Intelligent project collaboration platform for engineering management

Abstract

This invention relates to the field of project management knowledge graph technology, and more particularly to an intelligent project collaboration platform for engineering management. It constructs a supervisory role knowledge graph through a role graph construction module, decomposes real-time power supervision requirements and constructs a real-time task knowledge graph through a task graph construction module, merges the two graphs through a knowledge graph fusion module and constructs a role-task fusion graph, then uses a collaborative task analysis module to obtain the aggregated result using a graph neural network model, and finally obtains a role-task allocation scheme through a role-task allocation module. This invention constructs a knowledge network linking "people" and "tasks" through a knowledge graph, and finally uses a graph neural network model for information aggregation to obtain an allocation scheme that balances optimal personnel capabilities, reasonable task dependencies, and the highest collaborative efficiency. This effectively solves the problem of broken collaborative chains in multi-disciplinary cross-task allocation faced by existing power supervision tasks.

Description

Technical Field , ,

[0001] The present invention relates to the technical field of project management knowledge graphs, and particularly to an intelligent project collaboration platform for engineering management. Background Art

[0002] Electric power supervision is a key quality and safety control link in the process of electric power engineering construction. It refers to that a supervision unit and personnel with professional qualifications supervise and manage the whole life cycle of an electric power engineering project according to relevant national and industrial standards. Different from other engineering fields such as construction engineering and traffic engineering, the tasks of electric power supervision are highly professional and risk-sensitive, and the tasks of electric power engineering are often restricted by the operation and dispatching of the power grid and must be completed within a strict time window.

[0003] However, different from other fields, the tasks of electric power engineering often involve multi-disciplinary collaboration, and there are time-sequence dependencies and information transfer constraints between tasks (for example, the installation of the main transformer requires the joint participation of electrical primary supervision, secondary protection supervision, and civil engineering supervision, and the secondary supervision needs to immediately intervene in the wiring inspection after the installation of the primary equipment). The traditional allocation of electric power supervision tasks depends on manual experience, and there are often problems such as overlapping role responsibilities, mismatches between resources (personnel / equipment) and tasks, resulting in the "breakage of the collaboration chain" of multi-disciplinary cross tasks and making the tasks unable to be completed on time.

[0004] Therefore, there is a need for an intelligent project collaboration platform for engineering management that can solve the collaboration of multi-disciplinary cross tasks in electric power engineering. Summary of the Invention

[0005] Therefore, the present invention provides an intelligent project collaboration platform for engineering management to solve the problem of the breakage of the collaboration chain of multi-disciplinary cross tasks faced by the allocation of electric power supervision tasks in the prior art.

[0006] The present invention provides an intelligent project collaboration platform for engineering management, including: A role graph construction module, configured to construct a supervision role knowledge graph based on electric power supervision personnel information. The supervision role knowledge graph includes role nodes corresponding to each electric power supervision personnel, and the edges between the role nodes include collaboration scores indicating the collaboration ability between two electric power supervision personnel; A task graph construction module, configured to decompose real-time electric power supervision requirements into multiple sub-tasks and construct a real-time task knowledge graph. The real-time task knowledge graph includes task nodes corresponding to each sub-task, and the edges between the task nodes represent the dependency relationship between two sub-tasks; The knowledge graph fusion module is used to merge the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph. The edges connecting role nodes and task nodes in the role-task fusion graph include an adaptation score that represents the degree of compatibility between power supervisors and sub-tasks. The collaborative task analysis module is used to input the role-task fusion graph into a trained graph neural network model and obtain the aggregated output of the graph neural network model. The role-task allocation module is used to obtain a role-task allocation scheme based on the aggregation results.

[0007] In a preferred embodiment: based on information about power supervision personnel, a knowledge graph of supervision roles is constructed, including: Based on information about power supervision personnel, construct role nodes; Based on the information of power supervision personnel, we can obtain the historical number of collaborations, average communication efficiency, spatial distance adaptability, and time window overlap between each pair of power supervision personnel. The collaboration score between each pair of power supervisors is obtained based on the number of historical collaborations, average communication efficiency, spatial distance adaptability, and time window overlap. If the collaboration score exceeds the preset collaboration score threshold, an edge is established between the two role nodes corresponding to that collaboration score, thus obtaining the supervisor role knowledge graph.

[0008] In a preferred embodiment: the node feature encoding of role nodes includes spatiotemporal attributes, qualification attributes, experience attributes, and skill attributes; the node feature encoding of task nodes includes spatiotemporal data, qualification requirements, importance, and technical requirements corresponding to the above four attributes at their respective encoding positions; in the knowledge graph fusion module, the supervisor role knowledge graph and the real-time task knowledge graph are merged to construct a role-task fusion graph, including: Based on the corresponding codes in the node features of the character node and the task node, the spatiotemporal matching degree, qualification matching degree, experience matching degree and skill matching degree between the character node and the task node are obtained respectively. Based on the time-space matching degree, qualification matching degree, experience matching degree, and skill matching degree, the compatibility score between each character node and task node is obtained; If the adaptation score exceeds the preset adaptation score threshold, an edge is established between the role node and the task node corresponding to the adaptation score to obtain the role-task fusion graph.

[0009] In a preferred embodiment: the role-task fusion graph is input into a trained graph neural network model to obtain the aggregated output of the graph neural network model, including: Obtain a node from the character-task fusion graph; If the node is a role node, then the role nodes adjacent to the node are aggregated to the node based on the first weight matrix, and the task nodes adjacent to the node are aggregated to the node based on the second weight matrix; If the node is a task node, then the task nodes adjacent to the node are aggregated to the node based on the third weight matrix, and the role nodes adjacent to the node are aggregated to the node based on the second weight matrix; Output the aggregated node features of each role node as the aggregation result.

[0010] In a preferred embodiment: the aggregation result includes the aggregated node features of each role node; based on the aggregation result, a role-task allocation scheme is obtained, including: Based on the node features of each role node in the aggregation results, the adaptation score of each role node is obtained; The matching scores that exceed the preset matching score threshold are filtered out, and the power supervisors corresponding to the selected matching scores are counted to obtain a candidate list. The optimal role-task allocation scheme is obtained by optimizing the candidate list.

[0011] In a preferred solution: an optimal role-task allocation scheme is obtained by optimizing the candidate list, including: Based on the candidate list, remove the role nodes and corresponding edges in the role-task fusion graph that do not exist in the candidate list to obtain the candidate representation graph; An allocation scheme code is established based on the candidate representation graph, and an optimization algorithm is used to find the optimal role-task allocation scheme based on the allocation scheme code. The fitness calculated by the optimization algorithm includes the sum of the cooperation score and the fitness score corresponding to an allocation scheme code in the candidate representation graph.

[0012] In a preferred scheme: an allocation scheme code is established based on the candidate representation graph, and an optimization algorithm is used to find the optimal role-task allocation scheme based on the allocation scheme code, including: An allocation scheme code is established based on the candidate representation graph, where each bit in the allocation scheme code corresponds to the role-task edge in the candidate representation graph, and the value of each bit is used to indicate whether to allocate the role to the task. Based on the allocation scheme encoding, the sum of the corresponding cooperation score and fitness score in the candidate representation graph is calculated to obtain the first fitness score; Based on the allocation scheme code and the preset evaluation index, the second fitness score is obtained; The fitness is obtained based on the first fitness score and the second fitness score; Based on fitness, an optimization algorithm is used to find the optimal role-task allocation scheme.

[0013] This invention also provides an intelligent project collaboration method for engineering management, comprising: Based on the information of power supervision personnel, a knowledge graph of supervision roles is constructed. The knowledge graph of supervision roles includes role nodes corresponding to each power supervision personnel, and the edges between role nodes include collaboration scores representing the collaboration ability of two power supervision personnel. The real-time power monitoring requirements are decomposed into multiple sub-tasks, and a real-time task knowledge graph is constructed. The real-time task knowledge graph includes task nodes corresponding to each sub-task, and the edges between task nodes represent the dependency relationship between two sub-tasks. Merge the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph. The edges connecting role nodes and task nodes in the role-task fusion graph include an adaptation score that represents the degree of fit between power supervisors and sub-tasks. The role-task fusion graph is input into the trained graph neural network model to obtain the aggregated output of the graph neural network model; Based on the aggregation results, a role-task allocation scheme is obtained.

[0014] The present invention also provides an electronic device, comprising: Memory and processor; The memory is used to store the program, and the processor is used to execute the steps in the intelligent project collaboration method for engineering management described above when the program is executed.

[0015] The present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, enables the implementation of the steps in the above-described intelligent project collaboration method for engineering management.

[0016] The beneficial effects of adopting the above scheme are: This invention provides an intelligent project collaboration platform for engineering management. It constructs a supervisory role knowledge graph based on power supervisor information using a role graph construction module, where edges between role nodes include collaboration scores. A task graph construction module decomposes real-time power supervision requirements into multiple sub-tasks and constructs a real-time task knowledge graph, where edges between task nodes represent dependencies. A knowledge graph fusion module merges the supervisory role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph, where edges connecting role nodes and task nodes include adaptation scores. A collaborative task analysis module inputs the role-task fusion graph into a trained graph neural network model to obtain an aggregation result. Finally, a role-task allocation module determines a role-task allocation scheme based on the aggregation result. This invention first accurately characterizes the collaborative abilities of each supervisor with other supervisors through a supervisor role knowledge graph construction module, forming a knowledge network reflecting the true capabilities of the personnel. Then, it clearly defines the temporal dependencies between sub-tasks through a real-time task knowledge graph, constructing a task network that conforms to engineering logic. Next, it deeply integrates the personnel capability graph with the task requirement graph, accurately quantifying the matching degree between each supervisor and the sub-task through adaptation scores, forming a "person-task" related knowledge network. Finally, it uses a graph neural network model for information aggregation, effectively capturing the collaborative chain relationships and potential risks in multi-professional cross-tasks, avoiding the collaborative breakage problem caused by ignoring the collaborative history between personnel or the temporal constraints between tasks in traditional methods. Ultimately, based on the aggregation results of the graph neural network, it intelligently generates an allocation scheme that takes into account the optimal personnel capabilities, reasonable task dependencies, and the highest collaborative efficiency, effectively solving the problem of multi-professional cross-task collaborative chain breakage faced by power supervision task allocation in existing technologies. Attached Figure Description

[0017] Figure 1 A system architecture diagram of the intelligent project collaboration platform for engineering management provided by the present invention; Figure 2 A flowchart illustrating the intelligent project collaboration method for engineering management provided by this invention; Figure 3 for Figure 2 A detailed step diagram of step S205 is shown below; Figure 4 for Figure 3 The detailed steps of step S303 are shown in the diagram. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 A specific embodiment of the present invention discloses an intelligent project collaboration platform for engineering management, comprising: The role graph construction module 110 is used to construct a supervisory role knowledge graph based on the information of power supervisors. The supervisory role knowledge graph includes role nodes corresponding to each power supervisor, and the edges between role nodes include collaboration scores representing the collaboration ability of two power supervisors. The task graph construction module 120 is used to decompose the real-time power supervision requirements into multiple sub-tasks and construct a real-time task knowledge graph. The real-time task knowledge graph includes task nodes corresponding to each sub-task, and the edges between task nodes represent the dependency relationship between two sub-tasks. The knowledge graph fusion module 130 is used to merge the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph. The edges connecting role nodes and task nodes in the role-task fusion graph include an adaptation score that represents the degree of compatibility between power supervisors and sub-tasks. The collaborative task analysis module 140 is used to input the role-task fusion graph into the trained graph neural network model and obtain the aggregated result output by the graph neural network model. The role-task allocation module 150 is used to obtain a role-task allocation scheme based on the aggregation results.

[0020] Please refer to the following: Figure 2 The present invention also provides an intelligent project collaboration method for engineering management, comprising: S201. Based on the information of power supervision personnel, construct a knowledge graph of supervision roles. The knowledge graph of supervision roles includes role nodes corresponding to each power supervision personnel, and the edges between role nodes include collaboration scores representing the collaboration ability of two power supervision personnel. S202. Decompose the real-time power supervision requirements to obtain multiple sub-tasks, and construct a real-time task knowledge graph. The real-time task knowledge graph includes task nodes corresponding to each sub-task, and the edges between task nodes represent the dependency relationship between two sub-tasks. S203. Merge the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph. The edges connecting role nodes and task nodes in the role-task fusion graph include an adaptation score that represents the degree of compatibility between power supervisors and sub-tasks. S204. Input the role-task fusion graph into the trained graph neural network model to obtain the aggregated result output by the graph neural network model; S205. Based on the aggregation results, obtain the role-task allocation scheme.

[0021] It should be noted that the intelligent project collaboration platform for engineering management provided in the above embodiments is mainly a computer program product, which can realize the technical solutions described in the above methods. The specific implementation principles and advantages of the intelligent project collaboration platform and intelligent project collaboration method for engineering management are reflected in the following embodiments.

[0022] In the above process, the supervisor role knowledge graph, the real-time task knowledge graph, and the role-task fusion graph are all represented by graph data structures. The supervisor role knowledge graph can be maintained on a daily basis, and only the real-time task knowledge graph needs to be built when assigning tasks on a daily basis.

[0023] In the above process, the collaboration score in the supervisory role knowledge graph represents the collaborative ability of two power supervisors. For example, if two power supervisors are geographically far apart or have no overlapping free time, their collaborative ability is obviously poor. In practice, the calculation of the collaboration score can be flexibly specified according to specific needs, and this invention provides a preferred method: In one embodiment, step S201 above, constructing a knowledge graph of supervisor roles based on information about power supervisors, specifically includes: Based on information about power supervision personnel, construct role nodes; Based on the information of power supervision personnel, we can obtain the historical number of collaborations, average communication efficiency, spatial distance adaptability, and time window overlap between each pair of power supervision personnel. The collaboration score between each pair of power supervisors is obtained based on the number of historical collaborations, average communication efficiency, spatial distance adaptability, and time window overlap. If the collaboration score exceeds the preset collaboration score threshold, an edge is established between the two role nodes corresponding to that collaboration score, thus obtaining the supervisor role knowledge graph.

[0024] Specifically, the final collaboration score can be obtained through any existing method, such as weighted summation or polynomial calculation, of historical collaboration frequency, average communication efficiency, spatial distance fit, and time window overlap. The specific data format can be set according to the actual situation. For example, historical collaboration frequency can be a value normalized to [0,1], such as 10 collaborations in the past year being represented as 1.0. Average communication efficiency can be represented as 1.0 if each collaboration lasts less than 30 minutes and 0.2 if it lasts more than 2 hours. Current spatial distance fit can be represented as 1.0 if the current location is less than 5km away and 0.1 if it is more than 20km away. Time window overlap can be represented as 1.0 if the shared available time is greater than 80% and 0.1 if it is less than 20%.

[0025] In this embodiment, the collaboration score comprehensively considers key objective factors affecting personnel collaboration efficiency—quantifying the number of historical collaborations reflects the tacit understanding and trust foundation of past cooperation; quantifying the average communication efficiency measures the timeliness and accuracy of information transmission; quantifying the spatial distance adaptability reflects the on-site collaboration convenience brought about by geographical proximity; and quantifying the overlap of time windows ensures that both parties have shared available working time. Based on these multi-dimensional indicators, an optimal calculation formula is designed to flexibly adapt to the actual needs of different engineering scenarios, ensuring the reliability of the personnel collaboration foundation when allocating tasks from the source.

[0026] The goal of the above process S202, which decomposes real-time power supervision requirements into multiple sub-tasks and constructs a real-time task knowledge graph, is to break down power supervision tasks into assignable sub-tasks and clarify the relationships such as "technical requirements," "time constraints," "qualification requirements," and "collaborative dependencies" for each sub-task. For example, power supervision tasks can be broken down into atomic tasks (such as "site leveling acceptance before main transformer installation," "GIS equipment withstand voltage test monitoring," and "safety briefing for crossing frame erection") by stage (construction preparation, main construction, commissioning and acceptance) and type (safety supervision, quality supervision, progress supervision). Task requirements can be tagged according to the following: Technical requirements: Required professional knowledge (e.g., for "GIS equipment withstand voltage test", familiarity with "Field Acceptance Test Procedure for Gas Insulated Metal Enclosed Switchgear" GB / T 16927.1 is required); Time constraints: planned start / end time, power outage window (e.g., "must be completed between 00:00 and 04:00 on 2025-09-20"), buffer time (e.g., power must be restored 30 minutes after completion); Qualification requirements: Mandatory qualifications (e.g., "High-voltage equipment testing and monitoring" requires "Registered Electrical Engineer (Power Transmission and Power Generation)" qualification), and preferred qualifications (e.g., "Supervisor with experience in UHV projects"). Collaborative dependencies: Precedence tasks must be completed (e.g., "secondary wiring inspection" requires "primary equipment installation completion") and related roles must be involved (e.g., "concealed works acceptance" requires the presence of the construction party, design party, and supervision party).

[0027] Edges between task nodes represent the dependency relationship between two subtasks. This can be achieved by establishing directed edges to represent temporal relationships, or by establishing weighted undirected edges to represent the degree of dependency. The specific design of these edges is based on existing technologies.

[0028] In a more preferred embodiment, the node feature encoding of a role node includes spatiotemporal attributes, qualification attributes, experience attributes, and skill attributes, while the node feature encoding of a task node includes spatiotemporal data, qualification requirements, importance, and technical requirements corresponding to the above four attributes at their respective encoding positions. Specifically: The spatiotemporal attributes of the role node include the current location of the personnel and the available time window; the qualification attributes include the set of certificates held by the supervisor, such as "Registered Electrical Engineer (Power Transmission and Generation), Electrician's Grid Access Permit, etc."; the experience attributes include the set of historical projects, years of work experience, etc.; and the skill attributes include the set of skill tags, such as high-voltage test monitoring, GIS equipment installation and acceptance, etc.

[0029] The spatiotemporal data of the task nodes include equipment location, planned task time, etc.; qualification requirements include mandatory qualification sets, such as "500kV test monitoring requires a registered electrical engineer"; importance includes safety impact level, etc.; and technical requirements include skill tag sets, such as relay protection commissioning requires mastery of IEC 60255 standard.

[0030] In this context, "corresponding in encoding position" refers to the spatiotemporal attributes and spatiotemporal data being in the same position in the encoding (e.g., both being the first element), the qualification attributes and qualification requirements being in the same position, the experience attributes and importance being in the same position, and the skill attributes and technical requirements being in the same position. This embodiment innovatively places the feature encoding of role nodes and task nodes in strictly corresponding encoding positions (e.g., spatiotemporal attributes / data are both in the first dimension, qualification attributes / requirements are both in the second dimension, experience attributes / importance are both in the third dimension, and skill attributes / technical requirements are both in the fourth dimension). This ensures that during the aggregation process of the graph neural network, data with the same meaning in the feature vectors of role nodes and task nodes will be aligned, achieving precise fusion of semantic alignment during the graph neural network aggregation process. This allows for the direct capture of complementary or conflicting relationships between roles and tasks in the same semantic dimension during the feature aggregation stage, significantly improving the graph neural network's ability to model complex constraints (such as the mandatory qualification requirements for high-safety tasks and the precise spatiotemporal matching for emergency tasks). The final allocation scheme can not only meet the stringent requirements of power supervision tasks for professional qualifications, spatiotemporal conditions, and skill levels, but also optimize multi-objective collaboration (such as balancing resource load and task priority) through the adaptive learning of the neural network, significantly enhancing the scientific nature and execution reliability of task allocation.

[0031] Furthermore, based on the above encoding method, step S203, merging the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph, specifically includes: In the knowledge graph fusion module, the supervisor role knowledge graph and the real-time task knowledge graph are merged to construct a role-task fusion graph, including: Based on the corresponding codes in the node features of the character node and the task node, the spatiotemporal matching degree, qualification matching degree, experience matching degree and skill matching degree between the character node and the task node are obtained respectively. Based on the time-space matching degree, qualification matching degree, experience matching degree, and skill matching degree, the compatibility score between each character node and task node is obtained; If the adaptation score exceeds the preset adaptation score threshold, an edge is established between the role node and the task node corresponding to the adaptation score to obtain the role-task fusion graph.

[0032] This embodiment specifically defines the part for calculating the fit score. Specifically, the spatiotemporal matching degree can be obtained through spatiotemporal attributes and spatiotemporal data; the qualification matching degree can be obtained through qualification attributes and qualification requirements; the experience matching degree can be obtained through experience attributes and importance; and the skill matching degree can be obtained through skill attributes and technical requirements. It is understood that the specific calculation process for the above matching degrees can be implemented using any existing method based on the actual situation and the specific encoding method, which is something that those skilled in the art would conceive of, and will not be elaborated upon in this document.

[0033] It is worth noting that, based on the strict correspondence of the encoding mentioned above, since the character node and task node are encoded in the same way, only a direct vectorization comparison between the two is needed (e.g., directly calculating the cosine similarity of the two node features). This allows for the simultaneous decoupling of the results containing the above four matching degrees. This not only avoids the redundant computational overhead and feature misalignment risk caused by multiple rounds of independent calculation of the matching degree of each dimension in traditional methods (e.g., mistakenly associating skill tags with qualification requirements), but also ensures the semantic consistency of the calculation of each matching degree through the alignment of feature positions. This enables the final adaptation score to comprehensively reflect the overall degree of fit between the character and the task requirements in key dimensions such as spatiotemporal conditions, professional qualifications, experience accumulation, and skill level.

[0034] Furthermore, in a preferred embodiment, step S204, inputting the role-task fusion graph into the trained graph neural network model to obtain the aggregated result output by the graph neural network model, specifically includes: Obtain a node from the character-task fusion graph; If the node is a role node, then the role nodes adjacent to the node are aggregated to the node based on the first weight matrix, and the task nodes adjacent to the node are aggregated to the node based on the second weight matrix; If the node is a task node, then the task nodes adjacent to the node are aggregated to the node based on the third weight matrix, and the role nodes adjacent to the node are aggregated to the node based on the second weight matrix; Output the aggregated node features of each role node as the aggregation result.

[0035] The above polymerization process can be represented by the following formula: Aggregation of role nodes:

[0036] Aggregation of task nodes:

[0037] in, Represents a role node. For character nodes Node characteristics, For character nodes The aggregated node features, For the character node The set of adjacent role nodes, For a role node in this set, For character nodes Node characteristics, For the character node The set of adjacent task nodes, Indicates a task node. For task nodes Node characteristics, For task nodes The aggregated node features, To be related to task nodes The set of adjacent task nodes, This is a task node in the set. For task nodes Node characteristics, To be related to task nodes The set of adjacent role nodes, For collaboration score, To adapt the score, , and These are the first weight matrix, the second weight matrix, and the third weight matrix, respectively. and All of them are any existing activation functions.

[0038] This embodiment innovatively designs a differentiated weight matrix aggregation strategy to address the characteristics of heterogeneous nodes (role nodes and task nodes) in the role-task fusion graph. When aggregating role nodes, the first weight matrix focuses on integrating the collaborative capability information of adjacent role nodes, while the second weight matrix integrates the demand characteristics of adjacent task nodes, thereby comprehensively capturing the role's overall capabilities in team collaboration and task adaptation. When aggregating task nodes, the third weight matrix aggregates the temporal and dependency strength information of adjacent task nodes, while the second weight matrix simultaneously absorbs the capability characteristics of adjacent role nodes, thus accurately depicting the overall adaptation status of the task in terms of execution environment and resource matching. This differentiated aggregation method not only solves the problem of coarse modeling of heterogeneous node relationships in traditional homogeneous graph neural networks, but also enables the network to adaptively focus on the key information of different types of edges through targeted weight matrix learning. The final aggregation result reflects both the comprehensive competence of each role in the complex collaborative network and the resource requirements and constraints of each task in the execution chain.

[0039] After aggregation, the aggregated features of each role node are input into a fully connected layer to obtain the final allocation scheme. However, in practice, obtaining the allocation result directly through a graph neural network requires preparing a large amount of training data in advance. On the one hand, the number of power engineering supervisors is limited, often making it impossible to prepare enough samples. On the other hand, it is also difficult to know in advance the optimal personnel allocation scheme under different circumstances. In practice, it is often only possible to obtain a rough range of candidate personnel based on some simple constraints. For example, in different situations, it is more convenient to select personnel who are closest to the site and have available time. Therefore, for feasibility considerations, this invention also provides a preferred embodiment.

[0040] like Figure 3 As shown, in a preferred embodiment, step S205 above, obtaining a role-task allocation scheme based on the aggregation result, specifically includes: S301. Based on the node features of each role node after aggregation in the aggregation results, obtain the adaptation score of each role node; S302. Filter out the adaptation scores that exceed the preset adaptation score threshold, and count the power supervision personnel corresponding to the filtered adaptation scores to obtain the candidate list. S303. Optimize the candidate list to obtain the optimal role-task allocation scheme.

[0041] The calculation process for the aforementioned adaptation score can also be performed using a fully connected layer, or it can be calculated using the suggested linear function (equivalent to a simplified, parameter-scarce fully connected layer). Since the graph neural network in this embodiment only needs to determine the approximate range of personnel, accuracy can be appropriately sacrificed. From another perspective, because this embodiment pre-screens using a graph neural network, solving the "broken collaboration chain" problem in multi-disciplinary cross-tasks, it can greatly simplify the complexity of the subsequent secondary optimization process and further consider other factors (such as addressing issues like "uneven resource load," "time window delays," and "insufficient priority for high-security tasks"), significantly improving accuracy while ensuring efficiency.

[0042] This embodiment first utilizes the powerful feature extraction capabilities of graph neural networks and innovatively adopts a two-stage strategy of "fit score filtering - candidate list optimization". First, the aggregated features are transformed into intuitive fit scores through a fully connected layer, and a candidate list of supervisors with basic competence is selected based on a preset threshold (such as automatically excluding low-fit roles with time and space conflicts or unqualified qualifications), which greatly narrows the scope of optimization. Then, for the high-potential candidates after screening, the optimal allocation scheme is finally determined by combining the actual constraints of the engineering site (such as simple rules that can be quickly verified, such as proximity priority and idle time priority) and business experience knowledge, through methods with low computational cost such as heuristic search, greedy algorithms or lightweight optimization models. This design cleverly avoids the limitations of traditional methods that rely on massive amounts of labeled data to train end-to-end allocation models (in power supervision scenarios, due to the limited number of personnel and the scarcity of historical task data, it is difficult to obtain sufficient samples, and the optimal allocation scheme is often difficult to label in advance). It fully utilizes the advantages of graph neural networks in modeling complex collaborative relationships, and through the flexible mechanism of secondary optimization, it is compatible with the rapid decision-making needs of engineering sites. This significantly improves the feasibility and practicality of allocation schemes in data-scarce scenarios, while ensuring the strict satisfaction of key constraints (such as safety qualifications and spatiotemporal windows). It provides a feasible and scalable intelligent allocation solution for engineering fields such as power supervision that are highly specialized but have limited data.

[0043] It is conceivable that the secondary optimization process can employ any existing method, further incorporating other factors. Figure 4 In a preferred embodiment, step S303, optimizing the candidate list to obtain the optimal role-task allocation scheme, specifically includes: S401. Based on the candidate list, remove the role nodes and corresponding edges in the role-task fusion graph that do not exist in the candidate list to obtain the candidate representation graph. S402. Establish an allocation scheme code based on the candidate representation graph, and optimize the allocation scheme code using an optimization algorithm to obtain the optimal role-task allocation scheme. The fitness calculated by the optimization algorithm includes the sum of the cooperation score and the fit score corresponding to an allocation scheme code in the candidate representation graph.

[0044] This embodiment uses an optimization algorithm (such as a genetic algorithm) to achieve secondary optimization. After selecting a list of highly fit candidates, it innovatively constructs a simplified candidate representation graph by removing redundant nodes and related edges outside the candidate list. This significantly reduces the coding complexity and computational redundancy of subsequent optimization algorithms and improves computational efficiency.

[0045] Most importantly, in the optimization process, this embodiment directly reuses the high-quality collaboration score (reflecting the historical cooperation and information transmission efficiency among the supervision team) and the suitability score (comprehensive person-job matching degree in key dimensions such as spatiotemporal matching, qualification conformity, experience suitability, and skill compatibility) pre-calculated and stored in the role-task fusion graph. By summing the two scores as the evaluation index, the reliability of the collaboration chain and the suitability of the task in multi-professional cross-tasks are quickly and accurately quantified. This design not only avoids the high cost of repeatedly calculating complex collaboration relationships and multi-dimensional matching indicators in the optimization process of traditional methods, but also achieves one-click efficient evaluation of multi-professional collaboration effects by reusing the existing features of the fusion graph, ensuring the real-time response capability and computational economy of the algorithm in large-scale task allocation scenarios.

[0046] Specifically, in a preferred embodiment, step S402, which involves establishing an allocation scheme code based on the candidate representation graph and then optimizing the allocation scheme code using an optimization algorithm to obtain the optimal role-task allocation scheme, specifically includes: An allocation scheme code is established based on the candidate representation graph, where each bit in the allocation scheme code corresponds to the role-task edge in the candidate representation graph, and the value of each bit is used to indicate whether to allocate the role to the task. Based on the allocation scheme encoding, the sum of the corresponding cooperation score and fitness score in the candidate representation graph is calculated to obtain the first fitness score; Based on the allocation scheme code and the preset evaluation index, the second fitness score is obtained; The fitness is obtained based on the first fitness score and the second fitness score; Based on fitness, an optimization algorithm is used to find the optimal role-task allocation scheme.

[0047] This embodiment creatively integrates a dual-dimensional evaluation index at the fitness calculation level. The first fitness score directly reuses the pre-stored collaborative capabilities in the candidate representation graph, accurately quantifying the quality of the task allocation scheme in terms of professional collaboration and basic adaptation. The second fitness score, based on preset evaluation indicators (such as resource load balancing, emergency task priority, and compliance with power outage windows—actual constraints on the engineering site), further captures the performance of the allocation scheme on the global optimization objective. By obtaining the comprehensive fitness through a weighted sum of the two scores, this scheme achieves precise control over both "collaborative efficiency" and "engineering constraints," enabling optimization algorithms (such as genetic algorithms) to quickly converge to the optimal allocation scheme that balances team collaboration efficiency and actual engineering needs. This embodiment avoids redundant evaluation of collaboration and adaptation relationships and reduces computational redundancy through compact coding, significantly improving the optimization efficiency and feasibility of the scheme in large-scale power supervision task allocation scenarios.

[0048] Taking the genetic algorithm as an example, a specific implementation of the above embodiment is as follows: 1. Chromosome coding The role-task edges in the candidate representation graph are represented by unique codes, where 1 indicates that the role is assigned to the task, and 0 indicates that it is not assigned. Multiple individuals are randomly generated to obtain the initial population.

[0049] 2. Fitness Function Design The fitness function integrates multi-objective constraints. After obtaining the sum of collaboration score and fitness score (only the sum of collaboration score and fitness score corresponding to the role nodes with assigned tasks is calculated), as well as three preset evaluation indicators, namely, time window satisfaction (the overlap ratio between task completion time and planning window), resource load balancing degree (the variance of total working hours of the role, the smaller the variance, the higher the value), and emergency task priority (the number of emergency tasks assigned to highly fit roles), the fitness is normalized and weighted summed to obtain the fitness.

[0050] 3. Genetic manipulation Selection: Tournament selection is used to avoid premature elimination of highly fit individuals; Crossover: Two-point crossover (randomly select two crossover points and exchange offspring genes) to retain the superior allocation pattern of the parent generation; Mutation: Uniform mutation (flipping gene loci with probability p) introduces new solutions to avoid local optima; Constraint Repair: If the mutation violates hard constraints such as "only one high-risk task is assigned to the same role at the same time", the task is redistributed through a greedy algorithm (e.g., the conflicting task is assigned to the suboptimal role).

[0051] Based on the candidate list, the optimal task allocation scheme is obtained through multi-objective genetic algorithm optimization, which solves problems such as "uneven resource load", "time window delay" and "insufficient priority of high-security tasks".

[0052] The present invention also provides an electronic device, comprising: Memory and processor; The memory is used to store the program, and the processor is used to execute the steps in any of the above-mentioned intelligent project collaboration methods for engineering management when the program is executed.

[0053] The present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in any of the above-described intelligent project collaboration methods for engineering management.

[0054] This invention provides an intelligent project collaboration platform for engineering management. It constructs a supervisory role knowledge graph based on power supervisor information using a role graph construction module, where edges between role nodes include collaboration scores. A task graph construction module decomposes real-time power supervision requirements into multiple sub-tasks and constructs a real-time task knowledge graph, where edges between task nodes represent dependencies. A knowledge graph fusion module merges the supervisory role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph, where edges connecting role nodes and task nodes include adaptation scores. A collaborative task analysis module inputs the role-task fusion graph into a trained graph neural network model to obtain an aggregation result. Finally, a role-task allocation module determines a role-task allocation scheme based on the aggregation result. This invention first accurately characterizes the collaborative abilities of each supervisor with other supervisors through a supervisor role knowledge graph construction module, forming a knowledge network reflecting the true capabilities of the personnel. Then, it clearly defines the temporal dependencies between sub-tasks through a real-time task knowledge graph, constructing a task network that conforms to engineering logic. Next, it deeply integrates the personnel capability graph with the task requirement graph, accurately quantifying the matching degree between each supervisor and the sub-task through adaptation scores, forming a "person-task" related knowledge network. Finally, it uses a graph neural network model for information aggregation, effectively capturing the collaborative chain relationships and potential risks in multi-professional cross-tasks, avoiding the collaborative breakage problem caused by ignoring the collaborative history between personnel or the temporal constraints between tasks in traditional methods. Ultimately, based on the aggregation results of the graph neural network, it intelligently generates an allocation scheme that takes into account the optimal personnel capabilities, reasonable task dependencies, and the highest collaborative efficiency, effectively solving the problem of multi-professional cross-task collaborative chain breakage faced by power supervision task allocation in existing technologies.

[0055] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0056] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent project collaboration platform for engineering management, characterized in that, include: The role graph construction module is used to construct a knowledge graph of supervisory roles based on the information of power supervisors. The knowledge graph of supervisory roles includes role nodes corresponding to each power supervisor, and the edges between role nodes include collaboration scores representing the collaboration ability of two power supervisors. The task graph construction module is used to decompose the real-time power supervision requirements into multiple sub-tasks and construct a real-time task knowledge graph. The real-time task knowledge graph includes task nodes corresponding to each sub-task, and the edges between task nodes represent the dependency relationship between two sub-tasks. The knowledge graph fusion module is used to merge the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph. The edges connecting role nodes and task nodes in the role-task fusion graph include an adaptation score that represents the degree of compatibility between power supervisors and sub-tasks. The collaborative task analysis module is used to input the role-task fusion graph into a trained graph neural network model and obtain the aggregated output of the graph neural network model. The role-task allocation module is used to obtain a role-task allocation scheme based on the aggregation results.

2. The intelligent project collaboration platform for engineering management according to claim 1, characterized in that, Based on information about power supervision personnel, a knowledge graph of supervision roles is constructed, including: Based on information about power supervision personnel, construct role nodes; Based on the information of power supervision personnel, we can obtain the historical number of collaborations, average communication efficiency, spatial distance adaptability, and time window overlap between each pair of power supervision personnel. The collaboration score between each pair of power supervisors is obtained based on the number of historical collaborations, average communication efficiency, spatial distance adaptability, and time window overlap. If the collaboration score exceeds the preset collaboration score threshold, an edge is established between the two role nodes corresponding to that collaboration score, thus obtaining the supervisor role knowledge graph.

3. The intelligent project collaboration platform for engineering management according to claim 1, characterized in that, The node feature encoding of a character node includes spatiotemporal attributes, qualification attributes, experience attributes, and skill attributes. The node feature encoding of a task node includes spatiotemporal data, qualification requirements, importance, and technical requirements corresponding to the above four attributes at the encoding position. In the knowledge graph fusion module, the supervisor role knowledge graph and the real-time task knowledge graph are merged to construct a role-task fusion graph, including: Based on the corresponding codes in the node features of the character node and the task node, the spatiotemporal matching degree, qualification matching degree, experience matching degree and skill matching degree between the character node and the task node are obtained respectively. Based on the time-space matching degree, qualification matching degree, experience matching degree, and skill matching degree, the compatibility score between each character node and task node is obtained; If the adaptation score exceeds the preset adaptation score threshold, an edge is established between the role node and the task node corresponding to the adaptation score to obtain the role-task fusion graph.

4. The intelligent project collaboration platform for engineering management according to claim 1, characterized in that, The role-task fusion graph is input into a trained graph neural network model to obtain the aggregated output of the graph neural network model, including: Obtain a node from the character-task fusion graph; If the node is a role node, then the role nodes adjacent to the node are aggregated to the node based on the first weight matrix, and the task nodes adjacent to the node are aggregated to the node based on the second weight matrix; If the node is a task node, then the task nodes adjacent to the node are aggregated to the node based on the third weight matrix, and the role nodes adjacent to the node are aggregated to the node based on the second weight matrix; Output the aggregated node features of each role node as the aggregation result.

5. The intelligent project collaboration platform for engineering management according to claim 1, characterized in that, The aggregation result includes the node features of each role node after aggregation; Based on the aggregation results, a role-task allocation scheme is obtained, including: Based on the node features of each role node in the aggregation results, the adaptation score of each role node is obtained; The matching scores that exceed the preset matching score threshold are filtered out, and the power supervisors corresponding to the selected matching scores are counted to obtain a candidate list. The optimal role-task allocation scheme is obtained by optimizing the candidate list.

6. The intelligent project collaboration platform for engineering management according to claim 5, characterized in that, The optimal role-task allocation scheme is obtained by optimizing the candidate list, including: Based on the candidate list, remove the role nodes and corresponding edges in the role-task fusion graph that do not exist in the candidate list to obtain the candidate representation graph; An allocation scheme code is established based on the candidate representation graph, and an optimization algorithm is used to find the optimal role-task allocation scheme based on the allocation scheme code. The fitness calculated by the optimization algorithm includes the sum of the cooperation score and the fitness score corresponding to an allocation scheme code in the candidate representation graph.

7. The intelligent project collaboration platform for engineering management according to claim 6, characterized in that, An allocation scheme code is established based on the candidate representation graph, and an optimization algorithm is used to find the optimal role-task allocation scheme based on the allocation scheme code, including: An allocation scheme code is established based on the candidate representation graph, where each bit in the allocation scheme code corresponds to the role-task edge in the candidate representation graph, and the value of each bit is used to indicate whether to allocate the role to the task. Based on the allocation scheme encoding, the sum of the corresponding cooperation score and fitness score in the candidate representation graph is calculated to obtain the first fitness score; Based on the allocation scheme code and the preset evaluation index, the second fitness score is obtained; The fitness is obtained based on the first fitness score and the second fitness score; Based on fitness, an optimization algorithm is used to find the optimal role-task allocation scheme.

8. A smart project collaboration method for engineering management, characterized in that, include: Based on the information of power supervision personnel, a knowledge graph of supervision roles is constructed. The knowledge graph of supervision roles includes role nodes corresponding to each power supervision personnel, and the edges between role nodes include collaboration scores representing the collaboration ability of two power supervision personnel. The real-time power monitoring requirements are decomposed into multiple sub-tasks, and a real-time task knowledge graph is constructed. The real-time task knowledge graph includes task nodes corresponding to each sub-task, and the edges between task nodes represent the dependency relationship between two sub-tasks. Merge the supervisor role knowledge graph and the real-time task knowledge graph to construct a role-task fusion graph. The edges connecting role nodes and task nodes in the role-task fusion graph include an adaptation score that represents the degree of fit between power supervisors and sub-tasks. The role-task fusion graph is input into the trained graph neural network model to obtain the aggregated output of the graph neural network model; Based on the aggregation results, a role-task allocation scheme is obtained.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store the program, and the processor is used to execute the steps of the intelligent project collaboration method for engineering management in claim 8 when the program is executed.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, enable the implementation of the steps in the intelligent project collaboration method for engineering management as described in claim 8.

Images