Work order intelligent scheduling and resource optimization configuration method and system for energy industry

Abstract

This invention provides a method and system for intelligent work order scheduling and resource optimization in the energy industry, relating to the field of intelligent energy management technology. The method includes parsing work order text to generate work order type feature vectors and work order element sets; constructing a dynamic relationship graph based on an energy resource knowledge graph; recommending the optimal work order scheduling scheme; calculating the fit degree and outputting the optimal resource combination; generating a complete work order configuration scheme; executing the scheme and optimizing it in real time through dynamic monitoring and feedback, outputting the final scheduling execution result. The system includes modules for work order parsing and feature extraction, dynamic relationship graph construction, scheduling scheme recommendation, resource fit optimization, scheduling plan generation, and execution monitoring and feedback optimization. This invention achieves intelligent scheduling and resource optimization throughout the entire work order scheduling process, significantly improving plan achievement rate, reducing time deviation rate, shortening emergency repair response time, and improving resource utilization.

Description

Technical Field

[0001] This invention relates to the field of intelligent energy management technology, and in particular to a method and system for intelligent scheduling of work orders and optimized resource allocation for the energy industry. Background Technology

[0002] As the energy structure shifts towards a higher proportion of renewable energy, the large-scale integration of distributed energy resources such as photovoltaics, energy storage, and controllable loads significantly increases the complexity of operation and management in the energy industry. Virtual power plants, as key platforms for aggregating distributed resources, face increasingly heavy demands for work order scheduling and resource optimization while participating in electricity market transactions and providing ancillary services. Work order scheduling involves operational tasks such as equipment maintenance, fault handling, and energy distribution, while resource optimization requires efficient coordination of distributed resources to improve the economy and reliability of power grid operation.

[0003] Traditional work order scheduling and resource allocation methods often rely on human experience or static rule bases, employing a centralized decision-making mechanism. For example, work order allocation is often based on fixed priorities or historical data, failing to respond in real time to changes in power grid status, resource availability, and market dynamics; resource allocation typically uses predefined strategies, lacking adaptive adjustment capabilities. This traditional approach has significant limitations: as distributed resources expand, centralized scheduling is prone to single-point-of-failure risks, and static rules lead to delays in work order response and low resource utilization. Especially in highly volatile renewable energy scenarios, traditional methods struggle to achieve dynamic matching between work orders and resources. For example, during grid outages, a backlog of work orders may cause delays in the maintenance of critical equipment, while uneven resource allocation exacerbates supply-demand imbalances, affecting the operational efficiency and service quality of virtual power plants. Coal transportation scheduling has long relied on the experience and judgment of veteran employees, lacking scientific coordination, resulting in high transportation costs. In remote oil and gas fields, work order distribution still relies on the old methods of telephone and SMS, which not only easily leads to scheduling conflicts but also makes it impossible to track the progress of on-site operations in real time. This makes it difficult to ensure operational compliance and often causes inefficient internal friction due to repeated communication and information delays, and resource allocation lacks a global perspective.

[0004] In view of this, there is an urgent need for intelligent scheduling and resource optimization methods and systems for work orders in the energy industry, in order to at least address the above-mentioned shortcomings. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent scheduling and resource optimization of work orders in the energy industry, in order to solve the problems of rigid scheduling mechanisms, high risk of centralized architecture, lack of adaptive resource allocation, inefficient information transmission, insufficient global optimization, and low level of intelligence in existing technologies. The specific technical solution is as follows:

[0006] This invention provides a method for intelligent scheduling and resource optimization of work orders in the energy industry, comprising:

[0007] Step 1: Based on the received work order scheduling request data, parse the work order text, extract work order scheduling elements and classify work order types, and generate work order type feature vectors and work order element sets;

[0008] Step 2: Based on the work order type feature vector and work order element set, perform data preprocessing based on the energy resource knowledge graph to construct a dynamic relationship graph containing energy equipment, work order events, and geographical locations, and generate the dynamic relationship graph;

[0009] Step 3: Based on the work order type feature vector, recommend the optimal work order scheduling scheme through a federated learning model to generate a recommended scheduling scheme;

[0010] Step 4: Based on the recommended scheduling scheme and dynamic relationship graph, apply graph neural network to calculate the fit between candidate resources and work order scheduling targets, and output the optimal resource combination;

[0011] Step 5: Based on the optimal resource combination, automatically generate a work order scheduling plan that meets the resource optimization configuration requirements, and output a complete work order configuration scheme;

[0012] Step 6: Execute the complete work order configuration scheme, and optimize it in real time through dynamic monitoring and feedback, and output the final scheduling execution result.

[0013] Furthermore, in step 1:

[0014] The work order scheduling request data includes the work order text, work order number, submission time, and submitting user, which are described in natural language and received from the energy industry production management system.

[0015] The parsing of the work order text includes applying natural language processing technology to perform word segmentation, part-of-speech tagging, named entity recognition, and dependency parsing on the work order text.

[0016] The work order scheduling elements include equipment elements, event elements, resource requirement elements, and space elements;

[0017] The classification of work orders is achieved through a machine learning classification model, which categorizes work orders into predefined business categories, including emergency repairs, planned maintenance, equipment inspections, equipment installations, and energy distribution.

[0018] The work order type feature vector is generated by one-hot encoding of the work order type label or by a neural network model based on the full text of the work order.

[0019] The work order element set is stored in the form of key-value pairs, JSON objects, or database records.

[0020] Furthermore, in step 2:

[0021] The energy resource knowledge graph is a pre-constructed semantic network used to describe entities and their relationships in the energy field. It is stored in a graph structure, where nodes represent entities and edges represent relationships between entities. Entities include energy equipment, personnel, materials, geographical locations, and historical work orders. Relationships include location, belonging to, required skills, use of spare parts, and historical associated work orders.

[0022] The data preprocessing includes the operation of associating and fusing work order type feature vectors and work order element sets with the energy resource knowledge graph, specifically including:

[0023] S201. Match and link the entities mentioned in the work order elements with the corresponding entity nodes in the knowledge graph to confirm their unique identifiers.

[0024] S202. Based on existing relationships in the knowledge graph, automatically complete the implicit information related to the current work order;

[0025] S203. Obtain the current status and personnel status information of relevant device nodes in the knowledge graph from the external monitoring system in real time, update the graph, and form a dynamic context;

[0026] The dynamic relationship graph is a weighted directed or undirected graph temporarily constructed for the current work order scheduling task, and its nodes and edges are dynamically updated as time or work order context changes.

[0027] Furthermore, step 2, constructing the dynamic relationship graph, includes:

[0028] S211. Based on the dynamic information obtained from the energy resource knowledge graph after data preprocessing and the current set of work order elements to be scheduled, initialize the node set of the dynamic relationship graph. The node set consists of resource nodes, work order event nodes and location nodes.

[0029] Among them, resource nodes represent energy sector resource entities that can be scheduled and allocated, including energy equipment, personnel, and materials; each resource node is associated with characteristic attributes, including status attributes, capability attributes, and location attributes.

[0030] The work order event node represents a work order event that is currently to be scheduled. Its features are initialized by the work order type feature vector and fused with key parameters extracted from the work order element set.

[0031] Location nodes represent key geographical locations, derived from geographical location entities in the knowledge graph, and each location node contains at least its geographical coordinate features;

[0032] S212. Based on the business rules and scheduling logic of the energy industry, define and dynamically calculate the relationship between nodes in the graph to form a set of edges of the graph. Each edge is assigned a weight to quantify the strength, cost or matching degree of the relationship.

[0033] Edge types include membership edges, spatial edges, capability matching edges, and temporal edges;

[0034] S213. Output a dynamic relationship graph with the structure G = (V, E), where node V contains feature attributes and edge E contains the weights of the quantized relationship.

[0035] Furthermore, in step 2, updating the dynamic relationship graph is triggered by external state update events, new work order injection events, or scheduling decision feedback events:

[0036] External status update events: Real-time acquisition of equipment status, personnel location and status changes from SCADA system and asset management system, and update the characteristic attributes of corresponding resource nodes and the weights of related spatial relationship edges accordingly;

[0037] New work order injection event: When a new work order enters the scheduling queue, a new work order event node is dynamically added, and its capability matching edge and spatial edge with the existing resource node and location node are immediately created. At the same time, its temporal relationship edge with the existing work order node is checked.

[0038] Scheduling decision feedback event: When a resource is temporarily pre-allocated, its state attributes and the weights of related edges need to be updated immediately.

[0039] Furthermore, in step 3:

[0040] The term "optimal" refers to a multi-objective weighted optimality, with weights set according to the work order type. For emergency repair work orders, the time weight is 0.6, the cost weight is 0.3, and the reliability weight is 0.1; for planned maintenance work orders, the cost weight is 0.5, the time weight is 0.3, and the reliability weight is 0.2.

[0041] Step 3 includes:

[0042] S301. Generate a set of candidate scheduling schemes based on the work order type feature vector and the scheduling strategy knowledge base;

[0043] Among them, the scheduling strategy knowledge base stores historical best scheduling schemes, industry scheduling rules and constraints;

[0044] S302. Construct a federated learning scheduling recommendation model, which consists of a central server and local models deployed locally on each data source. This is achieved through collaborative training across multiple data sources without the data leaving the local machine.

[0045] S303. Calculate the recommendation score of the candidate scheduling scheme through the federated learning scheduling recommendation model;

[0046] S304. Sort the work orders according to the recommended scores and output the optimal work order scheduling scheme;

[0047] S305. Generate a recommended scheduling scheme, which specifies the resources and scheduling instructions to be allocated to the current work order.

[0048] Furthermore, in step 4:

[0049] The adaptability refers to the comprehensive score of resource-work order matching, which is used to quantitatively evaluate the overall cost, benefits and risks of a specific resource executing a target work order under the current system status. The higher the score, the better the adaptability.

[0050] The formula for calculating adaptability is: Adaptability = 0.4 × Capability Matching Degree + 0.3 × Spatial Distance Adaptability + 0.2 × Timing Conflict Degree + 0.1 × Resource Load Rate;

[0051] Among them, capability matching measures the degree of matching between resource capabilities and work order requirements; spatial distance adaptation measures the spatial adaptability between resources and work order execution locations; time sequence conflict measures the degree of time conflict between resources executing the current work order and existing tasks; and resource load rate quantifies the current load saturation of resources.

[0052] Furthermore, step 4 includes:

[0053] S400, Dynamic Relationship Graph Timeliness Verification: Confirm that the status of resource nodes and edge weights in the graph have been updated within the last 5 minutes; if the timeliness is exceeded, trigger the external status update event process in step 2 to re-obtain the latest graph structure;

[0054] S401. Based on the candidate scheduling scheme set in the recommended scheduling scheme and the resource screening conditions of each scheme, match them with the resource nodes in the dynamic relationship diagram, enumerate all the resource-work order pairs that meet the conditions, and generate a candidate resource-work order pair set.

[0055] S402. Construct a fitness calculation model based on graph neural networks. This model takes a dynamic relationship graph and the features of each node as input and adopts a graph neural network architecture.

[0056] S403. Calculate the fitness score for each pair using the fitness calculation model;

[0057] S404. Sort the resources according to their fit scores and output the optimal resource combination.

[0058] S405. Output the optimal resource combination, which is explicitly specified as one or more specific resource entities allocated to the current work order.

[0059] Furthermore, step 5 includes:

[0060] S501. Generate a work order scheduling milestone sequence based on the optimal resource combination and scheduling rule base;

[0061] Among them, the scheduling rule base stores the standard process for work order execution in the energy industry, timing logic, and resource coordination constraints;

[0062] S502. Based on dynamic relationship diagrams, resource coordination and conflict resolution are performed, and an executable scheduling instruction set is generated in detail.

[0063] The complete work order configuration scheme is output using a standardized data exchange format, including XML files or JSON objects. Its data format includes work order number, allocated resources, schedule, operation instructions, and emergency plan.

[0064] Furthermore, step 5, which refines the generation of the executable scheduling instruction set, includes:

[0065] S5021, Timing Orchestration: Assign a suitable specific time window to each operation node in the sequence to ensure that the timing constraints in the rule base are met, and take into account the time required for resource movement;

[0066] S5022, Resource Coordination Instruction Generation: For operations that require the coordination of multiple resources, generate specific coordination instructions;

[0067] S5023. Conflict Detection and Resolution: Check whether the current scheduling plan has resource conflicts or spatiotemporal overlaps with existing task schedules or other work order plans in the diagram. If a conflict exists, adjust it according to the preset strategy to resolve the conflict.

[0068] Furthermore, step 6 includes:

[0069] S601. Generate a full lifecycle monitoring project set based on the complete work order configuration scheme and execution monitoring template;

[0070] Among them, the execution monitoring template is used to generate monitoring items for each execution stage by referring to the work order execution strategy;

[0071] S602. Construct an execution verification network. This network is built based on the expected execution logic and monitoring item set of the complete work order configuration scheme and is used to simulate the expected execution process and ideal state of the work order.

[0072] S603. Obtain real-time execution data through a monitoring agent deployed locally on the resource;

[0073] S604. Perform spatiotemporal alignment and comparative analysis between the execution verification network and real-time execution data;

[0074] S605. Dynamic optimization and feedback adjustment based on comparison results;

[0075] For minor deviations, adjustment instructions are automatically generated and fed back to the execution end for fine-tuning;

[0076] For major deviations or sudden failures, an interruption request is generated, and the current execution status and deviation data are used as feedback inputs, which are then fed back to the corresponding steps according to the source of the deviation.

[0077] S606. Output the final scheduling execution result, which is a structured execution report. The content includes the final actual execution timeline, actual resource consumption, actual execution results and verification data of key operations, and overall execution performance indicators.

[0078] This invention also relates to an intelligent work order scheduling and resource optimization allocation system for the energy industry, comprising:

[0079] The work order parsing and feature extraction module is used to parse the work order text, extract work order scheduling elements and classify work order types based on the received work order scheduling request data, and generate work order type feature vectors and work order element sets.

[0080] The dynamic relationship graph construction module is used to preprocess data based on the energy resource knowledge graph according to the work order type feature vector and work order element set, and to construct a dynamic relationship graph containing energy equipment, work order events and geographical location.

[0081] The scheduling scheme recommendation module is used to recommend the optimal work order scheduling scheme based on the work order type feature vector and through a federated learning model, and generate the recommended scheduling scheme.

[0082] The resource adaptation and optimization module is used to calculate the fit between candidate resources and work order scheduling targets based on recommended scheduling schemes and dynamic relationship graphs, and output the optimal resource combination.

[0083] The scheduling plan generation module is used to automatically generate work order scheduling plans that meet the requirements of resource optimization configuration based on the optimal resource combination, and output a complete work order configuration scheme.

[0084] The execution monitoring and feedback optimization module is used to execute the complete work order configuration scheme, and to perform real-time optimization through dynamic monitoring and feedback, and output the final scheduling execution result.

[0085] Furthermore, the dynamic relationship graph construction module is also used for:

[0086] Update the node feature attributes and edge weights in the dynamic relationship graph based on external state update events, new work order injection events, or scheduling decision feedback events;

[0087] Among them, external status update events trigger the real-time acquisition of equipment status, personnel location and status changes from the SCADA system and asset management system.

[0088] The new work order injection event triggers the dynamic addition of new work order event nodes and the creation of related edges;

[0089] The scheduling decision feedback event triggers updates to resource status attributes and related edge weights.

[0090] Furthermore, in the resource adaptation and optimization module:

[0091] The fitness calculation model adopts a graph neural network architecture, taking a dynamic relationship graph and the features of each node as input, and performs message passing and node feature aggregation through multi-layer graph convolution or attention mechanism to output fitness score.

[0092] The formula for calculating the fit is: Fit = 0.4 × Capability Match + 0.3 × Spatial Distance Fit + 0.2 × Timing Conflict + 0.1 × Resource Load Rate.

[0093] Furthermore, the execution monitoring and feedback optimization module is also used for:

[0094] Generate a full lifecycle monitoring project set based on the execution monitoring template;

[0095] Construct an execution verification network to simulate the expected execution process and ideal state of work orders;

[0096] Obtain real-time execution data through a monitoring agent deployed locally on the resource;

[0097] The execution verification network and real-time execution data are spatiotemporally aligned and compared for analysis.

[0098] Dynamic optimization and feedback adjustments are made based on the comparison results, including fine-tuning instruction generation and major deviation backflow processing.

[0099] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method described herein.

[0100] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method.

[0101] The beneficial effects of this invention are as follows: By integrating natural language processing, energy resource knowledge graphs, federated learning, and graph neural network technologies, it achieves intelligent and automated scheduling throughout the entire work order process. In terms of scheduling efficiency and response speed, it enables automatic parsing and classification of work orders, intelligent recommendation and dynamic optimization of scheduling schemes, shortening emergency repair response time and improving plan achievement rate. In terms of resource allocation and cost control, by constructing a dynamic relationship graph and calculating resource-work order suitability, it achieves precise matching and collaborative optimization of resources (personnel, equipment, materials, geographical location), effectively reducing time deviation rate and scheduling costs, and improving resource utilization. In terms of system reliability and security, it employs federated learning to protect data privacy, utilizes knowledge graphs to complete implicit information, and combines dynamic monitoring and feedback mechanisms to ensure the accuracy and robustness of scheduling decisions, reducing operational risks caused by information lag or single-point failures. Overall, this invention solves the problems of rigid scheduling mechanisms, high risk of centralized architecture, lack of adaptive resource allocation, inefficient information transmission, and insufficient global optimization in traditional methods, providing the energy industry with a highly reliable, efficient, and low-cost intelligent scheduling solution.

[0102] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0103] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0104] Figure 1 This is a schematic diagram of a work order intelligent scheduling and resource optimization allocation method for the energy industry in an embodiment of the present invention;

[0105] Figure 2 This is a schematic diagram of an intelligent work order scheduling and resource optimization allocation system for the energy industry, as described in an embodiment of the present invention. Detailed Implementation

[0106] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0107] This invention provides a method for intelligent scheduling and resource optimization of work orders in the energy industry, such as... Figure 1 As shown, it includes:

[0108] Step 1: Based on the received work order scheduling request data, parse the work order text, extract work order scheduling elements and classify work order types, and generate work order type feature vectors and work order element sets;

[0109] The work order scheduling request data is the original information carrier received from the energy industry production management system, requesting resource scheduling and task allocation. It includes the work order text described in natural language, the work order number, the submission time, and the submitting user. For example, the work order text is: "The 35kV transformer B of substation A has an oil temperature too high alarm and needs to be urgently dispatched for maintenance. It is expected to take 4 hours and requires a high-voltage electrician certificate."

[0110] Parsing the work order text refers to the process of applying natural language processing technology to structure the work order text, including word segmentation, part-of-speech tagging, named entity recognition, and dependency parsing. The parsing process is implemented using open-source tools (such as the Chinese word segmentation tool Jieba, the natural language processing library spaCy, or Stanford CoreNLP) or based on pre-trained models (such as BERT) to identify lexical units, grammatical structures, and entity boundaries in the text.

[0111] Among them, work order scheduling elements refer to the key parameter entities that affect scheduling decisions extracted from the parsed work order text. Work order scheduling elements include equipment elements, event elements, resource demand elements, and space elements.

[0112] Equipment elements include the identification of the energy equipment that has failed or requires operation (e.g., transformer B) and the equipment type (e.g., 35kV transformer);

[0113] Event elements include the specific event type described in the work order (e.g., high oil temperature alarm) and the required operation (e.g., maintenance, urgency level (e.g., emergency);

[0114] Resource requirements include estimated working hours (e.g., 4 hours), required tools (e.g., insulation tester), and required personnel skill certificates (e.g., high-voltage electrician certificate).

[0115] Spatial elements include the geographical location or site where the equipment or event occurs (e.g., substation A).

[0116] The classification of work order types refers to categorizing work orders into predefined business categories based on their content. These work order types are predefined based on energy industry business rules and include "emergency repair," "planned maintenance," "equipment inspection," "equipment installation," and "energy distribution." The classification is achieved through a machine learning classification model. The input of this model is features extracted from the work order text (such as bag-of-words model features, word embedding features, or context embedding features), and the output is the probability distribution of the work order type or a specific category label. The model can be a support vector machine, random forest, or deep learning text classification model (such as TextCNN or BERT), and it is trained under supervision using historical work order data and its annotation types.

[0117] Among them, the work order type feature vector is a numerical representation of the work order type, which is used to quantify and compare work order features in the algorithm. Its generation methods include sparse vectors obtained by one-hot encoding of work order type labels, or dense continuous vectors generated based on the full text content of the work order through a neural network model (such as the sentence embedding model Sentence-BERT).

[0118] The work order element set is a collection of extracted work order scheduling elements organized in the form of structured data. It is usually stored in the form of key-value pairs, JSON objects or database records to facilitate subsequent querying and processing.

[0119] Step 2: Based on the work order type feature vector and work order element set, perform data preprocessing based on the energy resource knowledge graph to construct a dynamic relationship graph containing energy equipment, work order events, and geographical location, and generate the dynamic relationship graph;

[0120] The energy resource knowledge graph is a pre-built semantic network used to describe entities and their relationships within the energy field. It is stored in a graph structure, where nodes represent entities and edges represent relationships between entities. The knowledge graph includes energy equipment (such as transformers, switchgear, and photovoltaic inverters), personnel (such as technicians and work teams), materials (such as spare parts), geographical locations (such as substations and power supply stations), and historical work orders. Relationships between entities include "located in," "belongs to," "requires skills," "uses spare parts," and "historically associated work orders." The construction of the knowledge graph is based on industry standards, equipment ledgers, and organizational charts, and it is stored and queried using knowledge graph construction tools (such as Neo4j and Apache Jena).

[0121] The data preprocessing refers to the operation of associating and fusing the work order type feature vector and work order element set output in step 1 with the energy resource knowledge graph, including:

[0122] S201. Match and link the entities mentioned in the work order elements (such as transformer B, substation A) with the corresponding entity nodes in the knowledge graph to confirm their unique identifiers.

[0123] S202. Based on existing relationships in the knowledge graph, automatically complete implicit information related to the current work order; for example, by using the relationship, find the substation A node where the transformer B node is located, and further find all available technician nodes and their skill attributes at that station.

[0124] S203. Obtain information such as the current status of relevant equipment nodes and personnel status from external monitoring systems (such as SCADA systems and asset management systems) in real time, update the graph, and form a dynamic context.

[0125] The dynamic relationship graph is a weighted directed or undirected graph temporarily constructed for the current work order scheduling task. Its nodes and edges are dynamically updated as time or work order context changes. This graph is the direct input for the subsequent graph neural network (GNN) model to perform node embedding, relationship learning and final scheduling decision calculation.

[0126] The construction of a dynamic relationship graph includes:

[0127] S211. Based on the dynamic information obtained from the energy resource knowledge graph after data preprocessing and the current set of work order elements to be scheduled, initialize the node set of the dynamic relationship graph. The node set consists of resource nodes, work order event nodes and location nodes.

[0128] Resource nodes originate from the energy resource knowledge graph and represent energy sector resource entities that can be scheduled and allocated. Examples include energy equipment, personnel, and materials.

[0129] Each resource node is equipped with feature attributes extracted from the knowledge graph or obtained in real time from an external system. These feature attributes include state attributes, capability attributes, and location attributes.

[0130] Status attributes include the current operating status of the equipment and the current status of the personnel;

[0131] Capability attributes include the skills certificates of technical personnel, the technical parameters of equipment, and the specifications of materials.

[0132] Location attributes include the identifier of the geographical location node where the resource is currently located or belongs.

[0133] A work order event node represents a work order event currently awaiting scheduling. The features of each work order event node are initialized by the work order type feature vector generated in step S1, and are integrated with key parameters extracted from the work order element set, such as required skills (e.g., needing a "high voltage electrician's certificate"), target equipment (e.g., needing to operate "transformer B"), urgency level (e.g., "urgent"), and estimated working hours (e.g., "4 hours").

[0134] Location nodes represent key geographical locations, derived from geographical location entities in the knowledge graph. Examples include substations, power supply stations, and specific addresses where work orders are executed. Each location node contains at least its geographical coordinates (such as latitude and longitude).

[0135] S212. Based on the business rules and scheduling logic of the energy industry, define and dynamically calculate the relationships between nodes in the graph to form a set of edges E. Each edge is assigned a weight to quantify the strength, cost, or matching degree of the relationship.

[0136] Membership edges represent organizational affiliation or static associations between resources. For example, creating an edge from the node "Technician A" to the node "Maintenance Team 1" represents a "belongs to" relationship. The weight of this type of edge is usually a fixed value (e.g., 1.0).

[0137] Spatial relationship edges are used to connect resource nodes with the location nodes required by the work order, or to connect two resource nodes (if coordination is required). Their weight is determined by the spatial distance between them. Preferably, the distance is the actual road network distance. The formula for calculating the weight w_spatial is: w_spatial = distance / max_distance, where distance is the distance between two points, and max_distance is a preset maximum effective distance threshold. The shorter the distance, the smaller the weight value, indicating lower scheduling costs.

[0138] Capability matching edges connect resource nodes (especially personnel nodes) to work order event nodes. Their weights represent the degree to which the resource's capability matches the work order. The matching degree is calculated based on the resource node's capability attributes and the work order event node's requirement parameters. For example, if a work order requires skill S, and the technician holds certificate S, then the matching degree m = 1; otherwise, m = 0. More complex weighted calculations can also be performed. Ultimately, the weight w_capability = m; a higher matching degree corresponds to a larger weight value, indicating better suitability.

[0139] Temporal relationship edges are used to connect two work order event nodes, representing the potential temporal constraints between them when assigned to the same resource. By analyzing the existing task schedule of the resource, if the expected time period for executing work order A overlaps with or the interval between the required time period for work order B is insufficient, an edge is created between nodes A and B. Its weight w_temporal reflects the severity of the conflict (such as the amount of time overlap); the larger the weight value, the more severe the conflict.

[0140] S213. Output a dynamic relationship graph; specifically, the dynamic relationship graph structure is G = (V, E); where nodes V contain rich feature attributes, and edges E contain weights for quantized relationships. This graph structure will be converted into a tensor format supported by a graph neural network framework (such as PyTorchGeometric or DGL) and used as input for the graph neural network computation in step S4 to solve for the optimal resource combination.

[0141] Step 3: Based on the work order type feature vector, recommend the optimal work order scheduling scheme through a federated learning model to generate a recommended scheduling scheme;

[0142] Here, "optimal" refers to multi-objective weighted optimality, with weights set according to the work order type (the weight allocation for different work order types is adapted to the business priorities of the energy industry), as detailed below:

[0143] Emergency repair work orders: Time weight 0.6 (prioritize rapid fault response and reduce power outage losses), cost weight 0.3 (control dispatch costs under the premise of timeliness), reliability weight 0.1 (basic support resource capabilities meet standards);

[0144] Planned maintenance work orders: cost weight 0.5 (prioritize optimizing resource costs in non-urgent scenarios), time weight 0.3 (ensure completion within the planned period), reliability weight 0.2 (enhance maintenance quality and avoid secondary failures);

[0145] The subsequent recommendation score calculation and optimal solution selection for candidate scheduling schemes are all based on the multi-objective weights of the corresponding work order type.

[0146] Step 3: Based on the work order type feature vector, an optimal work order scheduling scheme is recommended using a federated learning model to generate a recommended scheduling scheme, including:

[0147] S301. Generate a set of candidate scheduling schemes based on the work order type feature vector and the scheduling strategy knowledge base;

[0148] The scheduling strategy knowledge base is a knowledge base that stores historical optimal scheduling schemes, industry scheduling rules and constraints, and is built based on energy industry scheduling experience and historical data. For example, the knowledge base contains the rule "For 'emergency repair' work orders, priority should be given to scheduling the technicians who are closest to the fault point and have the corresponding skills", or stores historical cases such as "When the 35kV transformer of substation A fails, the optimal solution is to dispatch technician Zhang San from the maintenance team and call the backup transformer C".

[0149] Generating a candidate scheduling scheme set refers to matching the work order type feature vector of the current work order with the rules and cases in the scheduling strategy knowledge base to deduce one or more feasible preliminary scheduling schemes; for example, for an emergency repair work order with excessive oil temperature, the knowledge base matches three candidate schemes: 1) dispatch Zhang San from the maintenance team 1, 2) dispatch Li Si from the maintenance team 2, and 3) request Wang Wu from the neighboring area for support.

[0150] S302. Construct a federated learning scheduling recommendation model;

[0151] The federated learning scheduling recommendation model is a machine learning model that is collaboratively trained across multiple data sources (such as different regional power bureaus) without the data leaving the local machine. It is used to evaluate and recommend the optimal scheduling scheme. The model consists of a central server and local models deployed at each data source.

[0152] Building this model refers to the central server initializing a global scheduling recommendation model. The input of this model is the work order type feature vector, and the output is the rating or ranking of the scheduling scheme. Specifically, the dense continuous vector generated based on Sentence-BERT in step 1 is used to adapt to the federated learning model's requirement for continuous value input and avoid the sparsity problem of one-hot encoding.

[0153] Distribute the initial model parameters to the local models of each participant;

[0154] Each local model uses local historical work order data (including work order type feature vectors and the ultimately proven optimal scheduling scheme) to train the model and uploads the updated model parameters to the central server;

[0155] The central server aggregates (e.g., weighted average) the parameter updates of all local models to generate a new generation of global models; through multiple iterations, the global model learns cross-domain general scheduling knowledge while protecting the data privacy of all parties.

[0156] S303. Calculate the recommendation score of the candidate scheduling scheme through the federated learning scheduling recommendation model;

[0157] Each scheme in the candidate scheduling scheme set generated by S301, combined with the current work order type feature vector, is input into the trained federated learning scheduling recommendation model; the model will output a recommendation score, which quantifies the effectiveness, cost-effectiveness and reliability of the scheme in historically similar scenarios.

[0158] S304. Sort the work orders according to the recommended scores and output the optimal work order scheduling scheme;

[0159] All candidate scheduling schemes are sorted in descending order according to their recommended scores, and the top 3 schemes with the highest scores are selected to form a candidate scheduling scheme set (including the scheme and its corresponding score) for output. For example, if the model calculates that scheme 1 (dispatch Zhang San) has a score of 95, scheme 2 (dispatch Li Si) has a score of 80, and scheme 3 (request support) has a score of 65, then the output {scheme 1, scheme 2, scheme 3} is the candidate scheduling scheme set.

[0160] S305. Generate a recommended scheduling scheme;

[0161] The output of this step is a structured recommended scheduling scheme, which specifies the resources (such as personnel, equipment, and materials) to be allocated to the current work order and their scheduling instructions (such as departure time and expected route).

[0162] Step 4: Based on the recommended scheduling scheme and dynamic relationship graph, apply graph neural network to calculate the fit between candidate resources and work order scheduling targets, and output the optimal resource combination;

[0163] Among them, the adaptability refers to the comprehensive score of resource-work order matching, which is used to quantitatively evaluate the overall cost, benefits and risks of a specific resource executing a target work order under the current system state. The higher the score, the better the adaptability. The calculation formula is: Adaptability = 0.4 × Capacity Matching Degree + 0.3 × Spatial Distance Adaptability Degree + 0.2 × Timing Conflict Degree + 0.1 × Resource Load Rate. The parameters of each dimension are defined as follows:

[0164] Capability matching degree: Quantifies the degree of matching between resource capabilities and work order requirements. 1 indicates a full match (e.g., the skill certificates held by technicians fully meet the work order requirements and the equipment parameters meet the operating standards), and 0 indicates a mismatch.

[0165] Spatial distance adaptability: Quantifies the spatial adaptability between resources and work order execution locations. The calculation method is "1-(actual distance / max_distance)", where "actual distance" is the road network distance between the current location of the resource and the work order execution location, and "max_distance" is the preset maximum effective scheduling distance (e.g., 10 kilometers, adapting to the distribution characteristics of energy industry sites).

[0166] Time Conflict Degree: Quantifies the degree of time conflict between the execution of the current work order and existing tasks. The calculation method is "1-(conflict time / total working hours)", where "conflict time" is the overlap between the expected execution time of the current work order and the time of the existing tasks of the resource, and "total working hours" is the estimated execution time of the current work order.

[0167] Resource load factor: Quantifies the current load saturation of resources. The calculation method is "1-(allocated working hours / total available working hours)", where "allocated working hours" is the total time spent on tasks undertaken by the resource, and "total available working hours" is the total available working time of the resource during the expected execution period of the work order (e.g., 8 hours / day).

[0168] Step 4: Based on the recommended scheduling scheme and dynamic relationship graph, a graph neural network is applied to calculate the fit between candidate resources and work order scheduling targets, and the optimal resource combination is output, including:

[0169] S400: Dynamic relationship graph timeliness check. Call the update log of the dynamic relationship graph to confirm that the status of resource nodes (personnel idle / busy, equipment available / faulty) and edge weights (spatial distance, capability matching degree) in the graph have been updated within the last 5 minutes. If the timeliness is exceeded, trigger the 'external status update event' process in step 2, re-obtain the latest graph structure and then execute S401.

[0170] S401. Based on the candidate scheduling scheme set in the recommended scheduling scheme and the resource screening conditions of each scheme (such as Scheme 1 requiring 'dispatch of technician Zhang San with a high voltage electrician certificate and within 10 kilometers of the fault point'), match them with the resource nodes (personnel status, location) in the dynamic relationship diagram, enumerate all the resource-work order pairs that meet the conditions, and generate a candidate resource-work order pair set.

[0171] S402: Construct an fitness calculation model based on graph neural networks;

[0172] The fitness calculation model is a machine learning model based on graph neural networks (GNN). It takes a dynamic relationship graph and the features of each node as input, and can comprehensively learn the complex topological relationships and node attributes in the graph, and output a fitness score that represents the degree of matching.

[0173] Specifically, the model is constructed using a classic graph neural network architecture (such as GraphSAGE, GAT, or GIN).

[0174] The model's input layer receives topological data (adjacency matrix) of a dynamic relational graph and feature matrices of all nodes; the model performs message passing and node feature aggregation through multi-layer graph convolution / attention mechanisms, and finally outputs an embedding vector representing the global graph state or a task-aware embedding for a specific node.

[0175] S403: Calculate the fitness score for each pair using the fitness calculation model;

[0176] Each pair in the candidate resource-work order pairing set generated by S401, along with the complete dynamic relationship graph, is input into the fit calculation model. The model calculates through forward propagation and outputs a fit score for each candidate pair. This score quantifies the overall cost, benefit, and risk of selecting this resource to execute the work order under the current system state described by the dynamic relationship graph. The higher the score, the better the fit.

[0177] S404: Sort and output the optimal resource combination based on the fit score;

[0178] All candidate resource-work order pairs are sorted in descending order based on their fit scores, and the resource combination corresponding to the pair with the highest score is selected as the final optimal resource combination output. For example, if the model calculates that the fit score of pair (work order, A) is 92, (work order, B) is 85, and (work order, C) is 78, then the final output is that technician A is the optimal resource.

[0179] S405: Output the optimal resource combination;

[0180] The output of this step is a structured optimal resource combination that explicitly specifies one or more specific resource entities (such as technician ID, vehicle number, equipment ID, and material ID) allocated to the current work order. This combination will be used as input to step S5 to generate the final scheduling plan.

[0181] Step 5: Based on the optimal resource combination, automatically generate a work order scheduling plan that meets the resource optimization configuration requirements, and output a complete work order configuration scheme;

[0182] Based on the optimal resource combination, an automatic work order scheduling plan that meets the resource optimization configuration requirements is generated, and a complete work order configuration scheme is output, including:

[0183] S501. Generate a work order scheduling milestone sequence based on the optimal resource combination and scheduling rule base;

[0184] The scheduling rule base is a knowledge base that stores the standard processes, timing logic, and resource coordination constraints for work order execution in the energy industry. It is set up based on industry operating procedures and historical optimization experience. For example, the rule base contains the rule "For maintenance work orders involving high-voltage equipment, the following series of operations must be executed in sequence: 'power outage isolation,' 'voltage testing and grounding,' 'permitted work,' 'work execution,' 'work termination,' and 'power restoration'", and defines the order and minimum time interval constraints between each operation.

[0185] Generating a work order scheduling milestone sequence refers to matching and instantiating the resource entities specified in the optimal resource combination with the standard process rules corresponding to the current work order type in the scheduling rule base to generate a time-constrained milestone sequence containing each key operation node and its logical order; for example, for the above maintenance work order, the generated sequence is: [T1: execute power outage isolation operation, T2: execute voltage testing and grounding operation,...], where T1 and T2 are relative or absolute time points;

[0186] S502: Based on dynamic relationship graphs, resource coordination and conflict resolution are performed, and a set of executable scheduling instructions is generated in detail.

[0187] The dynamic relationship diagram provides the latest status, spatial distribution, and relationships of resources. The detailed generation of executable scheduling instruction set refers to: based on the work order scheduling milestone sequence, combined with the real-time status and spatial relationships of resources in the dynamic relationship diagram, performing detailed scheduling and conflict checks, and generating instructions that directly drive resource execution.

[0188] The complete work order configuration scheme is output using a standardized data exchange format, such as an Extensible Markup Language (XML) file or a JavaScript Object Notation (JSON) object. Its data schema is predefined and includes the work order number, allocated resources, schedule, operation instructions, and emergency plan.

[0189] Specifically, the work order number is a unique identifier for the work order, associated with the original work order request; the allocated resources are a list of allocated resource entities, including the unique identifier and role of each resource; the schedule is a detailed scheduling timeline, including the start time, end time, and execution location of each operation; the operation instructions are a list of behavioral instructions for specific resources; and the contingency plan is a description of contingency measures for potential risks.

[0190] The complete work order configuration scheme is output to the work order execution system to drive the subsequent work dispatch and execution process.

[0191] Step 6: Execute the complete work order configuration plan, and optimize it in real time through dynamic monitoring and feedback, and output the final scheduling execution result;

[0192] Step 6 involves executing the complete work order configuration scheme, performing real-time optimization through dynamic monitoring and feedback, and outputting the final scheduling execution result, including:

[0193] S601. Generate a full lifecycle monitoring project set based on the complete work order configuration scheme and execution monitoring template;

[0194] The execution monitoring template is designed to generate monitoring items for each execution stage based on the work order execution strategy. It is set up according to energy industry work order execution standards and historical operation and maintenance experience. For example, when manually determining the execution process for a "transformer emergency repair" work order based on operation and maintenance experience, the template requires verifying the following during the planning and preparation stage: operator qualification certificates are valid; safety tools have passed inspection; and backup equipment is in good working order status. During the on-site execution stage, it verifies that the operation complies with safety regulations and procedures, and that key equipment status parameters are within acceptable limits. At the end of the execution stage, it verifies that the equipment has returned to normal operation and that the work order has been completed and archived. Therefore, the execution monitoring template generates monitoring items for the "transformer emergency repair" work order execution strategy, including: monitoring personnel qualifications, tool status, and spare parts availability before execution; monitoring compliance with safety regulations and procedures and equipment status parameters during execution; and monitoring equipment operating status and work order archiving status after execution.

[0195] Generating a full lifecycle monitoring project set refers to matching the complete work order configuration scheme with the corresponding type of monitoring rules in the execution monitoring template, and instantiating a specific set of monitoring and inspection items that covers the entire process before, during, and after work order execution.

[0196] S602. Construct an execution verification network;

[0197] The execution verification network is a digital twin subnet built based on the expected execution logic of the complete work order configuration scheme and the monitoring project set. It is used to simulate the expected execution process and ideal state of the work order, serving as a benchmark for verifying the actual execution data. The execution verification network is connected to the data comparison engine and can push expected state data according to the execution timeline.

[0198] S603. Obtain real-time execution data through a monitoring agent deployed locally on the resource;

[0199] Among them, the monitoring agent is a lightweight software module deployed locally on resource entities such as technicians' mobile terminals, vehicle IoT gateways, equipment monitoring and control devices, and material storage devices (such as spare parts warehouse sensors), and undertakes the tasks of on-site data collection, command reception, and preliminary feedback; obtaining real-time execution data refers to: periodically or event-triggeredly collecting real-time data from the monitoring agent of the relevant resources through the wireless communication network according to the requirements of the monitoring project set, such as: personnel location GPS coordinates, equipment operation records, on-site photos, and equipment status telemetry values;

[0200] S604. Perform spatiotemporal alignment and comparative analysis between the execution verification network and real-time execution data;

[0201] Spatiotemporal alignment refers to establishing a communication link and mapping relationship between the expected execution data of the execution verification network and the actual execution data uploaded by the monitoring agent, both with the same timestamp and spatial location. Comparison analysis refers to inputting the aligned two data streams into the data comparison engine and calculating the deviation for each monitoring item. For example, comparing the time difference between the expected operation time and the actual operation time, the latitude and longitude deviation between the expected operation location and the actual personnel location, and the difference between the expected equipment status parameters and the actual telemetry values.

[0202] S605. Dynamic optimization and feedback adjustment based on comparison results;

[0203] Dynamic optimization and feedback adjustment refers to the following: the data comparison engine judges the data based on a preset deviation threshold. If the deviation is within the allowable range, the verification is passed; if the deviation exceeds the limit, an alarm event is generated and the dynamic optimization process is triggered.

[0204] For minor deviations, adjustment instructions are automatically generated and fed back to the execution end for fine-tuning. For example, if it is detected that the technician has not moved along the optimal path, the path is automatically replanned and pushed to his mobile terminal.

[0205] For significant deviations or sudden failures (such as resource unavailability or technical difficulties encountered on-site), an interruption request is generated, with the current execution status and deviation data as feedback input, and the data is fed back according to the following rules:

[0206] If the deviation is caused by a change in resource status (such as a sudden busyness of personnel): return to step 2, trigger the 'scheduling decision feedback event' to update the dynamic relationship graph, and at the same time invalidate the original candidate scheduling scheme set generated in step 3. After the update, re-execute steps 3, 4 and 5.

[0207] If the deviation is due to an unreasonable scheduling strategy (e.g., the candidate scheme does not cover the backup resources): go back to step 3, reuse the dynamic relationship graph updated in step 2, input the deviation data (e.g., 'the original scheme did not consider backup personnel Wang Wu') as a new feature into the federated learning model, regenerate the candidate scheduling scheme set, and then execute steps 4 and 5.

[0208] S606. Output the final scheduling execution result; After all monitored items have been verified and the work order status reaches "execution completed", the final scheduling execution result is generated. This result is a structured execution report, which includes the final actual execution timeline, actual resource consumption, actual execution results and verification data of key operations, and overall execution performance indicators (such as plan achievement rate and work time deviation rate). This result is output and persistently stored in the work order history database for use as a basis for knowledge base updates and subsequent analysis.

[0209] The working principle and beneficial effects of the above technical solution are as follows: improved plan achievement rate, reduced working time deviation rate, shortened emergency maintenance response time, and improved resource utilization.

[0210] In one embodiment, step 2: Based on the work order type feature vector and work order element set, data preprocessing is performed using an energy resource knowledge graph to construct a dynamic relationship graph containing energy equipment, work order events, and geographical locations; this includes constructing the dynamic relationship graph, which comprises the following steps:

[0211] S211. Based on the dynamic information obtained from the energy resource knowledge graph after data preprocessing and the current set of work order elements to be scheduled, initialize the node set of the dynamic relationship graph. The node set consists of resource nodes, work order event nodes and location nodes.

[0212] Resource nodes originate from the energy resource knowledge graph and represent energy sector resource entities that can be scheduled and allocated. Examples include energy equipment, personnel, and materials.

[0213] Each resource node is equipped with feature attributes extracted from the knowledge graph or obtained in real time from an external system. These feature attributes include state attributes, capability attributes, and location attributes.

[0214] Status attributes include the current operating status of the equipment and the current status of the personnel;

[0215] Capability attributes include the skills certificates of technical personnel, the technical parameters of equipment, and the specifications of materials.

[0216] Location attributes include the identifier of the geographical location node where the resource is currently located or belongs.

[0217] A work order event node represents a work order event currently awaiting scheduling. The features of each work order event node are initialized by the work order type feature vector generated in step S1, and are integrated with key parameters extracted from the work order element set, such as required skills (e.g., needing a "high voltage electrician's certificate"), target equipment (e.g., needing to operate "transformer B"), urgency level (e.g., "urgent"), and estimated working hours (e.g., "4 hours").

[0218] Location nodes represent key geographical locations, derived from geographical location entities in the knowledge graph. Examples include substations, power supply stations, and specific addresses where work orders are executed. Each location node contains at least its geographical coordinates (such as latitude and longitude).

[0219] S212. Based on the business rules and scheduling logic of the energy industry, define and dynamically calculate the relationships between nodes in the graph to form a set of edges E. Each edge is assigned a weight to quantify the strength, cost, or matching degree of the relationship.

[0220] Membership edges represent organizational affiliation or static associations between resources. For example, creating an edge from the node "Technician A" to the node "Maintenance Team 1" represents a "belongs to" relationship. The weight of this type of edge is usually a fixed value (e.g., 1.0).

[0221] Spatial relationship edges are used to connect resource nodes with the location nodes required by the work order, or to connect two resource nodes (if coordination is required). Their weight is determined by the spatial distance between them. Preferably, the distance is the actual road network distance. The formula for calculating the weight w_spatial is: w_spatial = distance / max_distance, where distance is the distance between two points, and max_distance is a preset maximum effective distance threshold. The shorter the preset maximum effective distance, the smaller the weight value, indicating a lower scheduling cost.

[0222] Capability matching edges connect resource nodes (especially personnel nodes) to work order event nodes. Their weights represent the degree to which the resource's capability matches the work order. The matching degree is calculated based on the resource node's capability attributes and the work order event node's requirement parameters. For example, if a work order requires skill S, and the technician holds certificate S, then the matching degree m = 1; otherwise, m = 0. More complex weighted calculations can also be performed. Ultimately, the weight wcapability = m; a higher matching degree corresponds to a larger weight value, indicating better suitability.

[0223] Temporal relationship edges are used to connect two work order event nodes, representing the potential temporal constraints between them when assigned to the same resource. By analyzing the existing task schedule of the resource, if the expected time period for executing work order A overlaps with or the interval between the required time period for work order B is insufficient, an edge is created between nodes A and B. Its weight, wtemporal, reflects the severity of the conflict (such as the amount of time overlap); the larger the weight value, the more severe the conflict.

[0224] S213, Output dynamic relationship diagram;

[0225] Specifically, the dynamic relational graph structure is G = (V, E); where nodes V contain rich feature attributes, and edges E contain weights that quantify the relations. This graph structure will be converted into a tensor format supported by a graph neural network framework (such as PyTorchGeometric or DGL) and used as input for the graph neural network computation in step S4 to solve for the optimal resource combination.

[0226] The beneficial effects of the above technical solution are as follows: by using structured nodes and quantified edge weights, the relationship between resources, work orders, and locations in the energy field can be accurately depicted, providing real and complete topological and feature data support for subsequent graph neural network calculations of resource-work order fit, avoiding matching deviations caused by ambiguous entity relationships, and adapting to the complex needs of multi-entity collaborative scheduling in the energy industry.

[0227] In one embodiment, step 2: Based on the work order type feature vector and work order element set, perform data preprocessing based on the energy resource knowledge graph to construct a dynamic relationship graph containing energy equipment, work order events, and geographical locations; this includes updating the dynamic relationship graph, which is triggered by external state updates, new work order injections, or scheduling decision feedback events.

[0228] External status update events: Real-time acquisition of equipment status, personnel location and status changes from SCADA system and asset management system, and update the characteristic attributes of corresponding resource nodes and the weights of related spatial relationship edges accordingly.

[0229] New work order injection event: When a new work order enters the scheduling queue, a new work order event node is dynamically added, and its capability matching edge and spatial edge with the existing resource node and location node are immediately created. At the same time, its temporal relationship edge with the existing work order node is checked.

[0230] Scheduling decision feedback event: When a resource is temporarily pre-allocated, its status attributes (such as changing from "idle" to "assigned") and the weights of related edges need to be updated immediately for the graph neural network to perform the next round of calculation.

[0231] The beneficial effects of the above technical solution are: ensuring that the dynamic relationship graph can reflect the latest status of the system in real time (such as resource idle / busy, new work order demand, location change), avoiding invalid adaptation results in subsequent graph neural network calculations due to data lag, significantly improving the timeliness and accuracy of scheduling decisions, especially suitable for scenarios in the energy industry where equipment status changes frequently and emergency work orders are frequently issued.

[0232] In one embodiment, step 5: Based on the optimal resource combination, automatically generate a work order scheduling plan that meets the resource optimization configuration requirements, and output a complete work order configuration scheme;

[0233] The dynamic relationship diagram provides the latest status of resources, the spatial coordinates of location nodes (such as substation A, spare parts warehouse), and their relationships. Refining and generating an executable scheduling instruction set refers to performing refined scheduling and conflict checks based on the work order scheduling milestone sequence, combined with the real-time status and spatial relationships of resources in the dynamic relationship diagram, and generating instructions that directly drive resource execution. The refined generation of the executable scheduling instruction set includes the following steps:

[0234] S5021, Timing Orchestration: Assign a suitable specific time window to each operation node in the sequence to ensure that the timing constraints in the rule base are met, and take into account the time required for resource movement (calculated based on the spatial relationship in the graph);

[0235] S5022, Resource Coordination Instruction Generation: For operations that require the coordination of multiple resources, generate specific coordination instructions; for example: generate the instruction "Vehicle V arrives at location L1 at time T to pick up and drop off technician T at location L2";

[0236] S5023. Conflict Detection and Resolution: Check whether the current scheduling plan has resource conflicts or spatiotemporal overlaps with existing task schedules or other work order plans in the graph (i.e., the weight of the sequential relationship edge is greater than the threshold). If a conflict exists, adjust it according to the preset strategy (such as priority rules) to resolve the conflict.

[0237] The working principle and beneficial effects of the above technical solution are as follows: it transforms the abstract milestone sequence into highly implementable execution instructions, which not only ensures that the operation sequence complies with the safety regulations of the energy industry (such as the sequence of high-voltage equipment maintenance), but also avoids resource competition through conflict resolution, and realizes multi-resource collaboration, effectively reducing the work order execution delay rate and improving resource allocation efficiency and scheduling plan executability.

[0238] This invention provides an intelligent work order scheduling and resource optimization allocation system for the energy industry, such as... Figure 2 As shown, it includes:

[0239] The work order parsing and feature extraction module is used to parse the work order text, extract work order scheduling elements and classify work order types based on the received work order scheduling request data, and generate work order type feature vectors and work order element sets.

[0240] The dynamic relationship graph construction module is used to preprocess data based on the energy resource knowledge graph according to the work order type feature vector and work order element set, and to construct a dynamic relationship graph containing energy equipment, work order events and geographical location.

[0241] The scheduling scheme recommendation module is used to recommend the optimal work order scheduling scheme based on the work order type feature vector and through a federated learning model, and generate the recommended scheduling scheme.

[0242] The resource adaptation and optimization module is used to calculate the fit between candidate resources and work order scheduling targets based on recommended scheduling schemes and dynamic relationship graphs, and output the optimal resource combination.

[0243] The scheduling plan generation module is used to automatically generate work order scheduling plans that meet the requirements of resource optimization configuration based on the optimal resource combination, and output a complete work order configuration scheme.

[0244] The execution monitoring and feedback optimization module is used to execute the complete work order configuration scheme, and to perform real-time optimization through dynamic monitoring and feedback, and output the final scheduling execution result.

[0245] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for intelligent scheduling and resource optimization of work orders in the energy industry, characterized in that: include: Step 1: Based on the received work order scheduling request data, parse the work order text, extract work order scheduling elements and classify work order types, and generate work order type feature vectors and work order element sets; Step 2: Based on the work order type feature vector and work order element set, perform data preprocessing based on the energy resource knowledge graph to construct a dynamic relationship graph containing energy equipment, work order events, and geographical locations; Step 2's construction of the dynamic relationship graph includes the following steps: S211. Based on the dynamic information obtained from the energy resource knowledge graph after data preprocessing and the current set of work order elements to be scheduled, initialize the node set of the dynamic relationship graph. The node set consists of resource nodes, work order event nodes and location nodes. S212. Based on the business rules and scheduling logic of the energy industry, define and dynamically calculate the relationships between nodes in the graph to form a set of edges. Each edge is assigned a weight to quantify the strength, cost, or matching degree of the relationship. Edge types include membership edges, spatial edges, capacity matching edges, and temporal edges. S213. Output a dynamic relationship graph, where nodes contain feature attributes and edges contain weights for quantified relationships. Step 3: Based on the work order type feature vector, recommend the optimal work order scheduling scheme through a federated learning model to generate a recommended scheduling scheme; Step 4: Based on the recommended scheduling scheme and dynamic relationship graph, apply graph neural network to calculate the fit between candidate resources and work order scheduling target, and output the optimal resource combination; Step 5: Based on the optimal resource combination, automatically generate a work order scheduling plan that meets the resource optimization configuration requirements, and output a complete work order configuration scheme; Step 6: Execute the complete work order configuration scheme and optimize it in real time through dynamic monitoring and feedback, and output the final scheduling execution result; Real-time optimization through dynamic monitoring and feedback includes: S601, generating a full lifecycle monitoring project set based on the complete work order configuration scheme and execution monitoring template; S602. Construct an execution verification network. This network is built based on the expected execution logic and monitoring item set of the complete work order configuration scheme and is used to simulate the expected execution process and ideal state of the work order. S603. Obtain real-time execution data through a monitoring agent deployed locally on the resource; S604. Perform spatiotemporal alignment and comparative analysis between the execution verification network and real-time execution data; S605. Dynamic optimization and feedback adjustment based on comparison results, including: For deviations originating from changes in resource status, the current execution status and deviation data are used as feedback inputs, flowing back to step 2 to trigger the update of the dynamic relationship graph, and steps 3, 4 and 5 are re-executed. If the deviation is caused by an unreasonable scheduling strategy, the deviation data is used as a new feature input, and the process is fed back to step 3. After regenerating the recommended scheduling scheme, steps 4 and 5 are executed again. S606, Output the final scheduling execution result.

2. The intelligent scheduling and resource optimization allocation method for work orders in the energy industry as described in claim 1, characterized in that, The work order scheduling elements mentioned in step 1 include equipment elements, event elements, resource demand elements, and spatial elements; the classification of work order types is achieved through a machine learning classification model, which divides work orders into predefined business categories, including emergency repairs, planned maintenance, equipment inspections, equipment installations, and energy distribution; the work order type feature vector is generated by one-hot encoding of the work order type label or by a neural network model based on the full text of the work order.

3. The intelligent scheduling and resource optimization allocation method for work orders in the energy industry as described in claim 1, characterized in that, The energy resource knowledge graph mentioned in step 2 is a pre-constructed semantic network used to describe entities and their relationships in the energy field. It is stored in a graph structure, where nodes represent entities and edges represent relationships between entities. Entities include energy equipment, personnel, materials, geographical locations, and historical work orders. Relationships include location, belonging, required skills, use of spare parts, and historical associated work orders. The data preprocessing includes: S201. Match and link the entities mentioned in the work order elements with the corresponding entity nodes in the knowledge graph to confirm the unique identifier; S202. Based on existing relationships in the knowledge graph, automatically complete the implicit information related to the current work order; S203. Obtain the current status and personnel status information of relevant device nodes in the knowledge graph from the external monitoring system in real time, update the graph, and form a dynamic context.

4. The intelligent scheduling and resource optimization allocation method for work orders in the energy industry as described in claim 1, characterized in that, In step 2, the dynamic relationship graph is updated by external state update events, new work order injection events, or scheduling decision feedback events.

5. The intelligent scheduling and resource optimization allocation method for work orders in the energy industry as described in claim 1, characterized in that, Step 3 also includes: S301, generating a set of candidate scheduling schemes based on the work order type feature vector and the scheduling strategy knowledge base; S302. Construct a federated learning scheduling recommendation model, which consists of a central server and local models deployed locally on each data source. This is achieved through collaborative training across multiple data sources without the data leaving the local machine. S303. Calculate the recommendation score of the candidate scheduling scheme through the federated learning scheduling recommendation model; S304. Sort the work orders according to the recommended scores and output the optimal work order scheduling scheme; S305. Generate a recommended scheduling scheme.

6. The intelligent scheduling and resource optimization allocation method for work orders in the energy industry as described in claim 1, characterized in that, Step 5 also includes: S501, generating a work order scheduling milestone sequence based on the optimal resource combination and scheduling rule base; S502. Based on dynamic relationship graphs, resource coordination and conflict resolution are performed, and an executable scheduling instruction set is generated in detail.

7. A system for the intelligent scheduling and resource optimization allocation method for work orders in the energy industry as described in any one of claims 1-6, characterized in that, The system includes: The work order parsing and feature extraction module is used to parse the work order text, extract work order scheduling elements and classify work order types based on the received work order scheduling request data, and generate work order type feature vectors and work order element sets. The dynamic relationship graph construction module is used to preprocess data based on the energy resource knowledge graph according to the work order type feature vector and work order element set, and to construct a dynamic relationship graph containing energy equipment, work order events and geographical location. The scheduling scheme recommendation module is used to recommend the optimal work order scheduling scheme based on the work order type feature vector and through a federated learning model, and generate the recommended scheduling scheme. The resource adaptation and optimization module is used to calculate the fit between candidate resources and work order scheduling targets based on recommended scheduling schemes and dynamic relationship graphs, and output the optimal resource combination. The scheduling plan generation module is used to automatically generate work order scheduling plans that meet the requirements of resource optimization configuration based on the optimal resource combination, and output a complete work order configuration scheme. The execution monitoring and feedback optimization module is used to execute the complete work order configuration scheme, and to perform real-time optimization through dynamic monitoring and feedback, and output the final scheduling execution result.

8. The intelligent work order scheduling and resource optimization allocation system for the energy industry as described in claim 7, characterized in that, The dynamic relationship graph construction module is also used to update the node feature attributes and edge weights in the dynamic relationship graph based on external state update events, new work order injection events, or scheduling decision feedback events. Among them, external state update events trigger the real-time acquisition of changes in equipment status, personnel location, and status from the SCADA system and asset management system; new work order injection events trigger the dynamic addition of new work order event nodes and the creation of related edges; scheduling decision feedback events trigger the updating of resource status attributes and related edge weights. In the resource adaptation and optimization module: The fitness calculation model adopts a graph neural network architecture, taking a dynamic relationship graph and the features of each node as input, and performs message passing and node feature aggregation through multi-layer graph convolution or attention mechanism to output fitness score. The formula for calculating the fit is: Fit = 0.4 × Capability Match + 0.3 × Spatial Distance Fit + 0.2 × Timing Conflict + 0.1 × Resource Load Rate.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-6.

Images