Generator set dynamic intelligent maintenance path planning method and system

Abstract

The invention discloses a dynamic intelligent maintenance path planning method and system for a generator set, relates to the technical field of intelligent maintenance of power generation equipment, and solves the problems of insufficient dynamic quantification of potential safety hazards and low response efficiency of sudden working conditions in conventional maintenance planning of the generator set. The method comprises the steps that a constrained directed graph model is constructed based on a three-dimensional topological relation of equipment, nodes represent a to-be-overhauled part, and edges represent a feasible moving path; a temperature risk sensing mechanism is introduced, and the safety coefficient in the path edge weight is dynamically adjusted; and designing a hybrid optimization algorithm to solve the maintenance sequence with the lowest total cost. Project verification shows that the high-temperature operation risk can be effectively reduced, the construction period is shortened, and the operation and maintenance intelligence level of the generator set is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent maintenance of power generation equipment, and in particular to a method and system for dynamic intelligent maintenance path planning of a generator set. Background Art

[0002] In power generation equipment maintenance, traditional methods, due to their overreliance on static planning and manual experience, are unable to meet the stringent safety and efficiency requirements of highly complex operating environments. Key flaws include: a lack of dynamic quantification of safety risks. When critical parts suddenly heat up, fixed-safety path planning cannot accurately model risks based on real-time conditions and may still lead personnel closer to high-temperature areas, posing serious safety hazards. The lack of scientific quantification tools leads to a high reliance on subjective experience in maintenance paths, resulting in frequent rework due to illogical assembly and disassembly sequences, significantly reducing operational efficiency. Dynamic adaptability is severely limited, requiring static paths to be completely recalculated in response to sudden changes in operating conditions, resulting in delayed response and untimely adjustments, making them incapable of supporting high-risk, real-time operations. Furthermore, multi-objective collaborative optimization capabilities are weak, with existing intelligent algorithms focusing on single objectives and failing to integrate safety, resource accessibility, and time efficiency into a unified framework, limiting their practicality. These systemic bottlenecks urgently require breakthroughs through real-time risk perception, dynamic weight adjustment, and multi-objective collaborative optimization technologies to comprehensively improve the inherent safety and economic benefits of maintenance operations. Summary of the Invention

[0003] To address the issues of insufficient dynamic quantification of safety hazards and inefficient response to emergencies in traditional generator set maintenance planning, the present invention provides a method for dynamic intelligent maintenance path planning for generator sets, including:

[0004] S1: Construct a 3D constraint graph model of the generator set maintenance path. This model maps equipment components into nodes, defines a set of node attributes, and generates a set of directed edges based on the physical reachability between components to form a constrained maintenance path network.

[0005] S2: Based on the three-dimensional constraint graph model of S1, a temperature risk perception mechanism is introduced. When the temperature exceeds the safety threshold, the comprehensive weight is adaptively adjusted. When the node temperature data is invalid, it switches to the static weight mode.

[0006] S3: Based on the comprehensive weight of S2, a genetic algorithm is used to generate the initial path population, and Dijkstra local search is periodically embedded to optimize the optimal individual, outputting the maintenance sequence with the lowest total cost and the Pareto solution set;

[0007] S4: Continuously monitors the temperature data stream of the S1 node. When the temperature rise exceeds the threshold, S2 is triggered to adaptively adjust the comprehensive weight and S3 is linked to re-solve the path.

[0008] Furthermore, in S1, the node attribute set is obtained by:

[0009]

[0010] obtained, is a set of node attributes, is a unique identifier of a component, is a standard maintenance duration, is a real-time temperature, is a three-dimensional spatial coordinate.

[0011] Further, in S1, generating a directed edge set specifically includes:

[0012] S11: Constructing an occlusion matrix based on the three-dimensional topological relationship of the equipment ;

[0013] S12: Set the maximum allowable movement distance to ensure the movement feasibility when the maintenance personnel carry tools;

[0014] S13: Generate a directed edge set, which satisfies:

[0015]

[0016] wherein, represents a directed edge belongs to a feasible path set , is a space occlusion marker, is the Euclidean distance between equipment i and j.

[0017] Further, in S2, the comprehensive weight is obtained by:

[0018]

[0019] wherein, is a time item dynamic weight coefficient, is a temperature item dynamic weight coefficient, is a distance item dynamic weight coefficient, satisfying + + =1; is a standard maintenance duration; is a node temperature state of equipment i, is a node temperature state of equipment j; is the rated safe temperature of the equipment material; is the average moving speed of the maintenance personnel;

[0020] When the temperature exceeds the safety threshold, the adaptive adjustment of the comprehensive weight is specifically obtained by:

[0021]

[0022]

[0023] Implementations in which, is a default value, is a risk sensitivity factor.

[0024] Further, in S3, the genetic algorithm specifically includes:

[0025] S31: Adopting partial matching crossover to maintain process constraints;

[0026] S32: Randomly exchanging two points to mutate;

[0027] S33: After every 5 iterations, selecting the top 5% individuals by fitness to perform Dijkstra local search.

[0028] A dynamic intelligent maintenance path planning system for a generator set is also provided, which is implemented based on the foregoing method and includes:

[0029] A three-dimensional modeling module for constructing a three-dimensional constraint graph model of the maintenance path of the generator set, mapping equipment components as nodes, defining a set of node attributes, and generating a set of directed edges based on the physical accessibility between components to form a constrained maintenance path network;

[0030] A risk weight dynamic calculation module for introducing a temperature risk perception mechanism on the three-dimensional constraint graph model of S1, and adaptively adjusting the comprehensive weight when the temperature exceeds the safety threshold;

[0031] A hybrid path optimization module for generating an initial path population using a genetic algorithm based on the comprehensive weight of S2, periodically embedding Dijkstra local search to optimize the optimal individual, and outputting the maintenance sequence with the lowest total cost and the Pareto solution set;

[0032] A dynamic monitoring and response module for continuously monitoring the temperature data stream of S1 nodes, triggering adaptive adjustment of the comprehensive weight of S2 when the temperature rise exceeds the threshold, and coordinating S3 to re-solve the path.

[0033] The beneficial effects of the present application are:

[0034] 1. By fusing real-time perception and multi-objective optimization strategies, intelligent dynamic planning of the maintenance path of the generator set is achieved, significantly improving the safety of operations and emergency response capabilities. The system can actively avoid high-risk areas and achieve second-level path reconstruction in the event of an emergency, effectively reducing the risk of personnel exposure and enhancing the intrinsic safety of maintenance operations.

[0035] 2. Optimizing the sequencing and resource scheduling of maintenance tasks reduces unnecessary operator switching and tool changes, significantly shortening the overall maintenance cycle. The system supports frequent path weight updates, rapidly generating multi-objective equilibrium solutions that balance work hours, resources, and path safety, improving the overall efficiency and stability of overhaul operations.

[0036] 3. The system boasts excellent adaptability for engineering deployments, supports edge computing architectures, and reduces deployment costs. It also offers modular scalability and can integrate predictive maintenance features, enabling proactive intervention in equipment status. This significant reduction in unplanned downtime and the continued reduction in maintenance costs demonstrate the significant economic benefits and promotional value of this invention in practical engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Flow chart of the method of the present invention;

[0038] Figure 2 This is a flow chart of the system architecture of the present invention. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is further described below with reference to the embodiments, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention shall be included in the scope of protection of the present invention. The process equipment or devices not specifically noted in the following examples are all conventional equipment or devices in the art. Unless otherwise specified, the raw materials used in the examples of the present invention can be obtained commercially; unless otherwise specified, the technical means used in the examples of the present invention are all conventional means well known to those skilled in the art.

[0040] Example 1, combined Figure 1 This embodiment describes a method for planning a dynamic intelligent maintenance path for a generator set, including: S1: constructing a three-dimensional constraint graph model of the generator set maintenance path, mapping equipment components into nodes, defining a node attribute set, and generating a directed edge set based on physical reachability between components to form a constrained maintenance path network;

[0041] S2: Based on the three-dimensional constraint graph model of S1, a temperature risk perception mechanism is introduced. When the temperature exceeds the safety threshold, the comprehensive weight is adaptively adjusted. When the node temperature data is invalid, it switches to the static weight mode.

[0042] S3: Based on the comprehensive weight of S2, a genetic algorithm is used to generate the initial path population, and Dijkstra local search is periodically embedded to optimize the optimal individual, outputting the maintenance sequence with the lowest total cost and the Pareto solution set;

[0043] S4: Continuously monitors the temperature data stream of the S1 node. When the temperature rise exceeds the threshold, S2 is triggered to adaptively adjust the comprehensive weight and S3 is linked to re-solve the path.

[0044] Specifically, through Figure 1 It can be seen that the method described in the present invention first constructs a three-dimensional directed graph model based on physical constraints, and real-time temperature monitoring triggers the dynamic update or maintenance of the weight matrix. Then, the initial population of the genetic algorithm is generated and the core optimization operation is performed. When local convergence is detected, the top 5% elite individuals are selected and embedded in the Dijkstra local optimization to realize algorithm hybrid enhancement. Finally, the Pareto optimal solution set is output through dual convergence judgment, realizing a complete technical chain from real-time data perception, multi-algorithm collaborative optimization to intelligent decision-making output.

[0045] In S1, the node attribute set is:

[0046]

[0047] Get, its, is the node attribute set, Is the unique identifier of the component, For standard maintenance time, is the real-time temperature, is the three-dimensional space coordinate.

[0048] In S1, generating a directed edge set specifically includes:

[0049] S11: Construct an occlusion matrix based on the 3D topological relationship of the device ;

[0050] S12: Set the maximum allowable moving distance , ensuring the feasibility of movement of maintenance personnel when carrying tools;

[0051] S13: Generate a set of directed edges that satisfy:

[0052]

[0053] in, Represents a directed edge Belongs to the set of feasible paths , is the spatial occlusion marker, is the Euclidean distance between devices i and j.

[0054] Specifically, S1 implements structured path network modeling by disassembling the generator set into maintenance units and constructing an attributed graph model. Nodes contain attributes such as identification, maintenance time, temperature, and spatial coordinates. Directed edges are constrained by physical reachability and occlusion relationships, and maximum movement distances are set. This forms a constrained three-dimensional path graph, providing topological support for subsequent dynamic optimization.

[0055] In S2, the comprehensive weight is obtained by:

[0056]

[0057] Obtain, among which, is the dynamic weight coefficient of the time term, is the dynamic weight coefficient of the temperature term, is the dynamic weight coefficient of the distance term, satisfying + + =1; This is the standard maintenance time; is the node temperature state of device i, is the node temperature state of device j; The rated safety temperature of the equipment material; is the average moving speed of maintenance personnel;

[0058] When the temperature exceeds the safety threshold, the comprehensive weight is adaptively adjusted through:

[0059]

[0060]

[0061] Implementation, where is the default value, is a risk-sensitive factor.

[0062] Specifically, S2 introduces time, temperature, and distance factors to build a dynamic weighted model and assign comprehensive weights to path edges. The weight coefficient is adaptively adjusted as the working conditions change. When the temperature exceeds the threshold, the risk-sensitive adjustment mechanism is triggered to improve the ability to identify high-temperature areas. If the temperature data fails, the system automatically switches to static weight mode to ensure the continuity and robustness of path planning. In static weight mode, .

[0063] In S3, the genetic algorithm specifically includes:

[0064] S31: using partial matching cross-keeping process constraints;

[0065] S32: random two-point position swap mutation;

[0066] S33: After every 5 generations, the top 5% individuals in fitness are selected to perform Dijkstra local search.

[0067] Specifically, the optimal maintenance path is solved based on a genetic algorithm, with fitness measured by the total weight of the path. PMX cross-process constraints are introduced, combined with random mutation to enhance solution diversity, and Dijkstra local optimization is periodically embedded to improve convergence efficiency. Ultimately, the minimum-weight path and a multi-objective Pareto solution set are output, achieving an efficient fusion of global search and local refinement.

[0068] A dynamic intelligent maintenance path planning system for a generator set is also provided, which is implemented based on the above method and includes:

[0069] The 3D modeling module is used to construct a 3D constraint graph model of the generator set maintenance path. This module maps equipment components into nodes, defines a set of node attributes, and generates a set of directed edges based on the physical reachability between components to form a constrained maintenance path network.

[0070] The risk weight dynamic calculation module is used to introduce a temperature risk perception mechanism into the three-dimensional constraint graph model of S1, and adaptively adjust the comprehensive weight when the temperature exceeds the safety threshold;

[0071] The hybrid path optimization module is used to generate the initial path population based on the comprehensive weight of S2, periodically embed Dijkstra local search to optimize the optimal individual, and output the maintenance sequence with the lowest total cost and the Pareto solution set;

[0072] The dynamic monitoring and response module is used to continuously monitor the temperature data stream of the S1 node. When the temperature rise exceeds the threshold, it triggers S2 to adaptively adjust the comprehensive weight and links S3 to re-solve the path.

[0073] Example 2, combined with Figure 2 This embodiment is described by Figure 2 It can be seen that in this solution, the real-time temperature and position data of the generator set are collected by sensors to the data perception layer, and transmitted to the edge computing layer through the OPC-UA protocol to execute the method described in the present invention, that is, dynamic weight modeling (updating the adjacency matrix) and hybrid optimization solution (outputting the optimal path). The cloud decision layer stores historical maintenance data, optimizes the adaptive parameters of the weight coefficient, integrates the historical data analysis results and feeds back parameter self-optimization instructions to the edge layer. Finally, the AR smart terminal executes the maintenance path navigation.

Claims

1. A method for planning a dynamic intelligent maintenance path for a generator set, characterized in that: include: S1: Construct a 3D constraint graph model of the generator set maintenance path. This model maps equipment components into nodes, defines a set of node attributes, and generates a set of directed edges based on the physical reachability between components to form a constrained maintenance path network. S2: Based on the three-dimensional constraint graph model of S1, a temperature risk perception mechanism is introduced. When the temperature exceeds the safety threshold, the comprehensive weight is adaptively adjusted. When the node temperature data is invalid, it switches to the static weight mode. S3: Based on the comprehensive weight of S2, a genetic algorithm is used to generate the initial path population, and Dijkstra local search is periodically embedded to optimize the optimal individual, outputting the maintenance sequence with the lowest total cost and the Pareto solution set; S4: Continuously monitors the temperature data stream of the S1 node. When the temperature rise exceeds the threshold, S2 is triggered to adaptively adjust the comprehensive weight and S3 is linked to re-solve the path.

2. A method for dynamic intelligent maintenance path planning of a generator set according to claim 1, characterized in that: In S1, the node attribute set is: get, That, is the node attribute set, Is the unique identifier of the component, For standard maintenance time, is the real-time temperature, is the three-dimensional space coordinate.

3. A method for dynamic intelligent maintenance path planning of a generator set according to claim 1, characterized in that: In S1, generating a directed edge set specifically includes: S11: Construct an occlusion matrix based on the 3D topological relationship of the device ; S12: Set the maximum allowable moving distance , ensuring the feasibility of movement of maintenance personnel when carrying tools; S13: Generate a set of directed edges that satisfy: in, Represents a directed edge Belongs to the set of feasible paths , is the spatial occlusion marker, is the Euclidean distance between devices i and j.

4. A method for planning a dynamic intelligent maintenance path for a generator set according to claim 1, characterized in that: In S2, the comprehensive weight is obtained by: Obtain, among which, is the dynamic weight coefficient of the time term, is the dynamic weight coefficient of the temperature term, is the dynamic weight coefficient of the distance term, satisfying + + =1; This is the standard maintenance time; is the node temperature state of device i, is the node temperature state of device j; The rated safety temperature of the equipment material; is the average moving speed of maintenance personnel; When the temperature exceeds the safety threshold, the comprehensive weight is adaptively adjusted through: Implementation, where is the default value, is a risk-sensitive factor.

5. A method for planning a dynamic intelligent maintenance path for a generator set according to claim 1, characterized in that: In S3, the genetic algorithm specifically includes: S31: using partial matching cross-keeping process constraints; S32: random two-point position swap mutation; S33: After every 5 generations, the top 5% individuals in fitness are selected to perform Dijkstra local search.

6. A dynamic intelligent maintenance path planning system for a generator set, implemented based on the method described in any one of claims 1 to 5, characterized in that: include: The 3D modeling module is used to construct a 3D constraint graph model of the generator set maintenance path. This module maps equipment components into nodes, defines a set of node attributes, and generates a set of directed edges based on the physical reachability between components to form a constrained maintenance path network. The risk weight dynamic calculation module is used to introduce a temperature risk perception mechanism into the three-dimensional constraint graph model of S1, and adaptively adjust the comprehensive weight when the temperature exceeds the safety threshold; The hybrid path optimization module is used to generate the initial path population based on the comprehensive weight of S2, periodically embed Dijkstra local search to optimize the optimal individual, and output the maintenance sequence with the lowest total cost and the Pareto solution set; The dynamic monitoring and response module is used to continuously monitor the temperature data stream of the S1 node. When the temperature rise exceeds the threshold, it triggers S2 to adaptively adjust the comprehensive weight and links S3 to re-solve the path.

Images