponding to its own number is 1, and the rest are 0. Each agent reads the structure vector of its neighboring agents, and uses the formula... Perform iterative calculations; where, For the first k At +1 iteration, the agent iThe structure vector, For the first k In the next iteration, the agent j The structure vector, For elements of the adjacency matrix, when neighboring agents i , j When the electrical equipment of the next level is connected, Otherwise, it is 0; The iteration stops when the results of two consecutive iterations of the agent are the same. After the iteration ends, the agent checks the non-zero elements in its own structure vector and obtains the numbers of all agents connected to it and the global topology.

[0006] Furthermore, the specific steps for obtaining device status information in S1 are as follows: Collect the status information of all generators, loads and lines in the system, and construct the initial state vector of each intelligent agent. The length of the initial state vector is equal to the sum of the number of all devices in the system. Only the elements containing the status information of the agent itself are non-zero, and the rest are 0. Each agent reads the state vectors of its neighboring agents, and uses the formula... Perform iterative calculations; where, For the first k At +1 iteration, the agent i The state vector, For the first k In the next iteration, the agent j The state vector, for all... i Neighboring agents j , The physical on / off state is not considered; otherwise, the value is 0. Once the iteration converges, the agent checks the non-zero elements in its own state vector and obtains the state information of all devices in the system.

[0007] Furthermore, the specific steps for obtaining global partition information in S1 are as follows: After the structural information iteration is completed, a new partition structure vector is generated for each agent; the system is divided into multiple partitions, and the agents contained in each partition are identified by the ID number of the agent with the smallest ID number in that partition as the first element of its partition structure vector. The partition structure vector is iterated using the same iterative method as the state vector until the partition structure vectors of all agents converge, so that each agent knows the global partition status of the system.

[0008] Furthermore, in S2, when a fault occurs on a low-voltage distribution bus without a backup path, the intelligent agent in the area where the bus is located determines that the fault can be resolved independently and directly controls the disconnection of the tie switch connected to the bus.

[0009] Furthermore, in step S3, before performing the reconfiguration path search, the following rules guide the commissioning of the standby generator: At the start of the reconfiguration, all generators in standby mode are temporarily set to operational mode to participate in the reconfiguration calculation. After the reconfiguration calculation is completed, the power balance of each partition is re-examined. The status of the standby generators in the partition is changed from being in operation to being out of operation in turn, and the average load rate of the generators in the partition is calculated. If the average load rate of the generator is less than a preset threshold, it is determined that the standby generator does not need to be put into operation and remains in a shutdown state; if it is greater than or equal to the preset threshold, it is determined that the standby generator should be put into operation and its state is restored to the operation state.

[0010] Furthermore, the preset threshold is 90%; in S3, for regions with negative power balance, a priority queue is established according to the magnitude of the unbalanced power, with regions having larger unbalanced power ranked at the front of the queue; starting from the first region in the queue, the shortest power supply path is searched so that the power supply region formed by the new path has a positive power margin.

[0011] Furthermore, the power supply path for the key equipment in the first stage of S3 includes: All critical loads are sorted by importance to form a load set; For each load in the load set, search its power supply path in turn, and determine whether the generator load rate of the island exceeds the safety limit when the power is supplied according to the power supply path. When a suitable power supply path is found, or it is determined that the load cannot be powered by any path, it is removed from the load set, and the next load is selected to continue the refactoring; if the critical load is located on the main distribution board, the power supply path of that area is directly adopted.

[0012] Furthermore, the second phase of S3, which reconfigures the power supply path for non-critical equipment, includes: Classify non-critical loads in the system, excluding critical loads, by level; When calculating the load reconfiguration for each level, first set the state of all lower-level loads to 0; Once the load at this level is reconfigured, the lower-level loads are restored to their original state, and the power imbalance area is searched again to continue the reconfiguration calculation for the next level of load.

[0013] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the multi-agent-based fully distributed ship power supply system fault reconfiguration method.

[0014] The beneficial effects of this invention are as follows: This invention is based on a fully distributed ship power supply system fault reconstruction method with multiple agents. When a single point of failure occurs in the ship power system, this reconstruction method can effectively avoid the single point failure problem, improve information processing efficiency, and ensure the survivability of critical loads while taking into account the operating economy of generator sets (avoiding excessively low single-pole load rates). Attached Figure Description

[0015] Figure 1 This is a diagram of the agent structure in the multi-agent-based fully distributed ship power supply system fault reconstruction method of the present invention; Figure 2 This is a simplified structural diagram of a typical ship integrated electric power system and a schematic diagram of the MAS mapping method according to the present invention; Figure 3 This is a flowchart of an independent power system based on a heuristic algorithm, as described in this invention. Figure 4 This is a diagram illustrating an embodiment of the regional power distribution board reconfiguration path in the method of the present invention; Figure 5 This is a diagram illustrating an embodiment of the low-voltage distribution board reconfiguration path in the method of the present invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0017] This invention proposes a post-fault reconfiguration method for a fully distributed shipboard power system based on multi-agent technology. This method treats critical equipment (including generator sets, loads, circuit breakers, etc.) and devices in the system as a single agent, with each agent exchanging information only with its neighboring agents. Agents communicate efficiently and acquire global network information through information acquisition algorithms, enabling system reconfiguration upon the occurrence of a fault.

[0018] This invention establishes a novel MAS (Multi-Agent System) model and proposes a framework for a two-stage reconfiguration algorithm, including an information transfer algorithm. First, through an information acquisition algorithm, each agent can grasp global information. Second, based on the obtained global information, each agent quickly reconfigures the power supply paths of critical equipment in the system to maximize system survivability. Then, other equipment is reconfigured to restore as much of the system load as possible. One advantage of the MAS architecture proposed in this invention is the degree of autonomy of the agents. For load faults within some agents, communication with adjacent agents is not required; the fault can be resolved internally. When a fault occurs, the agent in the area where the fault occurs first needs to determine whether it can resolve the fault independently without assistance from other areas. For example, for a low-voltage distribution bus without a backup path, if a fault occurs, the tie switch connected to that bus is directly disconnected without the assistance of other agents.

[0019] The structure of each agent is as follows: Figure 1 As shown in the diagram, each agent has one or more interfaces, a parent node, and several child nodes. Each agent contains several variables and instructions. Variables are passed from child nodes to parent nodes, and parent nodes pass instructions to child nodes. Internally, the agent dynamically updates its instructions and variables based on the information it receives.

[0020] This invention, tailored to the characteristics of shipboard power systems, maps each power supply area within the system, centered on the main power distribution board of the power station, to a corresponding agent. Taking an 8-power station as an example, such as... Figure 2 As shown.

[0021] The advantages of this MAS structure are: (1) It reduces the complexity of the MAS. The MAS obtains the state of the power grid through information interaction with the physical power grid layer. (2) The function of the agent can be enhanced accordingly, so that the reconstruction results can be optimized.

[0022] For a fully distributed power system automation (MAS), each agent can only communicate with neighboring agents. Therefore, an independent distributed information acquisition algorithm for the power system is needed to collect global device information to facilitate system reconfiguration after a fault occurs. The specific steps of the information acquisition algorithm are as follows: Step 1: Obtaining power system structure information: Each agent is numbered from 1 to n, forming the initial state vector for each agent. Each agent reads the state vectors of its neighboring agents and iterates through them. , For the first k At +1 iteration, the agent i The structure vector, For the firstk In the next iteration, the agent j The structure vector, when neighboring agents i, j When the electrical equipment of the next level is connected, Otherwise, the value is 0; iteration stops when the results of two consecutive iterations of an agent are the same. The iteration ends when all agents stop iterating. Each agent can determine the numbers of all connected agents by checking the 1 elements in its own state vector.

[0023] Step 2: Obtaining Power System Status Information: The information to be collected includes the status information of all generators, loads, and lines in the system, and based on this information, a state vector is formed for each agent. The vector length is equal to the sum of the number of all power generation equipment, load equipment, and line equipment in the system.

[0024] Iterate , For the first k At +1 iteration, the agent i The state vector, For the first k In the next iteration, the agent j The state vector, where the matrix elements involved in the iteration are... With the structure vector Different, for all and i Neighboring agents j , Without considering physical continuity, otherwise The iteration stops when the results of two consecutive iterations are the same. The iteration ends when all agents stop iterating. Each agent can determine the state information of all agent devices in the system by checking the non-zero elements in its own state vector.

[0025] Step 3: Obtain global information about the power system: Since each agent can only determine the agents within its own partition through its state vector and cannot know the connection states of other agents, a new structure vector is generated for each agent after iterating through the agent structure information. Let agent 1 be a new agent... i -1 represents a zone, and the agent... i To intelligent agents n If it is another region, then the new structure vector is: .

[0026] For each partition, the ID of the agent with the smallest ID in that partition is used as the first element of its structure vector. This structure vector is iterated in the same way as the state vector, and finally, all structure vectors become: .

[0027] When a system failure occurs, the protection devices of the independent power system activate, causing circuit breakers around the faulty area to open and isolate the faulty area. If one busbar loses power, the equipment connected to that busbar automatically switches to another power source via automatic transfer switches before reconfiguration. Afterward, the agent begins information collection and calculation. The agent obtains basic state vectors by monitoring the status of local circuit breakers in real time and through communication between agents, and performs iterative calculations. When the iterative calculations are complete, each agent, in addition to knowing the equipment and structural information of its own region, also obtains global load and partition information for the system and can calculate the power balance of each region.

[0028] When a system failure occurs, the agents obtain global system information through a messaging algorithm. If the agents have not lost communication, each agent possesses the same global information, and the reconstructed structure calculated using the same algorithm should also be identical. Each agent compares its reconstructed calculation results with the global device information table stored within it to determine its own reconstructed operation.

[0029] In a ship's electrical system, the system's most critical survivability is determined by only a few key devices. Therefore, the reconfiguration algorithm can be divided into two parts. The first step considers only the critical devices in the system, rapidly reconfiguring their power supply paths to prioritize restoring power to loads crucial to system survival. The second step then considers other non-critical devices. Different centralized reconfiguration algorithms can be applied to the two stages of reconfiguration. Considering the requirement for computational speed in the first stage, and the relatively simple equipment and system structure involved, a heuristic algorithm can be used. Heuristic algorithms, as a type of reconfiguration algorithm, have advantages such as fast computation speed and clear logic. The general heuristic reconfiguration algorithm studied in this invention, combined with breadth-first search technology, can quickly search for the power supply paths of critical devices. In particular, when the system structure is relatively regular, this heuristic algorithm can also be used in the reconfiguration calculation of the second stage.

[0030] like Figure 3 As shown, the specific heuristic algorithm flow is as follows: (1) Operation of the automatic / manual switch: In marine electrical systems, critical loads are supplied with two power sources via automatic or manual transfer switches. If one busbar loses power, the automatic transfer switch connected to that busbar will automatically activate and switch to the other power source before reconfiguration. During reconfiguration, the switch can also be manually controlled to switch from one power source to the other.

[0031] (2) Faults that the agent can handle on its own: One of the advantages of the MAS architecture proposed in this invention is that the agents possess a certain degree of autonomy. For load faults within some agents, communication with adjacent agents can be avoided, and the fault can be resolved internally. When a fault occurs, the agent in the area where the fault occurs first needs to determine whether it can resolve the fault independently without assistance from other areas. For example, for a low-voltage distribution bus without a backup path, if a fault occurs, the tie switch connected to that bus can be directly disconnected without the need for assistance from other agents.

[0032] (3) Commissioning of the standby generator: When a system failure occurs, the backup generator can be put into operation in a very short time (10s). Considering that if some loads lose power, insufficient backup generators will lead to load shedding, while too many backup generators will result in excessively low single-pole load rates.

[0033] This algorithm uses the following principle to guide the activation of standby generators: When reconfiguration begins, all standby generators (state 0) are temporarily set to active (state 1) for reconfiguration calculation. After reconfiguration, the power balance of each partition is re-examined, and the states of the standby generators within each partition are sequentially reset from 1 to 0, and the average load rate of the generators in that partition is calculated. If the load rate is less than 90%, it indicates that there is no need to activate this standby generator; otherwise, it indicates that this standby generator should be activated, its state is reset to 1, and the calculation for that partition ends.

[0034] (4) Path search for reconfiguration in the negative power balance region: After the agents perform information transmission calculations, each agent knows the global system partitioning and the status and connection methods of power generation and load equipment in each region. Therefore, the power balance of different regions can be calculated. For regions with negative power balance, they are placed in a priority queue of agents according to the magnitude of the power imbalance. Regions with larger power imbalances are placed at the top of the queue. Starting from the first region in the queue, the shortest power supply path is searched to ensure that the power supply region formed by the new path has a positive power margin.

[0035] (5) Search for reconfiguration paths of key equipment: After the power supply area search is completed, if all areas have a positive power balance, no reconfiguration is needed, and existing loads can be restored to power. However, if there are still areas with a negative power balance, it means that some loads cannot be restored to power. In this case, critical equipment should be reconfigured first. All critical loads are sorted by importance to form a load set, with important loads listed first. Assuming that all loads are in a state of 0, the power supply to critical loads can be prioritized. In shipboard electrical systems, the power supply methods for critical loads mainly include the following two: (a) Powered by the main distribution board: Some critical loads are directly connected to the main switchboard of the generator vector. In this case, the power supply path for the critical load is the power supply path for that area. The power supply path for this load was found in the previous step. If the load still cannot be powered according to this path, then it should be disconnected. (b) Powered by a zone distribution board: In this situation, the possible power supply paths for the load are generally limited to a few. For example, in a certain ship's power system, there are two reconfiguration paths for the load on the area switchboard, such as... Figure 4 As shown; there are three reconfiguration paths for the load on the low-voltage distribution board below the area distribution board, as follows: Figure 5 As shown.

[0036] For each load in the load set, search for its power supply path and determine whether supplying power through this path would cause excessive generator load rate on the island containing the load. When a suitable power supply path is found, or it is determined that the load cannot be powered by any path, remove the load from the load set and select the next load for refactoring.

[0037] (6) Reconstruction path search for other devices: For more regular systems, this heuristic algorithm can also be used in the second-stage reconstruction calculation. The rules are the same as above. All loads in the system, besides critical loads, include loads divided into levels 1, 2, and 3. After calculating the reconstruction of each level of load, all lower-level loads should be set to 0. After the reconstruction of that level of load is complete, the lower-level loads are restored to their original state, the region of power imbalance is searched, and the reconstruction begins again.

[0038] The embodiments described above merely illustrate specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A fault reconfiguration method for a fully distributed ship power supply system based on multi-agent systems, characterized in that, The equipment in the ship's power supply system is mapped as intelligent agents, constructing a fully distributed multi-agent architecture. Each intelligent agent only interacts with neighboring intelligent agents. The method includes the following steps: S1, Distributed information acquisition: Each intelligent agent acquires global network topology information, equipment status information, and global partition information through local iterative calculation; S2, Fault isolation and autonomous processing: When a system fault occurs, causing the protection device to isolate the fault area, the intelligent agent in the fault area determines whether it can independently resolve the fault. If so, it autonomously performs the isolation operation without coordinating with intelligent agents in other areas; S3, Two-stage heuristic reconstruction: Based on the acquired global information, a heuristic algorithm is used for reconstruction calculation. In the first stage, the power supply path of critical equipment is prioritized for reconstruction to ensure system survivability. In the second stage, the power supply path of non-critical equipment is reconstructed to maximize the recovery of system load.

2. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 1, characterized in that, The specific steps for obtaining global network topology information in S1 are as follows: Each agent in the system is numbered, and an initial structure vector is constructed for each agent. In this initial structure vector, only the element corresponding to its own number is 1, and the rest are 0. Each agent reads the structure vectors of its neighboring agents and uses the formula... Perform iterative calculations; where, For the first k At +1 iteration, the agent i The structure vector, For the first k In the next iteration, the agent j The structure vector, For elements of the adjacency matrix, when neighboring agents i , j When the electrical equipment of the next level is connected, Otherwise, it is 0; the iteration stops when the results of the two iterations of the agent are the same. After the iteration ends, the agent checks the non-zero elements in its own structure vector and obtains the numbers of all agents connected to it and the global topology.

3. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 2, characterized in that, The specific steps for obtaining device status information in S1 are as follows: Collect the status information of all generators, loads, and lines in the system; construct an initial state vector for each agent; the length of the initial state vector is equal to the sum of the number of all devices in the system; only elements containing the status information of the agent itself are non-zero, and the rest are 0; each agent reads the state vectors of its neighboring agents and uses the formula... Perform iterative calculations; where, For the first k At +1 iteration, the agent i The state vector, For the first k In the next iteration, the agent j The state vector, for all... i Neighboring agents j , The physical on / off state is not considered; otherwise, the value is 0. After the iteration converges, the agent checks the non-zero elements in its own state vector to obtain the state information of all devices in the system.

4. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 3, characterized in that, The specific steps for obtaining global partition information in S1 are as follows: after the structure information iteration is completed, a new partition structure vector is generated for each agent; the system is divided into multiple partitions, and the agents contained in each partition are identified by the ID number of the agent with the smallest ID number in that partition as the first element of its partition structure vector. The partition structure vector is iterated using the same iterative method as the state vector until the partition structure vectors of all agents converge, so that each agent knows the global partition status of the system.

5. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 1, characterized in that, In S2, when a fault occurs on a low-voltage distribution bus without a backup path, the intelligent agent in the area where the bus is located determines that the fault can be resolved independently and directly controls the disconnection of the tie switch connected to the bus.

6. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 1, characterized in that, In step S3, before performing the reconfiguration path search, the following rules guide the commissioning of standby generators: At the start of reconfiguration, all generators in standby state are temporarily set to commissioning state to participate in the reconfiguration calculation; after the reconfiguration calculation is completed, the power balance of each partition is re-examined, and the status of standby generators in each partition is sequentially changed from commissioning to shutdown, and the average load rate of the generators in that partition is calculated; if the average load rate of the generators is less than a preset threshold, it is determined that the standby generator does not need to be commissioned and remains in shutdown state; if it is greater than or equal to the preset threshold, it is determined that the standby generator should be commissioned and its status is restored to commissioning state.

7. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 6, characterized in that, The preset threshold is 90%; in step S3, for regions with negative power balance, a priority queue is established according to the magnitude of the unbalanced power, with regions having larger unbalanced power ranked at the front of the queue; starting from the first region in the queue, the shortest power supply path is searched so that the power supply region formed by the new path has a positive power margin.

8. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 1, characterized in that, The first stage of S3, reconfiguring the power supply path of critical equipment, includes: sorting all critical loads by importance to form a load set; for each load in the load set, sequentially searching for its power supply path and determining whether the generator load rate of the island exceeds the safety limit when powered by the power supply path; when a suitable power supply path is found, or when it is determined that the load cannot be powered by any path, it is removed from the load set, and the next load is selected to continue reconfiguration; wherein, if the critical load is located on the main distribution board, the power supply path of that area is directly adopted.

9. The fault reconfiguration method for a fully distributed ship power supply system based on multi-agent technology according to claim 1, characterized in that, The second stage of S3 for reconfiguring the power supply path of non-critical equipment includes: classifying non-critical loads in the system, excluding critical loads, according to their levels; when calculating the reconfiguration of each level of load, first setting the state of all lower-level loads to 0; after the reconfiguration of the load at that level is completed, restoring the lower-level loads to their original state, re-searching for power imbalance areas, and continuing the reconfiguration calculation for the next level of load.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the fault reconfiguration method for a fully distributed ship power supply system based on multiple agents as described in any one of claims 1 to 9.

Images