An agent-based method and system for automated decision and control of a heating system

Abstract

This invention discloses an automated decision-making and control method and system for heating systems based on intelligent agents, belonging to the field of intelligent regulation of heating systems. Addressing the problems of fuzzy scheduling decisions, reliance on human experience, and security risks in existing technologies, this invention constructs a heating topology knowledge graph by acquiring basic data. Starting from the scheduling target node, it performs topology-constrained graph enhancement retrieval to extract physical subgraphs. Based on the physical subgraphs, a structured scheduling context is generated, and a code-based intelligent agent automatically generates strategy code. The strategy code is input into an automated sandbox environment for simulation execution and security assertion verification. After verification, it is converted into control commands for execution. This invention achieves fully automated closed-loop control of the heating system from perception, decision-making, verification to execution, improving the safety and executability of scheduling decisions.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for heating systems, and particularly relates to an automated decision-making and control method and system for heating systems based on intelligent agents. Background Technology

[0002] With the continuous expansion of urban centralized heating, heating systems are gradually exhibiting characteristics such as complex pipe network structures, a large number of equipment, and highly dynamic operating conditions. In actual operation, heating systems generally exhibit large time lags, strong coupling, and nonlinear characteristics, and the requirements for safety and energy efficiency are constantly increasing. Traditional operation and scheduling methods relying on manual experience or simple control rules are no longer sufficient to meet the needs of refined and intelligent operation. Existing heating scheduling technologies mainly include the following categories: First, methods based on traditional mechanisms of control. These methods are usually based on empirical formulas or simplified physical models, often using PID control or preset temperature curves to regulate the heating system. Their shortcomings include difficulty in accurately describing the nonlinear and time-varying characteristics of the heating system, high dependence on operator experience, difficulty in achieving unified scheduling optimization across regions and operating conditions, and limited adaptability to sudden disturbances and abnormal operating conditions. Second, there are predictive control methods based on deep learning. These methods train deep neural network models on historical operating data to predict heat load or supply and return water temperatures and adjust accordingly. However, their shortcomings include poor model interpretability, difficulty in clearly defining the physical basis of adjustment decisions, and the fact that model outputs are usually predicted values ​​that cannot directly generate executable control strategies. Furthermore, they lack robustness in scenarios involving equipment failure or sensor malfunctions. Third, there are auxiliary scheduling methods based on large language models. Existing patents attempt to introduce large language models to provide scheduling personnel with operational suggestions or textual guidance. However, these methods typically suffer from the following problems: the output of large models is mostly natural language descriptions, such as appropriately closing valves to increase supply water temperature, which are vague and semantically ambiguous, leading to non-executability issues. They also lack explicit modeling of the physical topology of the heating system, easily overlooking the hydraulic coupling relationships between equipment, making it difficult to form an automatic execution and feedback closed-loop control process, still requiring manual intervention. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes an automated decision-making and control method and system for heating systems based on intelligent agents, thereby resolving the issues present in the prior art.

[0004] In a first aspect, to achieve the above objectives, the present invention provides an automated decision-making and control method for a heating system based on an intelligent agent, comprising the following steps: Acquire basic data of the heating system, including heat source information, pipeline network information, heating station information, heating equipment ledger information, pipeline connection information, end-user information, as well as operation monitoring data and meteorological data; A heating topology knowledge graph is constructed based on the aforementioned basic data, where nodes represent entity objects in the heating system, edges represent physical or logical relationships between nodes, and corresponding operating parameters are associated with the nodes and edges. Upon receiving a scheduling request or detecting an operational anomaly, the node corresponding to the scheduling target is used as the starting point. Based on the hydraulic connection direction between nodes in the heating topology knowledge graph, a topology-constrained graph enhancement retrieval is performed to extract a physical subgraph containing upstream and downstream dependencies. A structured scheduling context is generated based on the physical subgraph, and the scheduling context is input into the code agent, which automatically generates policy code under preset interface specifications and control constraints. The strategy code is input into an automated sandbox environment for simulated execution, and the simulation execution results are verified based on preset security assertions. After the strategy code passes all security assertion checks, it is converted into control commands and sent to the heating system for execution.

[0005] Optionally, the process of obtaining basic data for the heating system includes: Collect information on heat sources, pipeline networks, heating stations, heating equipment ledgers, pipeline connections, end-users, as well as real-time operational monitoring data and meteorological data; The real-time operation monitoring data includes supply and return water temperature, pressure, flow rate, valve opening, pump frequency, and user indoor temperature; The meteorological data includes outdoor temperature, humidity, wind direction, wind force, air pressure, cloud cover, and weather conditions.

[0006] Optionally, the process of constructing a heating topology knowledge graph includes: Nodes represent heat sources, pipeline nodes, heat exchange stations, and user-side temperature monitoring nodes in the heating system; Edges are used to represent the hydraulic connections or heating service relationships between nodes; The flow-direction edges in the edges are associated with pipe segment length, network resistance, and transmission lag time parameters, and each node is associated with operating parameters and meteorological parameters.

[0007] Optionally, the process of performing topology-constrained graph-enhanced retrieval includes: Starting from the node corresponding to the scheduling target, trace upstream nodes that have hydraulic coupling with the target node in the reverse direction of the hydraulic connection relationship, and simultaneously trace downstream nodes that may be affected by the regulation behavior in the forward direction of the hydraulic connection relationship. Construct a physical subgraph containing upstream and downstream dependencies within a preset number of hops.

[0008] Optionally, the process of automatically generating policy code by the code agent includes: The code agent generates perception logic, judgment logic, and control logic based on the structured scheduling context. The perception logic is used to obtain the current running state of each node in the physical subgraph. The judgment logic is used to analyze the heating conditions based on the running state, meteorological conditions, and engineering rules. The control logic is used to generate specific control parameters.

[0009] Optionally, the process of simulating execution and performing security assertion verification on the policy code includes: The policy code is input into an automated sandbox environment for syntax validation, interface validation, and runtime validation based on security assertions; If the simulation execution result violates any security assertion, the policy code is prohibited from entering the execution phase, and an exception message is fed back to the code agent to trigger the regeneration of the policy code.

[0010] Secondly, the present invention also provides an agent-based automated decision-making and control system for a heating system, for implementing an agent-based automated decision-making and control method for a heating system, the system comprising: Acquire basic data of the heating system, including heat source information, pipeline network information, heating station information, heating equipment ledger information, pipeline connection information, end-user information, as well as operation monitoring data and meteorological data; A heating topology knowledge graph is constructed based on the aforementioned basic data, where nodes represent entity objects in the heating system, edges represent physical or logical relationships between nodes, and corresponding operating parameters are associated with the nodes and edges. Upon receiving a scheduling request or detecting an operational anomaly, the node corresponding to the scheduling target is used as the starting point. Based on the hydraulic connection direction between nodes in the heating topology knowledge graph, a topology-constrained graph enhancement retrieval is performed to extract a physical subgraph containing upstream and downstream dependencies. A structured scheduling context is generated based on the physical subgraph, and the scheduling context is input into the code agent, which automatically generates policy code under preset interface specifications and control constraints. The strategy code is input into an automated sandbox environment for simulated execution, and the simulation execution results are verified based on preset security assertions. After the strategy code passes all security assertion checks, it is converted into control commands and sent to the heating system for execution.

[0011] Thirdly, the present invention also provides a computer terminal device, comprising: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the agent-based automated decision-making and control method for heating systems described in the first aspect above.

[0012] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the agent-based automated decision-making and control method for heating systems described in the first aspect above.

[0013] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the agent-based automated decision-making and control method for heating systems described in the first aspect above.

[0014] Compared with the prior art, the present invention has the following advantages and technical effects: This invention provides an automated decision-making and control method and system for heating systems based on intelligent agents. By constructing a topological knowledge graph of the heating system and introducing a graph enhancement retrieval mechanism, this invention generates physical subgraphs strongly correlated with scheduling objectives under topological constraints, explicitly preserving the hydraulic coupling relationships and transmission lag characteristics between devices, thus ensuring physical consistency in scheduling decisions. By introducing a code-based intelligent agent to generate scheduling strategies in code form, the semantic ambiguity problems existing in natural language instructions are eliminated, making scheduling decisions compilable, verifiable, and reproducible. An automated sandbox verification module simulates and verifies the strategy code, identifying and blocking potential risky operations before deployment, ensuring the safe operation of the heating system. This invention achieves fully automated closed-loop control from perception, decision-making, verification to execution, improving the adaptive capability and operational stability of the heating system and reducing reliance on manual scheduling experience. Attached Figure Description

[0015] 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 undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the hierarchical architecture of the intelligent agent-based automated decision-making and control system for heating systems according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the topological constraint-based graph enhancement retrieval process according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the closed-loop control process for code agent generation and sandbox verification in an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0018] Example 1 like Figure 1 As shown, this invention is based on the actual physical structure and operating state of the heating system. It performs structured modeling of the heating system through a topological knowledge graph. In the scheduling decision-making stage, a graph-enhanced retrieval method with topological constraints is introduced to limit the scheduling decision space to physical subsystems that are strongly related to the target operating conditions. At the same time, a code-based intelligent agent is introduced to generate scheduling strategies in code form, and the strategies are verified and corrected in an automated sandbox environment. Ultimately, a safe and executable automated scheduling control of the heating system is achieved.

[0019] The automated decision-making and control system for heating systems of the present invention comprises, from top to bottom, a data acquisition layer, a topological knowledge graph layer, an agent decision-making layer, and an execution and control layer. These layers interact and coordinate control through standardized interfaces. This embodiment provides an agent-based automated decision-making and control method for heating systems, including: Acquire basic data of the heating system, including heat source information, pipeline network information, heating station information, heating equipment ledger information, pipeline connection information, end-user information, as well as operation monitoring data and meteorological data; A heating topology knowledge graph is constructed based on the aforementioned basic data, where nodes represent entity objects in the heating system, edges represent physical or logical relationships between nodes, and corresponding operating parameters are associated with the nodes and edges. Upon receiving a scheduling request or detecting an operational anomaly, the node corresponding to the scheduling target is used as the starting point. Based on the hydraulic connection direction between nodes in the heating topology knowledge graph, a topology-constrained graph enhancement retrieval is performed to extract a physical subgraph containing upstream and downstream dependencies. A structured scheduling context is generated based on the physical subgraph, and the scheduling context is input into the code agent, which automatically generates policy code under preset interface specifications and control constraints. The strategy code is input into an automated sandbox environment for simulated execution, and the simulation execution results are verified based on preset security assertions. After the strategy code passes all security assertion checks, it is converted into control commands and sent to the heating system for execution.

[0020] As one implementation method in this embodiment, the process of obtaining basic data of the heating system includes: Collect information on heat sources, pipeline networks, heating stations, heating equipment ledgers, pipeline connections, end-users, as well as real-time operational monitoring data and meteorological data; The real-time operation monitoring data includes supply and return water temperature, pressure, flow rate, valve opening, pump frequency, and user indoor temperature; The meteorological data includes outdoor temperature, humidity, wind direction, wind force, air pressure, cloud cover, and weather conditions.

[0021] As one implementation method in this embodiment, the process of constructing a heating topology knowledge graph includes: Nodes represent heat sources, pipeline nodes, heat exchange stations, and user-side temperature monitoring nodes in the heating system; Edges are used to represent the hydraulic connections or heating service relationships between nodes; The flow-direction edges in the edges are associated with pipe segment length, network resistance, and transmission lag time parameters, and each node is associated with operating parameters and meteorological parameters.

[0022] Specifically, nodes are used to represent physical objects or monitoring objects in the heating system, including at least heat source nodes, pipeline nodes, heat exchange station nodes, and user-side temperature monitoring nodes. Edges are used to represent the physical or logical relationships between nodes, including at least flow edges that represent the direction of heat medium transmission and energy supply edges that represent the relationship of heating services.

[0023] Each node is associated with corresponding operating parameters and meteorological parameters. The operating parameters include at least one or more of temperature, pressure, flow rate, valve opening degree, or pump frequency. The flow direction edge is further associated with pipe segment length, pipeline resistance, and transmission lag time parameters.

[0024] The topological knowledge graph constructed in the above manner can fully reflect the physical connection relationships and operational characteristics of the heating system.

[0025] As one implementation method in this embodiment, the process of performing topology-constrained graph enhancement retrieval includes: Starting from the node corresponding to the scheduling target, trace upstream nodes that have hydraulic coupling with the target node in the reverse direction of the hydraulic connection relationship, and simultaneously trace downstream nodes that may be affected by the regulation behavior in the forward direction of the hydraulic connection relationship. Construct a physical subgraph containing upstream and downstream dependencies within a preset number of hops.

[0026] A graph-enhanced retrieval module based on topological constraints; After the topological knowledge graph is constructed, when the system receives a scheduling request or detects an operational anomaly, it uses the node corresponding to the scheduling target as the starting point for retrieval and performs a topologically constrained graph enhancement retrieval operation based on the hydraulic connection direction of the heating system, such as... Figure 2 As shown. In a preferred embodiment, the retrieval operation includes: 1) Trace upstream nodes that have a hydraulic coupling relationship with the target node in the reverse direction of the hydraulic connection relationship; 2) Track downstream nodes that may be affected by regulatory behavior along the positive direction of the hydraulic connection; 3) Construct a physical subgraph containing upstream and downstream dependencies within a preset number of hops.

[0027] The physical subgraph obtained in the above manner explicitly retains the connection relationships and transmission lag characteristics between devices, thereby ensuring physical consistency in subsequent scheduling decisions.

[0028] As one implementation method in this embodiment, the process of automatically generating policy code by the code intelligence agent includes: The code agent generates perception logic, judgment logic, and control logic based on the structured scheduling context. The perception logic is used to obtain the current running state of each node in the physical subgraph. The judgment logic is used to analyze the heating conditions based on the running state, meteorological conditions, and engineering rules. The control logic is used to generate specific control parameters.

[0029] Scheduling strategy generation module based on code-based intelligent agents: After obtaining the physical subgraph, the system constructs a structured scheduling context and inputs it into the code agent. The code agent automatically generates strategy code for heating system control under preset interface specifications and control constraints. In a preferred embodiment, the strategy code includes at least: 1) The perception logic is used to acquire real-time operational data of each node within the physical subgraph from the heating system operation monitoring platform, and to uniformly organize and standardize the acquired data, thereby providing a data foundation for subsequent operating condition analysis and control decisions. In a preferred embodiment, the perception logic retrieves the operating status information of nodes related to the physical subgraph from the system database by calling a preset data reading interface. The operating status information includes at least parameters such as supply water temperature, return water temperature, pipeline pressure, flow rate, regulating valve opening, circulating pump operating frequency, and user-side indoor temperature. Simultaneously, the perception logic can also acquire meteorological data related to the heating load, such as outdoor temperature and wind speed, and encapsulate them together with the system operation data to form a set of state variables required for strategy execution. Through this method, the perception logic can achieve comprehensive perception of the operating status of the heating system within the physical subgraph and ensure the real-time performance and consistency of the data used in subsequent scheduling strategy calculations.

[0030] 2) Judgment logic, used to comprehensively analyze and judge the current operating condition of the heating system based on the operating status, meteorological conditions, and engineering rules obtained from the perception logic, thereby determining whether adjustment is needed and the specific direction of adjustment. In a preferred embodiment, the judgment logic first identifies areas of insufficient or excessive heating based on the deviation between the indoor temperature on the user side and the preset comfort temperature range; then, combined with meteorological conditions such as outdoor temperature, it calculates and evaluates the current heating load level and determines whether the system is operating under high or low load conditions. In addition, the judgment logic also analyzes the hydraulic balance of the heating system based on pipeline pressure, flow rate, and the operating status of heat exchange stations to identify whether there are uneven flow distribution or local hydraulic imbalance problems. To ensure that the analysis results conform to engineering reality, this invention introduces heating operation engineering rules into the judgment logic, such as the empirical relationship between supply water temperature and outdoor temperature, heat exchange station load distribution principles, and pipeline operation safety rules. By comprehensively considering operating data, meteorological conditions, and engineering rules, the judgment logic can form a structured judgment result on the current operating condition of the heating system and provide a basis for subsequent control parameter calculations.

[0031] 3) Control logic, used to generate specific control parameters based on the operating condition analysis results obtained from the judgment logic, and to serve as the control basis for the heating system's regulating equipment. In a preferred embodiment, the control logic calculates the adjustment amounts of key control variables such as the opening degree of the regulating valve at the heat exchange station, the operating frequency of the circulating pump, and the temperature of the heat source water supply, based on the heating system's operating target and the judgment results. For example, when the judgment logic identifies that the room temperature of users in a certain area is too low, the control logic can improve the heating capacity of that area by appropriately increasing the valve opening degree of the corresponding heat exchange station or increasing the frequency of the circulating pump; when the system is operating at low load, the water supply temperature can be appropriately reduced or the pump operating intensity can be reduced to reduce energy consumption. To avoid the adjustment process affecting the stability of the pipeline network, a gradual adjustment mechanism is also introduced into the control logic to limit the range of change of control variables, such as limiting the single change in valve opening degree and the rate of change in pump frequency, thereby reducing the risk of water hammer and ensuring system pressure stability. At the same time, the control logic must also meet preset safety operating constraints when generating control parameters, such as the upper limit of water supply temperature, the limit of pressure change rate, and the limit of equipment adjustment range. In this way, the control logic can generate executable control parameters while meeting safety constraints, providing a specific basis for the automated scheduling of the heating system.

[0032] By expressing scheduling strategies in code, ambiguity issues in natural language instructions can be eliminated, and scheduling decisions can be compiled, verified, and reproduced.

[0033] It also includes: the code-based intelligent agent is implemented based on a large code model and is subject to preset interface specifications and control constraints. The strategy code generated by the code-based intelligent agent includes at least perception logic, judgment logic, and control logic, wherein the perception logic is used to obtain the node's operating status, the judgment logic is used to analyze the heating conditions, and the control logic is used to generate specific control parameters.

[0034] As one implementation method in this embodiment, the process of simulating execution and verifying security assertions of the policy code includes: The policy code is input into an automated sandbox environment for syntax validation, interface validation, and runtime validation based on security assertions; If the simulation execution result violates any security assertion, the policy code is prohibited from entering the execution phase, and an exception message is fed back to the code agent to trigger the regeneration of the policy code.

[0035] Security control module based on sandbox verification: To prevent automatically generated policy code from causing security risks during actual operation, this invention sets up automated sandbox verification before policy execution to simulate the operation and security verification of the policy code, such as... Figure 3 As shown.

[0036] In a preferred embodiment, sandbox verification includes at least syntax verification, interface verification, and runtime verification based on security assertions. When the simulated execution result violates any security assertion, the system prevents the policy from entering the actual execution phase and feeds back the exception information to the code agent to trigger the regeneration of the policy code.

[0037] Automatic execution and closed-loop optimization module: Once the strategy code passes the sandbox verification, the system converts it into actual control instructions and sends them to the heating system for execution. During execution, the system continuously collects operational data and updates it to the topology knowledge graph, thereby achieving closed-loop feedback and continuous optimization of scheduling decisions.

[0038] Through the above methods, the present invention realizes closed-loop control of scheduling decision generation, verification and execution, and can continuously optimize the scheduling effect based on long-term operating data, thereby improving the overall operational stability and energy efficiency of the heating system.

[0039] It also includes: sandbox verification for simulating execution and verifying security assertions on the policy code. It is configured to report an exception to the code agent when a security assertion failure is detected, thereby triggering the regeneration of the policy code.

[0040] Safety assertions are used to limit the permissible range of operating parameters of a heating system, including at least the upper limit of the supply water temperature, the rate of pressure change, and the magnitude of changes in the parameters of regulating equipment.

[0041] This embodiment uses a centralized heating area in a city as an application scenario for illustration. This heating area includes two heat source plants, 220 heat exchange stations, several main and branch heating pipelines, and approximately 30,000 end-user heat users. The system adopts a layered architecture deployment, including a dispatch center server, a graph database server, a strategy generation and inference server, an automated sandbox verification environment, and edge controllers deployed at each heat exchange station. The dispatch center server communicates with the edge controllers at each heat exchange station via industrial Ethernet. The edge controllers connect to field devices such as regulating valves, circulating pumps, and temperature, pressure, and flow sensors via fieldbus, thereby enabling real-time acquisition of heating operation data and execution of control commands.

[0042] During system operation, the system first acquires basic data about the heating system through a data acquisition module. This data includes heat source information, pipeline connection relationships, equipment ledger information, operational data from multiple heat exchange stations, indoor temperature data on the user side, and outdoor meteorological data. The pipeline spatial topology is obtained through a GIS system, real-time operational data is collected through a SCADA system, and static equipment attributes are provided by the equipment management system. After unified processing, this data is input into a heating topology knowledge graph construction module. Nodes represent heat sources, pipeline nodes, heat exchange stations, and user-side monitoring objects, while edges represent the hydraulic connections and energy supply relationships within the heating system. Edge attributes record parameters such as flow direction, pipe length, hydraulic resistance, and transmission lag time. Each node is continuously bound to real-time operational parameters, achieving a structured representation of the heating system's operational status.

[0043] When the monitoring module detects an operational anomaly, such as a sustained drop in indoor temperature below a set threshold for users at multiple heat exchange stations within a certain area, the system automatically triggers a joint scheduling task. Unlike single-station control, this embodiment treats multiple heat exchange stations as joint control objects and identifies the hydraulic coupling relationships between them based on a heating topology knowledge graph. The graph enhancement retrieval module uses multiple target heat exchange station nodes as the starting point for joint retrieval, performing topology-constrained search operations along the flow direction within a preset hop count range. This includes tracing back along the flow direction to the common upstream heat source and main pipe segment nodes, and extending forward along the flow direction to each downstream user node, thereby extracting a joint physical subgraph containing multiple heat exchange stations and their shared pipe network resources. This physical subgraph not only contains the operating status data of each device but also retains the coupling influence relationships and transmission lag characteristics between heat exchange stations through the pipe network, and is converted into unified structured scheduling context data.

[0044] Subsequently, the strategy generation module automatically generates multi-station collaborative control strategy code based on the joint physical subgraph. The control strategy code includes perception logic, judgment logic, and control logic. The perception logic acquires real-time operating parameters of all relevant nodes in the joint subgraph; the judgment logic analyzes the load distribution relationship, hydraulic balance state, and terminal room temperature deviation among multiple heat exchange stations; and the control logic calculates the adjustment amounts of the regulating valve opening and circulating pump frequency for each heat exchange station according to the joint optimization principle. In this process, the control strategy not only considers the local objectives of a single station but also achieves multi-station collaborative optimization through shared network constraints to avoid problems such as decreased flow rates at other heat exchange stations or system pressure fluctuations due to single-station adjustments. The control strategy employs a gradual adjustment mechanism to limit the rate of change of control variables, thereby reducing the risk of water hammer and improving system stability.

[0045] To ensure the operational safety of the automatically generated control strategy, the strategy code is submitted to an automated sandbox verification environment for simulation before execution. The sandbox environment first performs syntax checks and interface consistency checks, then injects snapshot data of the current heating system operation for simulation execution, and verifies key parameters based on preset safety assertions, including upper limit constraints on supply water temperature, pressure change rate limits, and ensuring that the total flow allocation does not exceed the main pipeline capacity during multi-station joint regulation. When a safety assertion is detected as unmet, the system generates an anomaly feedback message and returns it to the strategy generation module for automatic correction of the control logic and regeneration of the strategy, thus forming a closed-loop process of strategy generation and verification.

[0046] Once the control strategy passes all security assertion verifications, the system compiles the strategy code into an executable instruction sequence for the edge controllers and simultaneously distributes it to the corresponding edge controllers at multiple heat exchange stations via the industrial communication network. Each edge controller adjusts the opening of the regulating valve and the operating frequency of the circulating pump step-by-step according to a unified scheduling strategy and continuously collects on-site operating data and feeds it back to the dispatch center. The system updates the real-time operating results to the heating topology knowledge graph, achieving dynamic updates to the heating system's operating status, thus forming a closed-loop control process of "data acquisition—topology perception—joint decision-making—security verification—collaborative execution—feedback optimization".

[0047] Through the above implementation methods, under the premise of ensuring the safe operation of the heating system, joint control based on the topological coupling relationship between multiple heat exchange stations can be realized, effectively avoiding systemic disturbances caused by local regulation, improving the stability of terminal heating and overall operating efficiency, and reducing the dependence on manual scheduling experience.

[0048] Based on this, this invention provides an agent-based automated decision-making and control method for heating systems. This invention constructs a topological knowledge graph of the heating system and introduces a graph enhancement retrieval mechanism. Under topological constraints, it generates physical subgraphs strongly correlated with scheduling objectives, explicitly preserving the hydraulic coupling relationships and transmission lag characteristics between devices, thus ensuring physical consistency in scheduling decisions. By introducing a code-based intelligent agent to generate scheduling strategies in code form, it eliminates the semantic ambiguity problems present in natural language instructions, making scheduling decisions compilable, verifiable, and reproducible. An automated sandbox verification module simulates and verifies the strategy code, identifying and blocking potential risky operations before deployment, ensuring the safe operation of the heating system. This invention achieves fully automated closed-loop control from perception, decision-making, verification to execution, improving the adaptive capability and operational stability of the heating system and reducing reliance on manual scheduling experience.

[0049] Example 2 In this embodiment, a computer terminal device is provided, including: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the above-described agent-based automated decision-making and control method for heating systems.

[0050] In this embodiment, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the above-described intelligent agent-based automated decision-making and control method for heating systems.

[0051] In this embodiment, an electronic device is also provided, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to execute the steps of the above-described agent-based automated decision-making and control method for heating systems.

[0052] In this embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the above-described agent-based automated decision-making and control method for heating systems.

[0053] The aforementioned program can run on a processor or be stored in memory (or a computer-readable medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0054] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented by different modules for different steps.

[0055] This embodiment provides such a device or system. The system, referred to as an agent-based automated decision-making and control system for heating systems, includes: The data acquisition module is used to acquire basic data of the heating system, including heat source information, pipeline network information, heating station information, heating equipment ledger information, pipeline connection information, end-user information, as well as operation monitoring data and meteorological data. The knowledge graph construction module is used to construct a heating topology knowledge graph based on the basic data, wherein nodes represent entity objects in the heating system, edges represent physical or logical relationships between nodes, and corresponding operating parameters are associated with the nodes and edges. The graph retrieval module is used to extract physical subgraphs containing upstream and downstream dependencies when a scheduling request is received or an operational anomaly is detected, starting from the node corresponding to the scheduling target and performing topologically constrained graph enhancement retrieval based on the hydraulic connection direction between nodes in the heating topology knowledge graph. The strategy generation module is used to generate a structured scheduling context based on the physical subgraph and input the scheduling context into the code agent, which then automatically generates strategy code under preset interface specifications and control constraints. The sandbox verification module is used to input the policy code into an automated sandbox environment for simulated execution, and to verify the simulation execution results based on preset security assertions. The execution module is used to convert the policy code into control commands and send them to the heating system for execution after the policy code has passed all security assertion verifications.

[0056] As one implementation method in this embodiment, the data acquisition module includes: The basic data acquisition unit is used to collect information on heat sources, pipeline networks, heating stations, heating equipment ledgers, pipeline connections, and end-users. The operation data acquisition unit is used to collect real-time operation monitoring data, including supply and return water temperature, pressure, flow rate, valve opening, pump frequency, and user indoor temperature. The meteorological data acquisition unit is used to collect meteorological data including outdoor temperature, humidity, wind direction, wind force, air pressure, cloud cover, and weather conditions.

[0057] As one implementation method in this embodiment, the knowledge graph construction module includes: The node definition unit is used to define nodes as heat sources, pipeline nodes, heat exchange stations, and user-side temperature monitoring nodes in the heating system. Edge definition unit, used to define the hydraulic connection relationship or heating service relationship between nodes; The parameter association unit is used to associate pipe segment length, network resistance and transmission lag time parameters for the flow-direction edge representing the hydraulic connection relationship, and to associate operating parameters and meteorological parameters for each node.

[0058] As one implementation method in this embodiment, the map retrieval module includes: The retrieval starting point determination unit is used to take the node corresponding to the scheduling target as the retrieval starting point; The reverse tracing unit is used to trace upstream nodes that have a hydraulic coupling relationship with the target node in the reverse direction of the hydraulic connection relationship; A forward tracking unit is used to track downstream nodes that may be affected by regulatory behavior along the forward direction of the hydraulic connection. The subgraph construction unit is used to construct a physical subgraph containing upstream and downstream dependencies within a preset number of hops.

[0059] As one implementation method in this embodiment, the strategy generation module includes: The context generation unit is used to generate a structured scheduling context based on the physical subgraph; the code intelligence unit is used to receive the scheduling context and automatically generate strategy code under preset interface specifications and control constraints. The strategy code includes perception logic, judgment logic and control logic, wherein the perception logic is used to obtain the current running state of each node in the physical subgraph, the judgment logic is used to analyze the heating conditions based on the running state, meteorological conditions and engineering rules, and the control logic is used to generate specific control parameters.

[0060] As one implementation method in this embodiment, the sandbox verification module includes: The simulation execution unit is used to input strategy code into an automated sandbox environment for simulated execution; The security verification unit is used to perform syntax verification, interface verification, and runtime verification based on security assertions on the simulated execution results. The exception feedback unit is used to prevent the policy code from entering the execution phase when the simulation execution result violates any security assertion, and to feed back exception information to the policy generation module to trigger the regeneration of the policy code.

[0061] The system or apparatus is used to implement the functions of the methods in the above embodiments. Each module in the system or apparatus corresponds to each step in the method, as has been described in the method and will not be repeated here.

[0062] The above implementation method solves the problem of automated decision-making and control of heating systems based on intelligent agents in related technologies, thereby ensuring that the problems existing in the prior art are resolved.

[0063] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An automated decision-making and control method for a heating system based on intelligent agents, characterized in that, Includes the following steps: Acquire basic data of the heating system, including heat source information, pipeline network information, heating station information, heating equipment ledger information, pipeline connection information, end-user information, as well as operation monitoring data and meteorological data; A heating topology knowledge graph is constructed based on the aforementioned basic data, where nodes represent entity objects in the heating system, edges represent physical or logical relationships between nodes, and corresponding operating parameters are associated with the nodes and edges. Upon receiving a scheduling request or detecting an operational anomaly, the node corresponding to the scheduling target is used as the starting point. Based on the hydraulic connection direction between nodes in the heating topology knowledge graph, a topology-constrained graph enhancement retrieval is performed to extract a physical subgraph containing upstream and downstream dependencies. A structured scheduling context is generated based on the physical subgraph, and the scheduling context is input into the code agent, which automatically generates policy code under preset interface specifications and control constraints. The strategy code is input into an automated sandbox environment for simulated execution, and the simulation execution results are verified based on preset security assertions. After the strategy code passes all security assertion checks, it is converted into control commands and sent to the heating system for execution.

2. The method according to claim 1, characterized in that, The process of obtaining basic data for a heating system includes: Collect information on heat sources, pipeline networks, heating stations, heating equipment ledgers, pipeline connections, end-users, as well as real-time operational monitoring data and meteorological data; The real-time operation monitoring data includes supply and return water temperature, pressure, flow rate, valve opening, pump frequency, and user indoor temperature; The meteorological data includes outdoor temperature, humidity, wind direction, wind force, air pressure, cloud cover, and weather conditions.

3. The method according to claim 1, characterized in that, The process of constructing a heating topology knowledge graph includes: Nodes represent heat sources, pipeline nodes, heat exchange stations, and user-side temperature monitoring nodes in the heating system; Edges are used to represent the hydraulic connections or heating service relationships between nodes; The flow-direction edges in the edges are associated with pipe segment length, network resistance, and transmission lag time parameters, and each node is associated with operating parameters and meteorological parameters.

4. The method according to claim 1, characterized in that, The process of performing topology-constrained graph-enhanced retrieval includes: Starting from the node corresponding to the scheduling target, trace upstream nodes that have hydraulic coupling with the target node in the reverse direction of the hydraulic connection relationship, and simultaneously trace downstream nodes that may be affected by the regulation behavior in the forward direction of the hydraulic connection relationship. Construct a physical subgraph containing upstream and downstream dependencies within a preset number of hops.

5. The method according to claim 1, characterized in that, The process of automatically generating policy code by a code-based intelligent agent includes: The code agent generates perception logic, judgment logic, and control logic based on the structured scheduling context. The perception logic is used to obtain the current running state of each node in the physical subgraph. The judgment logic is used to analyze the heating conditions based on the running state, meteorological conditions, and engineering rules. The control logic is used to generate specific control parameters.

6. The method according to claim 1, characterized in that, The process of simulating execution and performing security assertion verification on the policy code includes: The policy code is input into an automated sandbox environment for syntax validation, interface validation, and runtime validation based on security assertions; If the simulation execution result violates any security assertion, the policy code is prohibited from entering the execution phase, and an exception message is fed back to the code agent to trigger the regeneration of the policy code.

7. An automated decision-making and control system for a heating system based on intelligent agents, characterized in that, The system for implementing the method of any one of claims 1-6 comprises: Acquire basic data of the heating system, including heat source information, pipeline network information, heating station information, heating equipment ledger information, pipeline connection information, end-user information, as well as operation monitoring data and meteorological data; A heating topology knowledge graph is constructed based on the aforementioned basic data, where nodes represent entity objects in the heating system, edges represent physical or logical relationships between nodes, and corresponding operating parameters are associated with the nodes and edges. Upon receiving a scheduling request or detecting an operational anomaly, the node corresponding to the scheduling target is used as the starting point. Based on the hydraulic connection direction between nodes in the heating topology knowledge graph, a topology-constrained graph enhancement retrieval is performed to extract a physical subgraph containing upstream and downstream dependencies. A structured scheduling context is generated based on the physical subgraph, and the scheduling context is input into the code agent, which automatically generates policy code under preset interface specifications and control constraints. The strategy code is input into an automated sandbox environment for simulated execution, and the simulation execution results are verified based on preset security assertions. After the strategy code passes all security assertion checks, it is converted into control commands and sent to the heating system for execution.

8. A computer terminal device, characterized in that, include: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images