Power grid intelligent work order distribution method and system

Abstract

The invention belongs to the technical field of work order distribution, and particularly relates to a power grid intelligent work order distribution method and system, and the system comprises a work order convergence and feature extraction module which is used for constructing a unified work order pool and extracting comprehensive feature vectors, such as time and space, business, skill and topological association, of work orders; the mobile terminal state monitoring module is used for acquiring the position and the state of the terminal in real time and calculating the dynamic load rate and the predicted serviceable time of the terminal; the intelligent work order packaging module is used for aggregating the scattered work orders into work order packets conforming to geographical proximity, business association and timeliness overlapping based on a multi-constraint clustering algorithm; the dynamic matching and distribution engine generates an optimal assignment scheme between the work order packet and the mobile terminal by establishing a multi-objective optimization model, and issues the optimal assignment scheme based on a hybrid trigger mechanism; and the execution monitoring and closed-loop learning module is used for progress monitoring, risk early warning and driving system dynamic re-planning and parameter self-optimization to form an intelligent decision closed loop.

Description

Technical Field

[0001] This invention belongs to the field of work order dispatching technology, specifically a method and system for intelligent work order dispatching in power grids. Background Technology

[0002] The statements in this section only refer to the background technology related to this invention and do not necessarily constitute prior art.

[0003] In power systems, various operational activities, such as production operation and maintenance, marketing services, and equipment repair, are processed using work orders as the core carrier. However, different professional systems (such as distribution automation systems, electricity inspection systems, and metering management systems) dispatch work orders independently, typically through a unified task management center for data aggregation. Work orders involving multiple related specialties within the same transformer area or line (such as a distribution transformer inspection work order and a user meter fault work order for the same transformer area) are often assigned to different personnel, leading to repetitive back-and-forth work, low utilization of human resources, and increased operating costs.

[0004] At the same time, existing work order dispatch decisions rely on the subjective experience of managers and lack a quantitative decision support system. They cannot accurately match the multidimensional needs of work orders (such as skill qualification requirements, estimated working hours, urgency, and geographical coordinates) with the attribute characteristics of staff (such as skill level, current location, workload, and historical performance), easily leading to problems such as "skill mismatch" and "load imbalance," affecting the efficiency and quality of work order processing.

[0005] Therefore, the existing work order dispatch technology has inherent defects in terms of coordination, accuracy, risk control and dynamic adaptability, and can hardly meet the higher requirements of modern power grids for resource allocation efficiency, emergency response speed and lean management level. Summary of the Invention

[0006] This invention provides a method and system for intelligent work order dispatching in the power grid, constructing a dynamic and adaptive intelligent work order dispatching system to solve the problems of resource waste, extensive matching, lack of risk and rigid model in the existing work order dispatching mode of the power grid, and ultimately achieving efficient allocation of human resources, precise control of business risks and agile response to uncertainties on site.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention discloses a smart power grid work order dispatch system, comprising: The work order aggregation and feature extraction module is configured to: access work order data from multiple professional systems, build a unified work order pool, and extract the spatiotemporal attributes, business attributes, resource and skill requirements, and topological association attributes of the work orders as feature vectors; The mobile terminal status monitoring module is configured to: obtain the GPS location and working status identifier of each mobile terminal, and calculate the dynamic load rate and estimated service time based on the incomplete work orders bound to it; The work order intelligent packaging module is configured to: aggregate scattered work orders with similar feature vectors into work order packages that conform to geographical proximity, business relevance, and timeliness overlap, based on a multi-constraint clustering algorithm; The dynamic matching and dispatch module is configured to: take work order packages and available mobile terminals as inputs, establish a multi-objective optimization model, generate the optimal dispatch plan by comprehensively optimizing travel time, load balancing and skill matching, and send it to the terminal based on a hybrid event and periodic triggering mechanism; The execution monitoring and closed-loop learning module is configured to: monitor the execution progress of issued work orders, issue delay warnings, and use the feedback data to trigger dynamic replanning and optimize model parameters.

[0008] Furthermore, the spatiotemporal attributes include the latitude and longitude coordinates of the work order location, the expected operation time window, and the timeliness requirements; Business attributes include work order type, work order priority, and the power supply line or transformer area to which it belongs; Resource and skill requirements are transformed into computable feature vectors using one-hot encoding or word embedding techniques. Topology association attributes are based on power grid topology data to identify related business work orders on the same power supply line or in the same transformer area.

[0009] Furthermore, in the mobile terminal status monitoring module, the formula for calculating the dynamic load rate is: Load factor = Σ(remaining working hours × e^(-λ × risk coefficient)); Wherein, λ is the preset attenuation coefficient, and the risk coefficient is the delay risk value calculated based on the deviation of the work order execution progress.

[0010] Furthermore, the intelligent work order packaging module uses an aggregate utility model to evaluate work order packages, as shown in the following formula: U=α×C_geo+β×C_bus+γ×C_time; Where U is the aggregate utility score, C_geo is the geographic concentration degree representing the degree of spatial clustering of each work order in the work order package, C_bus is the business relevance degree quantified based on the predefined business rule base, C_time is the timeliness compactness representing the degree of overlap and continuity of the planned execution time windows of each work order in the work order package, and α, β, and γ are configurable weight coefficients, satisfying α+β+γ=1.

[0011] Furthermore, the multi-objective optimization model in the work order package-personnel dynamic matching and dispatch module is as follows: the objective function is to minimize the total cost Z, Min Z = W1 * C_travel + W2 * C_workload + W3 * C_skill + W4 * C_priority, where C_travel is the travel time, C_workload is the workload saturation variance, C_skill is the skill matching gap, C_priority is the work order priority default risk, and W1-W4 are weight coefficients.

[0012] Furthermore, the dynamic matching and dispatching module adopts a high-performance hybrid solution strategy to generate assignment schemes: based on work order priority and timeliness rules, an initial feasible solution is generated through a greedy insertion method; starting from the initial solution, iterative optimization is performed through multiple rounds of destruction-repair loops. Each round of loops includes a destruction operation to remove some work orders from the current solution and a repair operation to re-insert the removed work orders to generate a new solution; the optimization process must meet the constraints of skill matching, working time capacity, and unique allocation of work orders, and the search direction is to minimize the objective function Z value.

[0013] A second aspect of the present invention discloses a method for intelligent power grid work order dispatching, comprising the following steps: Acquire work order data from multiple professional systems, construct a unified work order pool, and extract the spatiotemporal attributes, business attributes, resource and skill requirements, and topological association attributes of the work orders as feature vectors; Obtain the GPS location and working status identifier of each mobile terminal, and calculate the dynamic load rate and estimated service time based on the incomplete work orders bound to them; Based on the multi-constraint clustering algorithm, scattered work orders with similar feature vectors are aggregated into work order packages that meet the criteria of geographical proximity, business relevance, and overlapping timeliness. Using work order packages and available mobile terminals as input, a multi-objective optimization model is established. The optimal assignment scheme is generated by comprehensively optimizing travel time, load balancing and skill matching, and then sent to the terminal based on a hybrid event and period triggering mechanism. Monitor the execution progress of issued work orders, issue delay warnings, and use the feedback data to trigger dynamic replanning and optimize model parameters.

[0014] A third aspect of the present invention discloses a computer program product including computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the aforementioned smart power grid work order dispatch method.

[0015] A fourth aspect of the present invention discloses an electronic device, including at least one processor and a memory connected to the processor, the memory being used to store a computer program; the processor being used to execute the computer program, enabling the electronic device to implement the above-described smart grid work order dispatching method.

[0016] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects: 1. By constructing a unified work order pool and performing comprehensive feature extraction, the "information silos" between various professional systems are fundamentally broken down. By introducing topological association attributes and work order packaging, multiple scattered work orders from the same region, the same route, or with business logic connections can be automatically identified and aggregated into a "work order package." This avoids redundant work trips caused by independent work assignments from different professions, achieving "one-time work, multiple task processing," significantly reducing the waste of manpower and vehicle resources, improving human resource utilization, and directly lowering on-site operation costs.

[0017] 2. By utilizing a mobile terminal status monitoring module and a multi-objective optimization model, the dispatching decision-making process shifts from relying on subjective experience to data-driven scientific calculation. By comprehensively optimizing multiple objectives such as travel time, workload balancing, skill matching gaps, and priority, the system can match the most suitable mobile terminal to each work order package, effectively solving the problems of "skill mismatch" and "load imbalance" in the traditional model. This not only ensures that work orders are processed by terminals with the appropriate qualifications and capabilities, improving operational safety and quality, but also shortens the overall processing time of work orders and improves response speed through optimal path planning and balanced allocation.

[0018] 3. By employing a rolling time-domain optimization mechanism and a hybrid triggering mechanism (event-driven + periodic scanning), the system acquires dynamic adjustment capabilities. When unexpected events such as sudden work order insertion failures, execution timeouts, or abnormal terminal states occur, the system can respond rapidly, triggering re-optimization immediately or in the next cycle to adjust the dispatch plan in real time. This ensures that the dispatch strategy always adapts to dynamically changing on-site conditions, overcoming the shortcomings of traditional static optimization models that quickly fail when faced with changes, and greatly enhancing the system's anti-interference capabilities and applicability in complex real-world environments.

[0019] 4. Through the execution monitoring and closed-loop learning modules, the system achieves proactive supervision of the entire dispatch and execution process. Furthermore, the proactive delay warning function transforms management from passive response to proactive intervention, further reducing the risk of work order delays and comprehensively improving the standardization of business management and risk control capabilities.

[0020] 5. A complete intelligent closed loop of "perception-decision-execution-learning" has been constructed. The system uses the actual execution results of each dispatch plan (such as actual working hours, path, and completion quality) as feedback data to continuously optimize the working hour prediction model, multi-objective weight coefficients, and clustering algorithm parameters. This self-learning mechanism enables the system to continuously learn from historical data. As the operating time increases, its predictions become more accurate and its decisions become better, forming an organic entity with autonomous learning and continuous improvement capabilities, ensuring the system's advanced nature and high efficiency in long-term operation. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 A schematic diagram of the smart power grid work order dispatch process provided in one or more embodiments of the present invention; Figure 2 A schematic diagram of the work order aggregation and comprehensive feature extraction process provided for one or more embodiments of the present invention; Figure 3 A schematic diagram of the process for constructing a dynamic profile of a person provided in one or more embodiments of the present invention; Figure 4 This is a schematic diagram of the dynamic matching and dispatch execution process provided for one or more embodiments of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] As described in the background section, currently, systems such as distribution automation, electricity consumption information collection, marketing management, and metering management typically operate independently, each with its own work order generation and dispatch channels. Even if some systems achieve simple data aggregation through a unified task center, this only superficially lists work order information and fails to achieve deep collaboration at the business logic level. This "information silo" model results in the inability to effectively identify and integrate multi-disciplinary related work orders (e.g., distribution transformer inspection work orders and batch meter failure work orders for users under the same distribution area) for the same transformer area, line, or user. These work orders are often assigned to different staff for processing. As a result, manual operations are highly fragmented, with staff repeatedly traveling back and forth to the same area, leading to a serious waste of human resources and transportation, and persistently high operating costs.

[0026] Meanwhile, the existing work dispatching decision-making process relies heavily on the manual experience of dispatchers or team leaders, lacking a quantitative and scientific decision support system. The system cannot comprehensively consider the multi-dimensional attributes of work orders (such as required skills and qualifications, estimated working hours, urgency, and precise geographical coordinates) and the dynamic characteristics of staff (such as skill level, real-time location, current workload, and historical performance) to perform accurate matching calculations. This crude work dispatching model is highly prone to "skill mismatch" (assigning employees without the corresponding qualifications) and "load imbalance" (some employees have a backlog of tasks while others are idle), directly affecting work order processing efficiency, first-time resolution rate, and final service quality.

[0027] Furthermore, the existing model lacks real-time monitoring and proactive risk intervention capabilities for the entire work order dispatch and execution process. Firstly, there is a lack of effective monitoring methods to ensure the workload is evenly distributed among staff, making it difficult to avoid uneven workloads and impacting employee morale and fairness. Secondly, the dispatch process lacks automated compliance verification mechanisms, failing to effectively prevent safety hazards such as unqualified operations and working beyond one's scope. Moreover, from a management perspective, the lack of transparent rules and constraints creates potential corruption risks due to specific individuals being responsible for high-value or high-authority work orders on a long-term, fixed basis, and there is a lack of technical preventative measures.

[0028] Furthermore, most research and practice on intelligent work order dispatching focuses on optimal path planning in static scenarios. Once the model parameters are set, the output decision scheme remains fixed. However, the power grid operating environment is highly uncertain, with frequent dynamic events such as the insertion of sudden fault work orders, timeouts in existing work order execution, and abnormal worker conditions. Such events can quickly render the optimal solution based on static assumptions ineffective, preventing the system from responding quickly and adjusting dynamically, exhibiting severe rigidity and lag, and insufficient robustness.

[0029] Therefore, this solution provides a method and system for intelligent work order dispatching in the power grid. Through a closed-loop architecture of "perception-decision-execution-learning", it integrates multi-source data to construct dynamic profiles of work orders and personnel, uses rolling optimization and multi-constraint clustering to package and accurately match work orders, and continuously optimizes based on real-time feedback to achieve dynamic dispatching and continuous evolution of the system.

[0030] A smart power grid work order dispatch system includes: The work order aggregation and feature extraction module is configured to: access work order data from multiple professional systems, build a unified work order pool, and extract the spatiotemporal attributes, business attributes, resource and skill requirements, and topological association attributes of the work orders as feature vectors; The mobile terminal status monitoring module is configured to: obtain the GPS location and working status identifier of each mobile terminal, and calculate the dynamic load rate and estimated service time based on the incomplete work orders bound to it; The work order intelligent packaging module is configured to: aggregate scattered work orders with similar feature vectors into work order packages that conform to geographical proximity, business relevance, and timeliness overlap, based on a multi-constraint clustering algorithm; The dynamic matching and dispatch module is configured to: take work order packages and available mobile terminals as inputs, establish a multi-objective optimization model, generate the optimal dispatch plan by comprehensively optimizing travel time, load balancing and skill matching, and send it to the terminal based on a hybrid event and periodic triggering mechanism; The execution monitoring and closed-loop learning module is configured to: monitor the execution progress of issued work orders, issue delay warnings, and use the feedback data to trigger dynamic replanning and optimize model parameters.

[0031] Correspondingly, such as Figure 1 As shown, a smart work order dispatching method for power grids includes the following steps: Acquire work order data from multiple professional systems, construct a unified work order pool, and extract the spatiotemporal attributes, business attributes, resource and skill requirements, and topological association attributes of the work orders as feature vectors; Obtain the GPS location and working status identifier of each mobile terminal, and calculate the dynamic load rate and estimated service time based on the incomplete work orders bound to them; Based on the multi-constraint clustering algorithm, scattered work orders with similar feature vectors are aggregated into work order packages that meet the criteria of geographical proximity, business relevance, and overlapping timeliness. Using work order packages and available mobile terminals as input, a multi-objective optimization model is established. The optimal assignment scheme is generated by comprehensively optimizing travel time, load balancing and skill matching, and then sent to the terminal based on a hybrid event and period triggering mechanism. Monitor the execution progress of issued work orders, issue delay warnings, and use the feedback data to trigger dynamic replanning and optimize model parameters.

[0032] The specific process of this plan is described in detail below.

[0033] Step S101: Work order aggregation and comprehensive feature extraction based on stream computing and vectorization technology.

[0034] Construct a unified, real-time view of work order data, specifically including: 1. Asynchronous access and work order pool construction: Asynchronous access to work order data from multiple professional systems such as power distribution automation system, marketing management system, and production management system is achieved through high-throughput message middleware (such as Apache Kafka). After standardizing and cleaning the heterogeneous data, a unified real-time work order pool is constructed. 2. Streaming Feature Engineering: Based on a stream processing engine (such as Apache Flink), perform real-time comprehensive feature extraction on each work order in the work order pool to generate a structured feature vector.

[0035] Feature vectors include: 1) Spatiotemporal attributes: work order location (latitude and longitude coordinates), estimated operation time window, and timeliness requirements (including promised delivery time); 2) Business attributes: Work order type (such as emergency repair, inspection, meter installation, fault handling, etc.), work order priority, power supply line / transformer area; 3) Resource and skill requirements: Required tools and vehicles (such as insulated bucket trucks, testing instruments), special operation qualifications (such as live-line work permits, high-voltage operation permits), and one-hot encoding or word embedding technology are used to transform the above requirements into computable feature vectors; 4) Topology association attributes: Based on the power grid topology data, identify work orders with association (such as multiple fault work orders on the same power supply line, or related business work orders in the same transformer area), and use a simple three-level quantification strategy (e.g., association strength of the same equipment / transformer area = 1.0, the same branch line = 0.7, the same main line = 0.4) to convert the association into numerical features for subsequent clustering calculations.

[0036] Step S102: Dynamic personnel profiling and capability modeling.

[0037] To achieve real-time, multi-dimensional quantitative representation of the status of field staff, specifically including: 1. Real-time status synchronization: Through stream processing technology, the GPS location data and status report data (such as "working", "idle", "on the way", "paused") of the staff are accessed in real time to dynamically update the real-time location and work status of the staff. 2. Multi-dimensional profile construction: Construct a three-dimensional personnel profile that covers static capabilities, dynamic workload, and historical performance.

[0038] 1) Static capability profile: Records fixed attributes such as the staff's skill qualification list, certification level, usual work area, and types of tools and vehicles that can be operated; 2) Dynamic load profile: including: ① Real-time location and movement trajectory; ② Workload saturation: Based on the estimated working hours of the assigned but unfinished work orders, a time decay factor is introduced to calculate the dynamic load rate over a future period. The load rate calculation formula is: Load rate = ∑(remaining working hours × e^(-λ × risk coefficient)), where λ is the decay coefficient and the risk coefficient is the delay risk value calculated based on the work order execution progress deviation. ③Estimated service time: Based on the current task's real-time progress, historical individual work efficiency, time required to reach the next location, and workload balancing factor, dynamically predict the time when the next new work order can be received. 3) Performance profile: Statistics on staff’s historical average completion time, proficiency scores for specific work order types, and work quality scores (such as rework rate and user satisfaction).

[0039] Step S103: Intelligent packaging of work orders based on multi-constraint clustering.

[0040] Aggregating scattered work orders into "work order packages" that conform to business logic improves batch dispatch efficiency, specifically including: 1. Aggregation Rules: A multi-rule aggregation strategy is adopted that integrates geographical proximity, business relevance, and time window overlap. 2. Enhanced Clustering Algorithm: An improved multi-constraint density clustering algorithm is adopted. Based on the spatial distance constraint of the standard DBSCAN algorithm, business similarity (such as consistency of work order type and business process correlation) and temporal proximity (such as the degree of overlap and continuity of expected execution time windows) are added as constraints to generate "work order packages" adapted to on-site operations. 3. Packaged scoring model: Calculate the aggregate utility score for each candidate work order package. The aggregate utility score is a weighted linear combination of geographical concentration, business relevance and timeliness. The calculation formula is: U=α×C_geo+β×C_bus+γ×C_time.

[0041] in: C_geo represents the geographic concentration, used to quantify the spatial clustering of work orders within a work order package; C_bus represents the business relevance, which is based on a predefined business rule base and quantifies the relevance of each work order within a work order package in terms of business processes. C_time is the timeliness compactness, used to measure the overlap and continuity of the execution time windows of each work order within a quantity work order package; α, β, and γ are weighting coefficients that satisfy α+β+γ=1. Their values ​​can be configured and adjusted according to different business scenarios (such as emergency repair scenarios and routine inspection scenarios). Only when the aggregation utility score U of a candidate work order package exceeds a preset threshold is the work order package confirmed as a valid aggregation package and sent to the subsequent optimization and matching process.

[0042] Step S104: Dynamic matching and dispatch of work order packages and personnel based on rolling time domain optimization.

[0043] To achieve dynamic and optimal matching between work order packages and workers, specifically including: 1. Rolling optimization window settings: Set a fixed optimization window (e.g., the next 2 hours) and a rolling period (e.g., 15 minutes). Within each rolling period, unexecuted work order packages in the optimization window are re-planned.

[0044] 2. Multi-objective optimization modeling: This involves fusing multiple conflicting objectives into a single total cost function and ensuring the feasibility of the proposed solution through constraints. (1) The decision variable is: x_ij: a binary variable, which means that work order package i is assigned to worker j (x_ij = 1 represents assignment, 0 represents no assignment); (2) Objective function: Minimize the total cost Z, as shown in the following equation: Min Z = W1 * C_travel + W2 * C_workload + W3 * C_skill + W4 * C_priority; Where C_travel is the travel time, C_travel = Σ_ij (x_ij * TravelTime(j, i)), and TravelTime(j, i) is the estimated travel time for person j from their current location or the end point of the previous task to the starting point of work order package i (real-time traffic conditions need to be connected). C_workload is the variance of workload saturation, C_workload = Variance( [Workload_j for jin AllPersonnel] ). Where Workload_j is the total estimated working hours of person j within the optimization window (including newly assigned work orders); C_skill represents the skill matching gap, where C_skill = Σ_ij (x_ij * SkillGap(i, j)). SkillGap(i, j) is a quantified value. For example, if a work order requires a high-voltage certificate but the personnel do not have one, the gap is infinite (hard constraint); if a work order requires a certain piece of equipment and the personnel have moderate proficiency, the gap is a relatively small value.

[0045] C_priority represents the default risk of a work order, where C_priority = Σ_ij (x_ij * Risk(i,j)). Risk(i,j) is the estimated delay risk after work order i is assigned to personnel j, which can be calculated based on (work order commitment time - current time - personnel j's estimated ready time - travel time). For work orders with higher priority, this risk coefficient should be set larger.

[0046] (3) Normalization. To ensure that the dimensions of each cost component in the objective function are consistent and that the weight coefficients can truly reflect business preferences, it is necessary to normalize each cost component.

[0047] The normalized objective function is: Min Z = W1 * (C_travel / T_norm) + W2 * (C_workload / W_norm) + W3 *(C_skill / S_norm) + W4 * (C_priority / R_norm); Where T_norm: is the maximum estimated travel time between all work order packages and all personnel within the current optimization window; W_norm: can be a fixed value set based on historical data, or a theoretical maximum value; S_norm: The sum of the maximum skill gaps for available personnel for each work order in the current window; R_norm: Uses the sum of the maximum possible risk values ​​in the current work order-person combination.

[0048] (4) Determination of weighting coefficients.

[0049] The weighting is a direct reflection of business strategy. Weight coefficients W1-W4 can be dynamically switched based on preset business scenario patterns (such as daily inspections, emergency repairs, and planned operations) to achieve strategic task dispatching. In the future, based on accumulated historical data, we can explore adaptive dynamic optimization of weights using reinforcement learning technology.

[0050] Several typical business scenario patterns are preset, and a fixed set of weights is configured for each pattern. For example, the implementation is as follows: 1) Routine inspection mode: W1=0.4, W2=0.3, W3=0.2, W4=0.1 (emphasizing efficiency and fairness); 2) Emergency Repair Mode: W1=0.2, W2=0.1, W3=0.2, W4=0.5 (extreme emphasis on priority and timeliness); 3) Planned work mode: W1=0.3, W2=0.4, W3=0.2, W4=0.1 (focusing on load balancing and skill matching); 4) Severe weather warning mode: W1=0.1, W2=0.1, W3=0.3, W4=0.5 (emphasis on safety and response speed).

[0051] Triggering conditions: The system can automatically switch modes based on information such as time (e.g., entering emergency repair mode at night), work order type ratio, and weather warnings.

[0052] (5) Constraints.

[0053] Hard skill constraint: If personnel j does not possess the mandatory skills or qualifications required by work order package i, assignment is prohibited, and x_ij = 0 (for all pairs of i, j that do not meet the skill requirements).

[0054] Hard capacity constraint: The total working hours of all work orders assigned to person j cannot exceed the remaining available working hours in this scrolling window.

[0055] Hard time constraint: The latest start time of work order package i must be later than the estimated service time of personnel j plus travel time.

[0056] Uniqueness constraint: For each work order i, there must be one and only one worker j assigned to it. That is: Σ_jx_ij = 1 (for all work orders i).

[0057] 3. High-performance hybrid solution.

[0058] To efficiently solve the multi-objective optimization model defined in section 2 above, this step employs a high-performance hybrid solution strategy. The core of this strategy is the Adaptive Large Neighborhood Search (ALNS) algorithm. This algorithm is guided by the model's objective function Min Z and strictly adheres to all hard constraints defined in the model during the search process, aiming to obtain a high-quality approximate optimal solution within a finite time. Its basic process is as follows: (1) Initial solution generation: A rule-based greedy algorithm is used to quickly construct an initial feasible solution. First, the work orders are sorted according to rules such as work order priority and timeliness; then, each work order is assigned to the feasible personnel that minimizes the insertion cost (such as additional travel time, skill gap, etc.) in sequence, thereby efficiently generating the search starting point of the ALNS algorithm; (2) Representation of solutions: A "solution" in the algorithm directly corresponds to a feasible assignment scheme in the model, that is, the assignment of a legal decision variable matrix x_ij; (3) Evaluation function: The algorithm uses the objective function Min Z defined in the model as the evaluation function. The quality of any solution is measured by its corresponding Z value; the smaller the Z value, the better the solution. (4) Constraint handling: Hard constraints defined in the model (such as skill, capacity, and uniqueness constraints) serve as feasibility criteria in the generation and transformation of solutions. Candidate solutions generated by the algorithm at any step must satisfy all of these constraints, otherwise they will be discarded. (5) Search mechanism: ALNS continuously generates new candidate solutions through “destruction” and “repair” operators and always moves in the direction of smaller Z value, thereby achieving rapid and effective exploration of the model solution space.

[0059] 4. Hybrid Triggering and Dispatch Mechanism: Event-driven triggering: When a new work order is inserted, a work order is canceled, an abnormal personnel status occurs, or an execution delay warning is issued, local or global re-optimization is immediately triggered; Periodic scan trigger: A global optimization is automatically triggered every rolling cycle to fine-tune the existing assignment scheme; Solution Validation and Deployment: The final generated assignment plan must undergo a lightweight rule validation, focusing on verifying whether all hard constraints (such as skill matching, work hour capacity, etc.) are met. If the validation passes, the plan is immediately deployed to the worker's mobile terminal for execution; if the validation fails, the plan is marked as invalid, triggering the system to re-optimize, and an alarm log is recorded for analysis.

[0060] Step S105: Execution process monitoring and proactive dynamic adjustment.

[0061] To establish a closed-loop decision-making process and ensure consistency between planning and actual execution, this includes: 1. Real-time progress monitoring and delay prediction: Multi-source data fusion: Integrates milestone events reported by mobile devices (departure, arrival at the work site, start of work, completion of work) with real-time GPS trajectory data of workers to dynamically track the progress of work orders; Intelligent delay warning: Using a lightweight machine learning model (such as gradient boosting tree), based on features such as "time elapsed / estimated working time", distance between the current location and the work site, historical work efficiency, traffic conditions, etc., the work order delay probability is calculated in real time. When the delay probability exceeds the preset threshold, an early warning is triggered.

[0062] 2. Feedback loop and self-learning: Dynamic replanning: The monitored signals such as work order delays, personnel abnormalities, and risk warnings are used as re-optimization trigger events and fed back to step S104 in real time to drive the system to make dynamic adjustments; Data accumulation and model update: The assignment plan and actual execution results (such as actual working time, actual driving route, and work quality) of each assignment are stored in the database and used periodically to retrain the work time prediction model, optimize the objective function weight coefficients and clustering algorithm parameters, so as to achieve continuous self-evolution of the system.

[0063] This solution avoids repetitive back-and-forth work by aggregating multi-source work order data and identifying topological associations, combined with a multi-constraint work order packaging strategy. Based on dynamic personnel profiles and multi-objective optimization matching, it achieves accurate matching of work orders and personnel, significantly improving human resource utilization and work order processing efficiency.

[0064] This solution employs a rolling time-domain optimization and hybrid triggering mechanism, dynamically adjusting the assigned scheme at fixed intervals or based on events. This effectively addresses uncertainties on-site, ensuring the scheme always adapts to actual working conditions and solving the problem of poor adaptability of traditional static models.

[0065] This solution combines proactive delay warnings to mitigate the risk of work orders exceeding time limits and improve the standardization and reliability of business processing.

[0066] This solution constructs a closed-loop architecture of "perception-decision-execution-learning", continuously feeding back actual execution data to optimize model parameters, enabling the system to have autonomous learning capabilities and continuously improve decision accuracy as running time increases.

[0067] Correspondingly, a computer program product includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the aforementioned smart grid work order dispatch method.

[0068] Accordingly, an electronic device includes at least one processor and a memory connected to the processor, the memory being used to store computer programs; the processor is used to execute the computer programs, enabling the electronic device to implement the aforementioned smart grid work order dispatching method.

[0069] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent work order dispatching in a power grid, characterized in that, Includes the following steps: The work order aggregation and feature extraction module is configured to: access work order data from multiple professional systems, build a unified work order pool, and extract the spatiotemporal attributes, business attributes, resource and skill requirements, and topological association attributes of the work orders as feature vectors; The mobile terminal status monitoring module is configured to: obtain the GPS location and working status identifier of each mobile terminal, and calculate the dynamic load rate and estimated service time based on the incomplete work orders bound to it; The work order intelligent packaging module is configured to: aggregate scattered work orders with similar feature vectors into work order packages that conform to geographical proximity, business relevance, and timeliness overlap, based on a multi-constraint clustering algorithm; The dynamic matching and dispatch module is configured to: take work order packages and available mobile terminals as inputs, establish a multi-objective optimization model, generate the optimal dispatch plan by comprehensively optimizing travel time, load balancing and skill matching, and send it to the terminal based on a hybrid event and periodic triggering mechanism; The execution monitoring and closed-loop learning module is configured to: monitor the execution progress of issued work orders, issue delay warnings, and use the feedback data to trigger dynamic replanning and optimize model parameters.

2. The method for intelligent power grid work order dispatch as described in claim 1, characterized in that, The spatiotemporal attributes include the latitude and longitude coordinates of the work order location, the expected operation time window, and the timeliness requirements; Business attributes include work order type, work order priority, and the power supply line or transformer area to which it belongs; Resource and skill requirements are transformed into computable feature vectors using one-hot encoding or word embedding techniques. Topology association attributes are based on power grid topology data to identify related business work orders on the same power supply line or in the same transformer area.

3. The method for intelligent power grid work order dispatch as described in claim 1, characterized in that, In the mobile terminal status monitoring module, the dynamic load rate is calculated using the following formula: Load factor = Σ(remaining working hours × e^(-λ × risk coefficient)); Wherein, λ is the preset attenuation coefficient, and the risk coefficient is the delay risk value calculated based on the deviation of the work order execution progress.

4. The method for intelligent power grid work order dispatch as described in claim 1, characterized in that, The intelligent work order packaging module uses an aggregate utility model to evaluate work order packages, as shown in the following formula: U=α×C_geo+β×C_bus+γ×C_time; Where U is the aggregate utility score, C_geo is the geographic concentration degree representing the degree of spatial clustering of each work order in the work order package, C_bus is the business relevance degree quantified based on the predefined business rule base, C_time is the timeliness compactness representing the degree of overlap and continuity of the planned execution time windows of each work order in the work order package, and α, β, and γ are configurable weight coefficients, satisfying α+β+γ=1.

5. The method for intelligent power grid work order dispatch as described in claim 1, characterized in that, The multi-objective optimization model in the dynamic matching and dispatch module is as follows: the objective function is to minimize the total cost Z, Z=W1×Σ(travel time)+W2×workload saturation variance+W3×Σ(skill matching gap)+W4×Σ(work order priority default risk), where W1-W4 are weight coefficients.

6. The method for intelligent power grid work order dispatch as described in claim 1, characterized in that, The dynamic matching and dispatching module uses a hybrid solution method based on adaptive large neighborhood search (ALNS) to generate an assignment scheme. Specifically, it generates an initial feasible solution based on work order priority and timeliness rules using a greedy insertion method. Then, starting from the initial feasible solution, it performs multiple rounds of destruction-repair loops to continuously optimize. In each round of the loop, a destruction operation is performed to remove some assigned work orders from the current solution, and a repair operation is performed to reinsert the removed work orders into the current solution to generate a new solution.

7. The method for intelligent power grid work order dispatch as described in claim 6, characterized in that, The generated new solution must satisfy constraints on skill matching, working hours capacity, and uniqueness allocation, and the search direction is based on the optimization objective function Z.

8. A method for intelligent power grid work order dispatch, implemented by the intelligent power grid work order dispatch system according to any one of claims 1-7, characterized in that, Includes the following steps: Acquire work order data from multiple professional systems, construct a unified work order pool, and extract the spatiotemporal attributes, business attributes, resource and skill requirements, and topological association attributes of the work orders as feature vectors; Obtain the GPS location and working status identifier of each mobile terminal, and calculate the dynamic load rate and estimated service time based on the incomplete work orders bound to them; Based on the multi-constraint clustering algorithm, scattered work orders with similar feature vectors are aggregated into work order packages that meet the criteria of geographical proximity, business relevance, and overlapping timeliness. Using work order packages and available mobile terminals as input, a multi-objective optimization model is established. The optimal assignment scheme is generated by comprehensively optimizing travel time, load balancing and skill matching, and then sent to the terminal based on a hybrid event and period triggering mechanism. Monitor the execution progress of issued work orders, issue delay warnings, and use the feedback data to trigger dynamic replanning and optimize model parameters.

9. A computer program product, characterized in that, It includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to perform the steps in the smart power grid work order dispatch method as described in claim 8.

10. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, the memory being used to store a computer program; the processor is used to execute the computer program, enabling the electronic device to implement the steps in the smart power grid work order dispatching method as described in claim 8.

Images