Intelligent heat supply work order dispatching method and system based on multi-objective optimization
By employing a multi-objective optimization and knowledge graph-guided intelligent work order dispatching method, the problems of slow response and unreasonable resource allocation in the traditional heating work order scheduling mode are solved, achieving efficient and accurate work order dispatching and dynamic scheduling, and improving the system's intelligent decision-making and anti-disturbance capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHIJIAZHUANG HUADIAN HEATING GRP CO LTD
- Filing Date
- 2025-12-25
- Publication Date
- 2026-04-24
AI Technical Summary
Traditional heating work order scheduling relies on manual experience, resulting in slow response, unreasonable resource allocation, and difficulty in dealing with the massive and highly uncertain work order scheduling problem during peak heating periods.
A multi-objective optimization-based intelligent dispatching method for heating work orders is adopted. By acquiring multi-source heterogeneous data in real time, a multi-objective optimization model integrating domain knowledge is constructed. A hybrid meta-heuristic algorithm guided by knowledge graph is used for collaborative solution. Dynamic rescheduling is carried out through rolling time-domain optimization and model predictive control framework. Finally, the dispatching decision is realized by combining human-computer interaction interface.
It achieves efficient and accurate work order dispatch, can cope with dynamic changes during peak heating periods, ensures the stability and reliability of the scheduling plan, improves the system's intelligent decision-making ability and anti-disturbance ability, and realizes efficient scheduling through human-machine collaboration.
Smart Images
Figure CN121920728A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent heating and operations optimization, and particularly to a method and system for intelligent dispatching of heating work orders based on multi-objective optimization. Background Technology
[0002] With the continuous expansion of centralized heating areas in urban areas and the increasing demands of users for heating service quality, heating companies are facing increasingly severe pressure in work order scheduling. The traditional scheduling model relies heavily on the personal experience of dispatchers, using methods such as telephone communication, manual recording, and experience-based work order assignment, which results in problems such as slow response, unreasonable resource allocation, and difficulty in coping with sudden peak periods.
[0003] The current technological development in the field of heating work order scheduling presents a "strong at both ends and weak in the middle" situation; that is, the intelligent progress of work order generation (front end) and pipeline network control (back end) is rapid, but the key decision-making link of "work order-technician" precise matching and path planning (middle end), which connects the front and the back, still lacks an intelligent solution that can integrate multi-source information, coordinate multiple objectives, adapt to dynamic changes, and efficiently collaborate with people.
[0004] Especially when dealing with the massive, concurrent, and highly uncertain work orders during the peak winter heating season, the limitations of existing technologies become more apparent. Therefore, a multi-objective optimization-based intelligent work order dispatching method and system are proposed to address the aforementioned issues. Summary of the Invention
[0005] The purpose of this invention is to provide a method and system for intelligent dispatching of heating work orders based on multi-objective optimization, which aims to address the limitations of existing technologies when dealing with massive, concurrent, and highly uncertain work orders during the peak winter heating season.
[0006] Specifically: The intelligent dispatching method for heating work orders based on multi-objective optimization includes the following steps:
[0007] S1. Real-time acquisition and fusion of multi-source heterogeneous data, including work order data, technician status data, geographic information data and dynamic environment data;
[0008] S2. Construct a multi-objective optimization model that integrates domain knowledge. The multi-objective optimization model includes at least five conflicting optimization objectives: response timeliness, operational economy, skill adaptability, dynamic load balancing, and service reliability.
[0009] S3. A knowledge graph-guided hybrid metaheuristic algorithm is used to collaboratively solve the multi-objective optimization model and output the Pareto optimal solution set.
[0010] S4. Based on the rolling time-domain optimization and model predictive control framework, the scheduling scheme currently in execution is dynamically monitored and rescheduled in real time.
[0011] S5. By integrating a multi-channel human-computer interaction and explainable artificial intelligence decision support interface, the final dispatch plan is selected from the Pareto front and executed.
[0012] The technical solution of this application will be further described below:
[0013] In one embodiment, in step S1, the work order data collection involves obtaining structured work order flow in real time through a standardized interface of the customer service system, including user identifier, fault code, text description, priority identifier, appointment time window, and geographic location text; and receiving automated work orders generated by abnormal alarms from smart heat meters and indoor temperature sensors through an IoT protocol proxy.
[0014] Technician status data is collected through a dedicated mobile terminal application deployed on the technician's side, continuously reporting BeiDou positioning coordinates, equipment status, task execution status, and image evidence; and through connection to low-power Bluetooth or IoT gateway, biometric indicators such as heart rate and body temperature of technicians' smart badges or wearable devices are collected for auxiliary assessment of workload and fatigue.
[0015] Geographic information data acquisition involves integrating enterprise-level geographic information system services to obtain vector topology of heating pipelines, coordinates of valves and heat exchange stations, and historical polygon data of construction restricted areas; and by calling third-party map platform APIs to obtain real-time road network vector data, traffic restriction information, and precise coordinates of points of interest.
[0016] Dynamic environmental data acquisition involves subscribing to meteorological data service interfaces to obtain gridded weather forecasts and real-time alerts, including temperature, precipitation, wind speed, and visibility; parsing real-time traffic flow data released by traffic management departments or road condition APIs from commercial navigation services to obtain road speed and congestion index; and aggregating time-series monitoring data from pressure, flow, and temperature sensors deployed at key pipeline nodes through an IoT platform.
[0017] In one embodiment, in step S3, the knowledge graph-guided hybrid metaheuristic algorithm includes:
[0018] a. In the initialization phase, an initial population is generated using a case-based reasoning strategy, where the case library is constructed from historical high-quality scheduling schemes;
[0019] b. In the evolutionary stage, an improved NSGA-III framework is used for iterative optimization, in which the selection probability of crossover and mutation operators is dynamically adjusted by the domain rules of real-time reasoning of the knowledge graph;
[0020] c. In the guidance phase, in each generation of evolution, the graph neural network is used to embed the relationship between "fault-skill-equipment" in the knowledge graph, and the generated embedding vector is used as auxiliary gradient information to guide the population to search for high-quality solution regions.
[0021] d. In the collaborative phase, a parallel adaptive local search thread is introduced to perform fine-tuning of the elite solution based on the constraint satisfaction problem, and exchanges information with the global evolution thread through a shared memory pool.
[0022] In one embodiment, step S4 specifically includes:
[0023] a. Initiate the rescheduling process using a fixed time window, or events such as newly added emergency work orders and technician task delays;
[0024] b. Treat the currently executing and uninterrupted tasks as fixed constraints, and unfreeze the tasks that have not yet started as optimizable variables to form a rolling time-domain optimization problem.
[0025] c. A lightweight model predictive control algorithm is adopted, with minimizing scheduling disturbances as one of the primary objectives, to quickly solve the optimization problem in the finite time domain;
[0026] d. Only issue the adjusted route instructions to the affected technicians to ensure the overall stability of the scheduling system.
[0027] In one embodiment, in step S5, the decision support interface includes:
[0028] A Pareto front visualization dashboard with multi-view fusion, including at least a parallel coordinate plot, a 3D scatter matrix, and a radar chart;
[0029] An interactive module integrating implicit preference learning and explicit rule setting, used to capture decision-maker intent;
[0030] An interpretable AI module based on counterfactual reasoning is used to generate comparative explanations for recommendation schemes;
[0031] A three-point collaborative decision-making workflow that supports asynchronous annotation by multiple experts and consensus calculation.
[0032] Another objective of this invention is to provide a multi-objective optimization-based intelligent dispatching system for heating work orders, used to implement the aforementioned multi-objective optimization-based intelligent dispatching method for heating work orders. The system includes:
[0033] The data perception and fusion layer is used to collect and standardize multi-source heterogeneous data in real time through IoT devices, customer service system interfaces, and third-party data APIs;
[0034] The digital twin and simulation layer constructs a virtual heating service environment that includes physical entities, operating rules, and random events, for scenario simulation and risk assessment.
[0035] The core optimization and decision-making layer includes a knowledge graph management module, a multi-objective optimization solution engine, a dynamic rescheduler, and a human-machine collaborative decision-making module.
[0036] The solution execution and feedback layer is responsible for distributing the optimized solution to the technicians' mobile terminals and collecting real-time feedback data during the execution process to form an optimization closed loop.
[0037] Compared with the prior art, the present invention has the following advantages:
[0038] 1. This invention constructs an integrated technical solution that integrates data fusion, knowledge-driven, multi-objective collaboration, rolling optimization, and human-machine co-decision-making. It models the scheduling of heating work orders as a high-dimensional, dynamic, multi-objective combinatorial optimization problem, introduces domain knowledge as a guide for search and decision-making, and solves the problem efficiently through a hierarchical collaborative intelligent algorithm. Finally, the optimization decision is implemented through a human-machine collaborative interface to form a closed loop.
[0039] Unlike solutions that only optimize business processes or work order address matching, this invention directly addresses the core of scheduling—multi-objective resource optimization—and provides a complete decision-making chain from modeling and solving to dynamic adjustment.
[0040] Compared to automatic order dispatch based on fixed rules or collaborative platforms that mainly rely on human judgment, this invention introduces knowledge graphs and advanced optimization algorithms, enabling the system to have an understanding and optimization capability similar to expert experience, and achieving a higher level of intelligent decision-making.
[0041] Through rolling time-domain optimization and model predictive control framework, the system of this invention has strong online adjustment and anti-disturbance capabilities, and can cope with various emergencies with ease, ensuring the stability and reliability of the scheduling plan, which is not available in most static optimization systems.
[0042] By using explainable AI and multimodal collaborative design, this invention breaks the black box of intelligent algorithms, enabling dispatchers not only to use the system, but also to understand, trust, and effectively supervise the system's decisions, achieving true human-machine intelligent integration rather than simple replacement.
[0043] 2. Construct a multi-objective optimization model that integrates domain knowledge, going beyond the traditional spatiotemporal constraint model. Creatively, it quantifies and integrates knowledge graphs of the heating domain, such as equipment failure models and the relationship between technician skills, into the objective function, so that the optimization model not only pursues mathematical optimality, but also business rationality and accuracy.
[0044] 3. Based on a knowledge graph-guided hybrid metaheuristic algorithm, a hybrid algorithm framework of knowledge-guided global evolution and model-driven local search was designed. The knowledge graph and graph neural network provide directional guidance for the metaheuristic algorithm, which significantly improves search efficiency and solution quality, and solves the contradiction between convergence speed and solution accuracy in large-scale real-time optimization.
[0045] 4. A dynamic rescheduling mechanism with minimal disturbance is proposed. A rescheduling strategy based on rolling time-domain optimization and model predictive control is proposed. The dynamic scheduling problem is transformed into a series of short-time-domain static optimization problems. Minimizing the disturbance to the original plan is the important optimization objective, ensuring the smooth transition and executability of scheduling instructions. Attached Figure Description
[0046] Figure 1 This is a diagram illustrating the overall architecture of the intelligent dispatching system for heating work orders based on multi-objective optimization, as described in this invention. Detailed Implementation
[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. The specific implementation of the invention will be described in detail below with reference to specific embodiments.
[0048] In this embodiment of the invention, firstly, a decision support interface integrating multi-channel human-computer interaction and explainable artificial intelligence is deployed to select and execute the final dispatch plan from the Pareto front. The decision support interface includes a multi-view fusion Pareto front visualization dashboard, which at least includes a parallel coordinate graph, a three-dimensional scatter matrix, and a radar chart; an interactive module integrating implicit preference learning and explicit rule setting to capture the decision-maker's intent; an explainable artificial intelligence module based on counterfactual reasoning to generate comparative explanations for recommended plans; and a three-point collaborative decision-making workflow supporting asynchronous annotation and consensus calculation by multiple experts. The data perception layer interfaces with the enterprise's existing customer service system, vehicle GPS system, meteorological data interface, and production scheduling platform; the knowledge graph management module extracts knowledge from historical maintenance records, equipment manuals, and training materials to construct an initial domain knowledge graph.
[0049] When a batch of new work orders enters the system, the optimization engine starts.
[0050] Target Calculation: The system calculates the scores of each potential "work order-technician" assignment pair on five objectives in real time. For example, when calculating skill fit, the system does not simply compare skill tags, but queries a knowledge graph: if the work order description is "heat exchanger plate leakage," the knowledge graph associates it with skills such as "sealing welding" and "pressure vessel knowledge," and points out that this fault is often accompanied by "corrosion" issues. Based on this, the system more accurately assesses the matching depth between the technician's skill vector and the fault characteristics, achieving "knowledge-enhanced matching."
[0051] Algorithm solution: The multi-objective optimization model that integrates domain knowledge specifically includes:
[0052] Objective function 1: Minimize risk-weighted response time
[0053]
[0054] Where, ρ i ω represents the potential risk coefficient of a work order based on knowledge graph reasoning. i For the priority weight of the work order, t i start t represents the actual start time of processing. i expect The user's expected response time;
[0055] Objective function 2: Minimize total lifecycle operating costs
[0056]
[0057] Where η(traffic) is the real-time traffic condition impact factor, c carbon ⋅E ij The cost of carbon emissions based on estimated energy consumption;
[0058] Objective function 3: Maximize the skill fit of knowledge enhancement
[0059]
[0060] Where KG-Match(.) is a matching function based on a knowledge graph in the heating field, Fi is the feature vector of the work order, and P j H is the vector image of the technician. ij This is a record of their historical collaboration.
[0061] Objective function four: Maximize resilience-driven dynamic load balancing.
[0062]
[0063] Among them, W t Let Θ(.) be the technician load vector at time t, and let Θ(.) be the parameter representing the technician's real-time fatigue level ΔL.j,t The decay function;
[0064] Objective Function 5: Maximize the service reliability of multi-source information fusion
[0065]
[0066] Among them, R ij base Based on reliability, λ i weather As a weather risk factor, Π ij spare This is the backup support coefficient calculated based on the status of nearby technicians.
[0067] We construct a multi-objective optimization model that integrates domain knowledge, going beyond traditional spatiotemporal constraint models. We creatively incorporate knowledge graphs of the heating domain, such as equipment failure models and the relationship between technician skills, into the objective function, so that the optimization model not only pursues mathematical optimality, but also business rationality and accuracy.
[0068] The initial population was generated by adjusting historical high-quality scheduling schemes with similar weather and time periods within the past week. In each generation of NSGA-III evolution, the system uses graph neural networks to analyze the "fault-solving" path patterns formed by the current elite solutions in the knowledge graph. If certain associations are found to be frequently used, an auxiliary gradient signal is generated to encourage the algorithm to explore such combinations, thereby guiding the search direction.
[0069] Output result: After approximately 120 iterations (taking 45 seconds), the algorithm outputs a Pareto front consisting of 28 non-dominated solutions.
[0070] In this embodiment of the invention, the intelligent dispatching method for heating work orders based on multi-objective optimization includes the following steps:
[0071] S1. Real-time acquisition and fusion of multi-source heterogeneous data, including work order data, technician status data, geographic information data and dynamic environment data;
[0072] S2. Construct a multi-objective optimization model that integrates domain knowledge. The multi-objective optimization model includes at least five conflicting optimization objectives: response timeliness, operational economy, skill adaptability, dynamic load balancing, and service reliability.
[0073] S3. A knowledge graph-guided hybrid metaheuristic algorithm is used to collaboratively solve the multi-objective optimization model and output the Pareto optimal solution set.
[0074] S4. Based on the rolling time-domain optimization and model predictive control framework, the scheduling scheme currently in execution is dynamically monitored and rescheduled in real time.
[0075] S5. By integrating a multi-channel human-computer interaction and explainable artificial intelligence decision support interface, the final dispatch plan is selected from the Pareto front and executed.
[0076] This invention constructs an integrated technical solution that integrates data fusion, knowledge-driven, multi-objective collaboration, rolling optimization, and human-machine co-decision-making. It models the scheduling of heating work orders as a high-dimensional, dynamic, multi-objective combinatorial optimization problem, introduces domain knowledge as a guide for search and decision-making, and solves the problem efficiently through a hierarchical collaborative intelligent algorithm. Finally, the optimization decision is implemented through a human-machine collaborative interface to form a closed loop.
[0077] Unlike solutions that only optimize business processes or work order address matching, this invention directly addresses the core of scheduling—multi-objective resource optimization—and provides a complete decision-making chain from modeling and solving to dynamic adjustment.
[0078] Compared to automatic order dispatch based on fixed rules or collaborative platforms that mainly rely on human judgment, this invention introduces knowledge graphs and advanced optimization algorithms, enabling the system to have an understanding and optimization capability similar to expert experience, and achieving a higher level of intelligent decision-making.
[0079] Through rolling time-domain optimization and model predictive control framework, the system of this invention has strong online adjustment and anti-disturbance capabilities, and can cope with various emergencies with ease, ensuring the stability and reliability of the scheduling plan, which is not available in most static optimization systems.
[0080] By employing explainable AI and multimodal collaborative design, this invention breaks down the black box of intelligent algorithms, enabling dispatchers not only to use the system but also to understand, trust, and effectively supervise its decisions, achieving true human-machine intelligent integration rather than simple replacement.
[0081] Furthermore, in step S1, the work order data collection involves obtaining structured work order flow in real time through the standardized interface of the customer service system, including user identifier, fault code, text description, priority identifier, appointment time window, and geographical location text; and receiving automated work orders generated from abnormal alarms from smart heat meters and indoor temperature sensors through an IoT protocol proxy.
[0082] Technician status data is collected through a dedicated mobile terminal application deployed on the technician's side, continuously reporting BeiDou positioning coordinates, equipment status, task execution status, and image evidence; and through connection to low-power Bluetooth or IoT gateway, biometric indicators such as heart rate and body temperature of technicians' smart badges or wearable devices are collected for auxiliary assessment of workload and fatigue.
[0083] Geographic information data acquisition involves integrating enterprise-level geographic information system services to obtain vector topology of heating pipelines, coordinates of valves and heat exchange stations, and historical polygon data of construction restricted areas; and by calling third-party map platform APIs to obtain real-time road network vector data, traffic restriction information, and precise coordinates of points of interest.
[0084] Dynamic environmental data acquisition involves subscribing to meteorological data service interfaces to obtain gridded weather forecasts and real-time alerts, including temperature, precipitation, wind speed, and visibility; parsing real-time traffic flow data released by traffic management departments or road condition APIs from commercial navigation services to obtain road speed and congestion index; and aggregating time-series monitoring data from pressure, flow, and temperature sensors deployed at key pipeline nodes through an IoT platform.
[0085] In step S3, the knowledge graph-guided hybrid metaheuristic algorithm includes:
[0086] a. In the initialization phase, an initial population is generated using a case-based reasoning strategy, where the case library is constructed from historical high-quality scheduling schemes;
[0087] b. In the evolutionary stage, an improved NSGA-III framework is used for iterative optimization, in which the selection probability of crossover and mutation operators is dynamically adjusted by the domain rules of real-time reasoning of the knowledge graph;
[0088] c. In the guidance phase, in each generation of evolution, the graph neural network is used to embed the relationship between "fault-skill-equipment" in the knowledge graph, and the generated embedding vector is used as auxiliary gradient information to guide the population to search for high-quality solution regions.
[0089] d. In the collaborative phase, a parallel adaptive local search thread is introduced to perform fine-tuning of the elite solution based on the constraint satisfaction problem, and exchanges information with the global evolution thread through a shared memory pool.
[0090] Based on a knowledge graph-guided hybrid metaheuristic algorithm, a hybrid algorithm framework of knowledge-guided global evolution and model-driven local search was designed. The knowledge graph and graph neural network provide directional guidance for the metaheuristic algorithm, which significantly improves search efficiency and solution quality, and solves the contradiction between convergence speed and solution accuracy in large-scale real-time optimization.
[0091] In step S4, the dynamic rescheduling step specifically includes:
[0092] a. Initiate the rescheduling process using a fixed time window, or events such as newly added emergency work orders and technician task delays;
[0093] b. Treat the currently executing and uninterrupted tasks as fixed constraints, and unfreeze the tasks that have not yet started as optimizable variables to form a rolling time-domain optimization problem.
[0094] c. A lightweight model predictive control algorithm is adopted, with minimizing scheduling disturbances as one of the primary objectives, to quickly solve the optimization problem in the finite time domain;
[0095] d. Only issue the adjusted route instructions to the affected technicians to ensure the overall stability of the scheduling system.
[0096] The dynamic rescheduling mechanism with minimal disturbance proposes a rescheduling strategy based on rolling time-domain optimization and model predictive control. This strategy transforms the dynamic scheduling problem into a series of short-time-domain static optimization problems, with minimizing the disturbance to the original plan as the key optimization objective, ensuring the smooth transition and executability of scheduling instructions.
[0097] In step S5, the decision support interface includes:
[0098] A Pareto front visualization dashboard with multi-view fusion, including at least a parallel coordinate plot, a 3D scatter matrix, and a radar chart;
[0099] An interactive module integrating implicit preference learning and explicit rule setting, used to capture decision-maker intent;
[0100] An interpretable AI module based on counterfactual reasoning is used to generate comparative explanations for recommendation schemes;
[0101] A three-point collaborative decision-making workflow that supports asynchronous annotation by multiple experts and consensus calculation.
[0102] In embodiments of the present invention, such as Figure 1 This paper provides a multi-objective optimization-based intelligent dispatching system for heating work orders, used to implement the aforementioned multi-objective optimization-based intelligent dispatching method for heating work orders. The system includes:
[0103] The data perception and fusion layer is used to collect and standardize multi-source heterogeneous data in real time through IoT devices, customer service system interfaces, and third-party data APIs;
[0104] The digital twin and simulation layer constructs a virtual heating service environment that includes physical entities, operating rules, and random events, for scenario simulation and risk assessment.
[0105] The core optimization and decision-making layer includes a knowledge graph management module, a multi-objective optimization solution engine, a dynamic rescheduler, and a human-machine collaborative decision-making module.
[0106] The solution execution and feedback layer is responsible for distributing the optimized solution to the technicians' mobile terminals and collecting real-time feedback data during the execution process to form an optimization closed loop.
[0107] Furthermore, the knowledge graph management module is specifically used for:
[0108] Build and maintain a domain knowledge graph covering heating equipment, fault types, repair processes, technician skills, and spare parts and materials;
[0109] Based on the work order text description, natural language processing technology is used to automatically extract entities and relationships for real-time expansion of the knowledge graph;
[0110] It provides graph-based reasoning services, supporting root cause inference for complex faults and multi-skill collaborative dispatch suggestions.
[0111] The dynamic rescheduler is implemented using a microservice architecture, including:
[0112] Event listeners continuously monitor the work order status stream, technician GPS location stream, and external event stream;
[0113] Impact evaluator, based on network flow model, quickly assesses the impact range of sudden events on the current scheduling plan;
[0114] A fast solver with an embedded lightweight optimization algorithm for responding to rescheduling requests in seconds;
[0115] The instruction differential compares the old and new scheduling schemes and generates a minimum set of incremental adjustment instructions.
[0116] The human-machine collaborative decision-making module provides three switchable working modes:
[0117] In AI-driven mode, the system automatically recommends the optimal solution and executes it directly, which is suitable for normal working conditions.
[0118] In the loop mode, the system provides a finite number of optimal solutions and their visual explanations, with the final decision made by the dispatcher.
[0119] In the manual-led mode, the dispatcher can manually adjust the plan, and the system will then perform local optimizations based on the manual plan and suggest areas for improvement.
[0120] In the description of this invention, although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for intelligent dispatching of heating work orders based on multi-objective optimization, characterized in that, Includes the following steps: S1. Real-time acquisition and fusion of multi-source heterogeneous data, including work order data, technician status data, geographic information data and dynamic environment data; S2. Construct a multi-objective optimization model that integrates domain knowledge. The multi-objective optimization model includes at least five conflicting optimization objectives: response timeliness, operational economy, skill adaptability, dynamic load balancing, and service reliability. S3. A knowledge graph-guided hybrid metaheuristic algorithm is used to collaboratively solve the multi-objective optimization model and output the Pareto optimal solution set. S4. Based on the rolling time-domain optimization and model predictive control framework, the scheduling scheme currently in execution is dynamically monitored and rescheduled in real time. S5. By integrating a multi-channel human-computer interaction and explainable artificial intelligence decision support interface, the final dispatch plan is selected from the Pareto front and executed.
2. The intelligent dispatching method for heating work orders based on multi-objective optimization according to claim 1, characterized in that, In step S1, work order data collection involves obtaining structured work order flow in real time through a standardized interface of the customer service system, including user identifier, fault code, text description, priority identifier, appointment time window, and geographic location text; and receiving automated work orders generated by abnormal alarms from smart heat meters and indoor temperature sensors through an IoT protocol proxy. Technician status data is collected through a dedicated mobile terminal application deployed on the technician's side, continuously reporting BeiDou positioning coordinates, equipment status, task execution status, and image evidence; and through connection to low-power Bluetooth or IoT gateway, biometric indicators such as heart rate and body temperature of technicians' smart badges or wearable devices are collected for auxiliary assessment of workload and fatigue. Geographic information data acquisition involves integrating enterprise-level geographic information system services to obtain vector topology of heating pipeline networks, coordinates of valves and heat exchange stations, and historical polygon data of construction restricted areas; and by calling third-party map platform APIs to obtain real-time road network vector data, traffic restriction information, and precise coordinates of points of interest. Dynamic environmental data acquisition involves subscribing to meteorological department data service interfaces to obtain gridded weather forecasts and real-time alerts, including temperature, precipitation, wind speed, and visibility. By analyzing real-time traffic flow data released by traffic management departments or road condition APIs from commercial navigation services, road speed and congestion index can be obtained; and time-series monitoring data from pressure, flow, and temperature sensors deployed at key pipeline nodes can be aggregated through an IoT platform.
3. The intelligent dispatching method for heating work orders based on multi-objective optimization according to claim 1, characterized in that, In step S2, the multi-objective optimization model that integrates domain knowledge specifically includes: Objective function 1: Minimize risk-weighted response time Where, ρ i ω represents the potential risk coefficient of a work order based on knowledge graph reasoning. i For the priority weight of the work order, t i start t represents the actual start time of processing. i expect The user's expected response time; Objective function 2: Minimize total lifecycle operating costs Where η(traffic) is the real-time traffic condition impact factor, c carbon ⋅E ij The cost of carbon emissions based on estimated energy consumption; Objective function 3: Maximize the skill fit of knowledge enhancement Where KG-Match(.) is a matching function based on a knowledge graph in the heating field, Fi is the feature vector of the work order, and P j H is the vector image of the technician. ij This is a record of their historical collaboration. Objective function four: Maximize resilience-driven dynamic load balancing. Among them, W t Let Θ(.) be the technician load vector at time t, and let Θ(.) be the parameter representing the technician's real-time fatigue level ΔL. j,t The decay function; Objective Function 5: Maximize the service reliability of multi-source information fusion Among them, R ij base Based on reliability, λ i weather As a weather risk factor, Π ij spare This is the backup support coefficient calculated based on the status of nearby technicians.
4. The intelligent dispatching method for heating work orders based on multi-objective optimization according to claim 1, characterized in that, In step S3, the knowledge graph-guided hybrid metaheuristic algorithm includes: a. In the initialization phase, an initial population is generated using a case-based reasoning strategy, where the case library is constructed from historical high-quality scheduling schemes; b. In the evolutionary stage, an improved NSGA-III framework is used for iterative optimization, in which the selection probability of crossover and mutation operators is dynamically adjusted by the domain rules of real-time reasoning of the knowledge graph; c. In the guidance phase, in each generation of evolution, the graph neural network is used to embed the relationship between "fault-skill-equipment" in the knowledge graph, and the generated embedding vector is used as auxiliary gradient information to guide the population to search for high-quality solution regions. d. In the collaborative phase, a parallel adaptive local search thread is introduced to perform fine-tuning of the elite solution based on the constraint satisfaction problem, and exchanges information with the global evolution thread through a shared memory pool.
5. The intelligent dispatching method for heating work orders based on multi-objective optimization according to claim 1, characterized in that, In step S4, the dynamic rescheduling step specifically includes: a. Initiate the rescheduling process using a fixed time window, or events such as newly added emergency work orders and technician task delays; b. Treat the currently executing and uninterrupted tasks as fixed constraints, and unfreeze the tasks that have not yet started as optimizable variables to form a rolling time-domain optimization problem. c. A lightweight model predictive control algorithm is adopted, with minimizing scheduling disturbances as one of the primary objectives, to quickly solve the optimization problem in the finite time domain; d. Only issue the adjusted route instructions to the affected technicians to ensure the overall stability of the dispatching system.
6. The intelligent dispatching method for heating work orders based on multi-objective optimization according to claim 1, characterized in that, In step S5, the decision support interface includes: A Pareto front visualization dashboard with multi-view fusion, including at least a parallel coordinate plot, a 3D scatter matrix, and a radar chart; An interactive module integrating implicit preference learning and explicit rule setting, used to capture decision-maker intent; An interpretable AI module based on counterfactual reasoning is used to generate comparative explanations for recommendation schemes; A three-point collaborative decision-making workflow that supports asynchronous annotation by multiple experts and consensus calculation.
7. A heating work order intelligent dispatching system based on multi-objective optimization, characterized in that, For implementing the intelligent dispatching method for heating work orders based on multi-objective optimization as described in any one of claims 1-6, the system comprises: The data perception and fusion layer is used to collect and standardize multi-source heterogeneous data in real time through IoT devices, customer service system interfaces, and third-party data APIs; The digital twin and simulation layer constructs a virtual heating service environment that includes physical entities, operating rules, and random events, for scenario simulation and risk assessment. The core optimization and decision-making layer includes a knowledge graph management module, a multi-objective optimization solution engine, a dynamic rescheduler, and a human-machine collaborative decision-making module. The solution execution and feedback layer is responsible for distributing the optimized solution to the technicians' mobile terminals and collecting real-time feedback data during the execution process to form an optimization closed loop.
8. The intelligent dispatching system for heating work orders based on multi-objective optimization according to claim 7, characterized in that, The knowledge graph management module is specifically used for: Build and maintain a domain knowledge graph covering heating equipment, fault types, repair processes, technician skills, and spare parts and materials; Based on the work order text description, natural language processing technology is used to automatically extract entities and relationships for real-time expansion of the knowledge graph; It provides graph-based reasoning services, supporting root cause inference for complex faults and multi-skill collaborative dispatch suggestions.
9. The intelligent dispatching system for heating work orders based on multi-objective optimization according to claim 7, characterized in that, The dynamic rescheduler is implemented using a microservice architecture, including: Event listeners continuously monitor the work order status stream, technician GPS location stream, and external event stream; Impact evaluator, based on network flow model, quickly assesses the impact range of sudden events on the current scheduling plan; A fast solver with an embedded lightweight optimization algorithm for responding to rescheduling requests in seconds; The instruction differential compares the old and new scheduling schemes and generates a minimum set of incremental adjustment instructions.
10. The intelligent dispatching system for heating work orders based on multi-objective optimization according to claim 7, characterized in that, The human-machine collaborative decision-making module provides three switchable working modes: In AI-driven mode, the system automatically recommends the optimal solution and executes it directly, which is suitable for normal working conditions. In the loop mode, the system provides a finite number of optimal solutions and their visual explanations, with the final decision made by the dispatcher. In the manual-led mode, the dispatcher can manually adjust the plan, and the system will then perform local optimizations based on the manual plan and suggest areas for improvement.