Heating system optimization planning and operation method based on heterogeneous information network

Abstract

This invention discloses a method for optimizing the planning and operation of a heating system based on a heterogeneous information network, comprising: constructing a heterogeneous information network for the heating system and forming multiple static attribute meta-paths and multiple dynamic attribute meta-paths for the heating system; based on the multiple static attribute meta-paths and prediction model of the heating system, setting a set of planning schemes for each cycle of transformation and construction, and constructing an optimized planning model for the heating system by combining the weights and constraints of each node with the goal of minimizing the transformation and construction cost for each cycle; based on the multiple dynamic attribute meta-paths and prediction model of the heating system, and combining the planning schemes for each cycle of transformation and construction, integrating economic and safety objectives, constructing an optimized operation model for the heating system; and through continuous iterative optimization, while analyzing the impact of the planning scheme and operation scheduling scheme of the previous cycle on the optimization of the next cycle, finally forming the optimal planning scheme and operation scheduling scheme for each cycle of transformation and construction.

Description

Technical Field

[0001] This invention belongs to the field of heating system technology, specifically relating to a method for optimizing the planning and operation of a heating system based on a heterogeneous information network. Background Technology

[0002] The core elements of urban heating systems will continue to change in the medium to long term. For example, the construction of new urban areas, renovation of old areas, and energy-saving renovation of buildings on the heat load side will lead to an increase in the total load or changes in its distribution. On the equipment side, core equipment such as heat sources and pipelines have service cycles and will face aging, retirement, and upgrading. In addition, the energy structure will be transformed from coal to clean energy.

[0003] In the planning and operation of heating systems, the equipment and heat network layout are complex, and the heat transmission and regulation processes are highly coupled. If there is a lack of medium- and long-term planning and operation scheduling, and decisions are made based solely on short-term demand, the system may be feasible in the short term but unstable and unreliable in the long term. Therefore, it is necessary to anticipate changes in advance, carry out system equipment renovation and construction, and introduce new heating technologies to ensure long-term matching of supply and demand, reduce heating safety risks, and balance the cost expenditures of each cycle.

[0004] Heterogeneous information networks (HINFs) are complex network types containing various node types and relationships. They can characterize the interconnected logic of multiple elements in heterogeneous scenarios, transforming various elements into nodes and their interactions, influences, and constraints into connections. This forms a comprehensive system panorama reflecting global interconnections, offering significant technical insights for accurately characterizing the heterogeneity of heating systems and providing interconnected logic for the full-cycle planning and operation of heating systems. Therefore, the key lies in applying HINFs to the planning and operation of heating systems. Through advance prediction, cost control, overall coordination, and dynamic adjustment, HINFs can adapt to various dynamic changes in long-term planning and operation, achieving the core objectives of economic optimization, safe and stable supply, and improved heating reliability.

[0005] Based on the above technical problems, it is necessary to design a new method for optimizing the planning and operation of heating systems based on heterogeneous information networks. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method for optimizing the planning and operation of a heating system based on a heterogeneous information network. To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: The present invention provides a method for optimizing the planning and operation of a heating system based on a heterogeneous information network, which includes: S1, in the heterogeneous information network construction layer, identifying and selecting multiple types of core nodes in the planning and operation process of the heating system, defining the relevant attributes of the nodes, analyzing the correlation between the nodes to form a relation matrix, and then constructing a heterogeneous information network for the heating system; S2, in the network learning layer, using the constructed heterogeneous information network of the heating system combined with historical operation data, embedding learning is performed on each core node, analyzing the parameter changes of a single typical pipe section, the impact of parameter measurement errors on the optimization result parameters of the pipe section and the propagation path, and conducting a chain reaction analysis of the fluctuations in the state parameters of heat source and heat user, forming multiple static attribute meta-paths and multiple dynamic attribute meta-paths of the heating system through topological integration of the heterogeneous information network; S3, in the load supply and demand layer, based on the multiple static and dynamic attribute meta-paths in the heterogeneous information network of the heating system, selecting meta-paths related to the heat load prediction type and meta-paths related to the heat supply prediction, and integrating the supply and demand... Using multi-source data from the heating system's sources, network, and load, machine learning algorithms are employed to establish long-term heat load demand and heat supply prediction models for the heating system. S4. At the optimization planning layer, the heating system's transformation and construction planning cycle is divided into multiple cycles. Based on the multi-type static attribute paths and prediction models of the heating system, a set of planning schemes for each transformation and construction cycle is set. With the goal of minimizing the transformation and construction cost for each cycle, and combining the weights and constraints of each node, an optimization planning model for the heating system is constructed, outputting the planning schemes for each transformation and construction cycle. S5. At the optimization operation layer, based on the multi-type dynamic attribute paths and prediction models of the heating system, and combining the planning schemes for each transformation and construction cycle, an optimization operation model for the heating system is constructed, integrating economic and safety objectives, and outputting the operation scheduling schemes for each transformation and construction cycle. S6. Through iterative optimization at the optimization planning layer, load supply and demand layer, and optimization operation layer, and simultaneously based on the network learning layer, the impact of the planning schemes and operation scheduling schemes of the previous cycle on the optimization of the next cycle is analyzed, ultimately forming the optimal planning schemes and operation scheduling schemes for each transformation and construction cycle.

[0007] Furthermore, S1 specifically includes: S101, identifying and selecting multiple types of core nodes in the planning and operation process of the heating system, including at least: physical entity nodes, parameter attribute nodes, benefit target nodes, and decision behavior nodes; the physical entity nodes include heat sources, heating networks, heat users, heating network equipment, and sensors; the parameter attribute nodes include heating parameters and economic parameters; the benefit target nodes include economic costs and carbon emissions; the decision behavior nodes include planning decisions and operational decisions; S102, defining the relevant attributes of each node, including at least: the output scale of the heat source, energy type and operating efficiency, the heat load scale of the heat user, the heating period and temperature demand, the diameter and length of the heating network pipes, the adjustment accuracy and operating parameters of the heating network equipment, the measurement accuracy and response time of the sensors, and the geographical coordinates and pipe dimensions of each physical entity node. The network topology; and the supply and return water temperatures, pressures, flow rates, heat loads, and heat transfer rates in the heating parameters; energy prices in the economic parameters; system transformation construction and operating costs in the economic costs; carbon emissions and carbon emission intensity in the carbon emissions; construction sequence, transformation cycle, and equipment selection constraints in the planning decisions; and equipment scheduling strategies in the operation decisions; S103, analyze the relationships between nodes, including at least: the relationship between heat sources and the heating network, the relationship between the heating network and heat users, the relationship between heating network equipment and parameter attribute nodes, and the relationship between parameter attribute nodes and benefit target nodes, and form a relationship matrix based on the number of nodes, node attributes, and the relationship between edges between nodes, considering the edge representation values ​​and edge directions; S104, construct a heterogeneous information network for the heating system based on multiple types of core nodes, the relevant attributes of each node, and the relationship matrix between nodes. , For each core node set, The set of attributes possessed by each node. It is the set of relationships between nodes in the network.

[0008] Further, S2 specifically includes: S201, in the network learning layer, a graph embedding algorithm is used to transform the nodes in the heterogeneous information network of the heating system into low-dimensional vector representations, capturing the structural and attribute features of the nodes: extracting historical operating data of each core node, including at least: heat source output, supply and return water temperature, pressure, flow rate, heat user load, sensor monitoring values, economic cost, and carbon emission time series data; associating the heterogeneous information network, including: mapping historical operating data with nodes and edges in the network to form a weighted heterogeneous information network; using a graph embedding algorithm to adapt to the multi-relationship characteristics of the heterogeneous information network, outputting the embedding vectors of each core node; S202, setting perturbations of the heating parameters of the pipe section, generating a perturbation dataset, and re-inputting the perturbation dataset into the heterogeneous information network to obtain new embedding vectors for the pipe section and associated nodes; calculating the vector difference between the original embedding and the perturbation-post-embedded embedding, and combining the heterogeneous information... The network edge weights are used to identify the propagation path with the strongest impact; and the changes in pipe segment parameters are associated with benefit target nodes to quantify the impact of changes in benefit target nodes; S203, random errors are introduced into sensor monitoring data to generate a noisy dataset; the difference between the original embedding vector and the noisy embedding vector is compared to identify the node most sensitive to the error; the propagation path of error from sensor nodes to other nodes is tracked through the topology of the heterogeneous information network; S204, fluctuations in heat source output and heat source equipment renovation, and fluctuations in heat user status parameters are set to generate multi-scenario datasets; the differences in embedding vectors of heat source, pipe segment, and heat user under different fluctuation scenarios are calculated; the chain of influence between heat source, pipe segment, and heat user is constructed by combining the edge weights of the heterogeneous information network; S205, based on the topology of the heterogeneous information network and the embedding learning results, multiple static attribute meta-paths and multiple dynamic attribute meta-paths are extracted.

[0009] Furthermore, the static attribute meta-paths include at least: static meta-paths for heat load demand forecasting, static meta-paths for heat source heat supply forecasting, meta-paths for heat load changes affecting source-network planning, meta-paths related to external transmission pipelines, meta-paths related to heat source decommissioning / new pipelines and pipeline schemes, meta-paths for large temperature difference units / new thermal storage schemes, and meta-paths related to multi-heat source routing change schemes; the dynamic attribute meta-paths include at least: dynamic meta-paths for heat load demand forecasting, dynamic meta-paths for heat source heat supply forecasting, meta-paths for heat load changes affecting source-network optimization, meta-paths related to renewable energy and heat pump operation, meta-paths related to thermal storage and multi-heat source flow, meta-paths related to grid electricity prices and water pump operation, and meta-paths related to user demand and multi-heat source response.

[0010] Further, S3 specifically includes: S301, selecting from multiple types of static and dynamic attribute meta-paths: static meta-path for heat load demand prediction, dynamic meta-path for heat load demand prediction, static meta-path for heat source heat supply prediction, and dynamic meta-path for heat source heat supply prediction; S302, extracting the inherent characteristics of heat users based on the static meta-path for heat load demand prediction to calculate the baseline load value for each heat user, and then capturing the time-series change characteristics of heat load based on the dynamic meta-path for heat load demand prediction to correct the baseline load value; S303, using the inherent characteristics of heat users and the time-series change characteristics of heat load as joint inputs, and substituting them into the first machine... The learning algorithm performs collaborative learning of two types of features and outputs the heat load demand prediction results of the heating system under different cycles; S304, the inherent features of the heat source are extracted based on the static meta-path of heat source heat supply prediction to clarify the upper limit of the heat supply capacity of the heat source equipment and calculate the heat supply benchmark value of the heat source equipment. Then, the heat source heat supply prediction dynamic meta-path captures the heat fluctuation time series change features and dynamically corrects the supply amount; S305, the inherent features of the heat source and the heat fluctuation time series change features are used as joint inputs and substituted into the second machine learning algorithm to perform collaborative learning of the two types of features and output the heat supply prediction results of the heating system under different cycles.

[0011] Furthermore, S4 specifically includes: S401, dividing the heating system transformation and construction planning cycle into multiple cycles based on the technological iteration of the heating system, urban planning development, and heat load growth; S402, setting node weights for each cycle of heating system planning based on the key points of heating system planning in each cycle, and obtaining multi-source data and corresponding heat load demand forecasting results and heat supply forecasting results under multiple static attribute meta-paths corresponding to the heating system in each cycle, as input to the planning model; S403, clarifying the constraint information between nodes based on the static attribute meta-paths, including at least physical constraints, supply and demand balance constraints, transformation and construction timing, and economic constraints; S404, constructing an optimized planning model for the heating system with the goal of minimizing the transformation and construction cost in each cycle, expressed as: ; The cost of heat source modification, construction, and decommissioning in the t-th cycle; Let t represent the cost of pipeline modification and new construction in the t-th cycle. The cost of putting into operation the large temperature difference unit and new thermal storage technology in the t-th cycle; Let n be the system operation and maintenance cost for the t-th cycle; n is the number of cycles; S405, solve the heating system optimization planning model and output the planning scheme for each cycle of transformation and construction.

[0012] Furthermore, the planning schemes for each phase of the transformation and construction include: variables related to the renovation / decommissioning of existing heat sources on the heating network side, variables related to the type and capacity of newly built heat sources, variables related to the commissioning of large temperature difference units and thermal storage; variables related to the expansion / new construction of external transmission pipelines on the pipeline network side, variables related to the renovation / new construction of regional pipelines, variables related to the adjustment of multiple heat source routes; and the implementation time of each planning decision variable.

[0013] Furthermore, S5 specifically includes: S501, acquiring multi-source data and corresponding heat load demand forecasting results and heat supply forecasting results under multiple dynamic attribute element paths corresponding to the heating system in each cycle, and using the planning schemes for each cycle of transformation and construction as input to the planning model; S502, constructing an optimized operation model for the heating system with the goal of minimizing operating costs and maximizing safety in each cycle of the heating system, expressed as: ; S503. Set the constraints on each node and the relationships between nodes during the operation of the heating system, including: supply and demand balance constraints, dynamic output constraints, economic constraints, safety constraints, and planning connection constraints; S504. Solve the optimized operation model of the heating system and output the operation scheduling scheme for each cycle of the transformation and construction.

[0014] Furthermore, the operation and scheduling schemes for each cycle of the transformation and construction include: the output plan of each heat source equipment on the heat source side, the charging and discharging power of the thermal storage system; the operating time and frequency of the water pumps on the pipeline side, the distribution of multiple heat source flows; and the implementation time of each optimization decision.

[0015] Furthermore, in S6, based on the network learning layer, the impact of the planning scheme and operation scheduling scheme of the previous cycle on the optimization of the next cycle is analyzed. Specifically, this includes: after simulating and implementing the initial planning scheme and operation scheduling scheme of the first cycle, collecting actual operation data and simulated operation data for the entire cycle, and using the network learning layer to quantify the impact of the first cycle scheme on the optimization of the second cycle based on the heterogeneous information network, outputting the influence coefficient matrix, and repeating the iteration until the optimization of all cycles is completed.

[0016] The beneficial effects of this invention are: (1) This invention uses a heterogeneous information network as the core carrier to integrate multiple types of nodes such as heat sources, pipe networks, users, and technical equipment, as well as multi-dimensional data such as static attributes and dynamic attributes. By breaking down data fragmentation through nodes and relationship matrices, the model can utilize the entire chain of source, network, and load data, avoiding decision-making bias caused by a single data dimension; (2) This invention uses a heterogeneous information network to quantitatively analyze the chain effect of parameter changes of a single node / pipe segment on the entire system and identify the path of influence propagation, allowing planning and operation schemes to consider the impact of local decisions on the global system, avoiding local optima but system imbalance; (3) The static attribute meta-path of this invention provides the prediction benchmark boundary, and the dynamic attribute meta-path captures fluctuation characteristics. Combined with machine learning algorithms, it significantly reduces the deviation rate of medium- and long-term heat load and heat supply prediction, providing accurate input for planning and operation; (4) The static and dynamic meta-paths of this invention clarify which characteristics affect decision-making. The nodes and relationship topology of the heterogeneous information network make the decision basis traceable, solving the problem that traditional machine learning models only know what but not why, making it easier for technicians and decision-makers to understand and implement; (5) This invention uses a static attribute meta-path Solidify the inherent constraints of the system, combine multi-cycle division and node weight setting, so that the planning scheme of each cycle can not only meet the minimum construction cost of a single cycle, but also adapt to the long-term transformation sequence, avoiding the contradiction of saving money in the short term but being redundant in the long term or being aggressive in the long term but not feasible in the short term. By integrating all elements through heterogeneous information network, the planning scheme can form the optimal combination of cycle connection and element coordination throughout the cycle; (6) This invention captures the real-time fluctuation of the system through dynamic attribute element path, so that the optimized operation model can dynamically adjust the scheduling strategy, while reducing operating costs such as fuel costs and electricity costs, and through overpressure / over-limit early warning, supply and demand balance The balance constraint ensures the reliability of heating supply, achieves a conflict-free balance between the lowest cost and safety compliance, and allows the scheduling scheme to adapt to real-time operating conditions through dynamic metapath and dual-objective model, taking into account both economy and safety; (7) The network learning layer of this invention analyzes the implementation impact of the previous cycle scheme through heterogeneous information network, outputs a quantitative impact coefficient matrix, and reversely corrects the prediction model, planning constraints and operation strategy of the subsequent cycle, so that the scheme gradually converges from the initial feasible to the global optimum of the whole cycle, avoiding repeated pitfalls. Through closed-loop iteration, the scheme of each cycle is optimized based on the previous experience and adapted to the dynamic changes of the system.

[0017] Other features and advantages will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for optimizing the planning and operation of a heating system based on a heterogeneous information network, according to the present invention.

[0021] Figure 2 This is a schematic diagram illustrating the principle of optimized planning and operation of the heating system of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 , Figure 2As shown, this embodiment provides a method for optimizing the planning and operation of a heating system based on a heterogeneous information network. It includes: S1. In the heterogeneous information network construction layer, identifying and selecting multiple types of core nodes in the planning and operation process of the heating system, defining the relevant attributes of the nodes, and analyzing the relationships between nodes to form a relation matrix, thereby constructing a heterogeneous information network for the heating system; S2. In the network learning layer, using the constructed heterogeneous information network of the heating system combined with historical operating data, embedding learning is performed on each core node to analyze the parameter changes of a single typical pipe section, the impact of parameter measurement errors on the optimization result parameters of that pipe section, and the propagation path, as well as the chain reaction analysis of fluctuations in the state parameters of heat sources and heat users. Through topological integration of the heterogeneous information network, multiple static attribute meta-paths and multiple dynamic attribute meta-paths of the heating system are formed; S3. In the load supply and demand layer, based on the multiple static and dynamic attribute meta-paths in the heterogeneous information network of the heating system, meta-paths related to heat load prediction type and meta-paths related to heat supply prediction are selected, and the multi-source data of the heating system's source, network, and load are integrated. According to the above, machine learning algorithms are used to establish a long-term heat load demand prediction model and a heat supply prediction model for the heating system; S4, at the optimization planning layer, the heating system transformation and construction planning cycle is divided into multiple cycles. Based on the multi-class static attribute meta-path and prediction model of the heating system, a set of planning schemes for each transformation and construction cycle is set. With the goal of minimizing the transformation and construction cost of each cycle, combined with the weights and constraints of each node, an optimization planning model for the heating system is constructed, and the planning schemes for each transformation and construction cycle are output; S5, at the optimization operation layer, based on the multi-class dynamic attribute meta-path and prediction model of the heating system, combined with the planning schemes for each transformation and construction cycle, the economic and safety objectives are integrated to construct an optimization operation model for the heating system, and the operation scheduling schemes for each transformation and construction cycle are output; S6, through iterative optimization of the optimization planning layer, load supply and demand layer, and optimization operation layer, and based on the network learning layer, the impact of the planning schemes and operation scheduling schemes of the previous cycle on the optimization of the next cycle is analyzed, and finally the optimal planning schemes and operation scheduling schemes for each transformation and construction cycle are formed.

[0024] In this embodiment, S1 specifically includes: S101, identifying and selecting multiple types of core nodes in the planning and operation process of the heating system, including at least: physical entity nodes, parameter attribute nodes, benefit target nodes, and decision behavior nodes; the physical entity nodes include heat sources, heating networks, heat users, heating network equipment, and sensors; the parameter attribute nodes include heating parameters and economic parameters; the benefit target nodes include economic costs and carbon emissions; the decision behavior nodes include planning decisions and operational decisions; S102, defining the relevant attributes of each node, including at least: the output scale of the heat source, energy type and operating efficiency, the heat load scale of the heat user, the heating period and temperature demand, the diameter and length of the heating network pipes, the adjustment accuracy and operating parameters of the heating network equipment, the measurement accuracy and response time of the sensors, and the geographical coordinates of each physical entity node. The network topology; and the supply and return water temperatures, pressures, flow rates, heat loads, and heat transfer rates in the heating parameters; energy prices in the economic parameters; system transformation construction and operating costs in the economic costs; carbon emissions in the carbon emission quantity and intensity; construction sequence, transformation cycle, and equipment selection constraints in the planning decisions; and equipment scheduling strategies in the operation decisions; S103, analyze the relationships between nodes, including at least: the relationship between heat sources and the heating network, the relationship between the heating network and heat users, the relationship between heating network equipment and parameter attribute nodes, and the relationship between parameter attribute nodes and benefit target nodes, and form a relationship matrix based on the number of nodes, node attributes, and the relationship between edges between nodes, considering the edge representation value and edge direction; S104, construct a heterogeneous information network for the heating system based on multiple types of core nodes, the relevant attributes of each node, and the relationship matrix between nodes. , For each core node set, The set of attributes possessed by each node. It is the set of relationships between nodes in the network.

[0025] It should be noted that the heterogeneous information network of the heating system includes not only physical entities and parameters, but also sensors, benefit objectives, decision-making behaviors, etc. Each node has attributes, and the edges are heterogeneous of multiple types, covering physical constraints (such as the thermal impact of heat source output on supply and return water temperatures), economic mapping (such as the cost relationship between electricity price and water pump operation), topological relationships (such as the flow distribution between pipe segments), and decision-making impacts (such as the temporal relationship between planning schemes and heat source decommissioning). By integrating the full-element relationships of physical entities, parameters, benefits, and decisions, it supports long-term time-series optimization at the planning level and short-term scheduling optimization at the operation level.

[0026] Examples illustrating the relationships between nodes include: heat source - heating network pipe segment: if the heat source outputs heat to the pipe segment, then an edge exists; heating network pipe segment - heat user: if the pipe segment delivers a certain amount of heat to the user, then an edge exists; heating network equipment - parameter attribute node: if the equipment affects the parameter, then an edge exists; parameter attribute node - benefit target node: if the parameter affects the economic cost, then an edge exists.

[0027] In this embodiment, S2 specifically includes: S201, in the network learning layer, a graph embedding algorithm is used to transform the nodes in the heterogeneous information network of the heating system into low-dimensional vector representations, capturing the structural and attribute features of the nodes; extracting historical operating data of each core node, including at least: heat source output, supply and return water temperature, pressure, flow rate, heat user load, sensor monitoring values, economic cost, and carbon emission time series data; associating the heterogeneous information network, including: mapping historical operating data with nodes and edges in the network to form a weighted heterogeneous information network; using a graph embedding algorithm to adapt to the multi-relationship characteristics of the heterogeneous information network, outputting the embedding vectors of each core node; S202, setting perturbations of the heating parameters of the pipe section, generating a perturbation dataset, and re-inputting the perturbation dataset into the heterogeneous information network to obtain new embedding vectors for the pipe section and associated nodes; calculating the vector difference between the original embedding and the perturbation-post-embedded embedding, and combining the heterogeneous... The edge weights of the information network are used to identify the propagation path with the strongest impact; and the changes in pipe segment parameters are associated with benefit target nodes to quantify the impact of changes in benefit target nodes; S203, random errors are introduced into sensor monitoring data to generate a noisy dataset; the difference between the original embedding vector and the noisy embedding vector is compared to identify the node most sensitive to error; the propagation path of error from sensor nodes to other nodes is tracked through the topology of the heterogeneous information network; S204, fluctuations in heat source output and heat source equipment renovation, and fluctuations in heat user status parameters are set to generate multi-scenario datasets; the differences in embedding vectors of heat source, pipe segment, and heat user under different fluctuation scenarios are calculated; the chain of influence between heat source, pipe segment, and heat user is constructed by combining the edge weights of the heterogeneous information network; S205, based on the topology of the heterogeneous information network and the embedding learning results, multiple static attribute meta-paths and multiple dynamic attribute meta-paths are extracted.

[0028] It should be noted that, for example, historical operating data can be mapped to nodes (such as heat source S1, pipe segment T1, heat user D1) and edges (such as the heating relationship between S1 and T1) in the network to form a weighted heterogeneous information network (the edge weights are the statistical characteristics of historical parameters).

[0029] To adapt to the multi-relationship characteristics of heterogeneous information networks, the embedding vectors of each core node are output, including: defining head node embedding, relation embedding, and tail node embedding for each triple; for each relation type (such as heating relation, monitoring relation, and economic impact relation), a projection matrix is ​​introduced to project the head and tail nodes onto the relation-specific space, and a scoring function is used to measure the rationality of the triples, minimizing the error to learn the embedding vectors; finally, the embedding vectors of each core node are output, capturing the structural position, attribute features, and relational dependencies of the nodes.

[0030] The parameter impact analysis of a single pipe segment, taking the water supply temperature change of pipe segment T1 as an example, is explained as follows: Parameter change simulation: A perturbation of ±5℃ in the water supply temperature of pipe segment T1 is set in historical data to generate a simulation dataset; Embedded vector update: The perturbation data is re-inputted into the embedding model to obtain new embedding vectors for pipe segment T1 and related nodes (such as heat source S1 and heat user D1); Quantification of impact propagation path: The difference between the vectors of the original embedding and the perturbation-reset embedding is calculated, and combined with the edge weights of the heterogeneous network, the propagation path with the greatest impact intensity is identified (e.g., the edge weight change of heat transport from T1 to D1 is the largest, indicating that the heat load of heat users is most significantly affected by the pipe segment temperature); Correlation of optimization results: The changes in pipe segment parameters are correlated with the optimization objective to quantify the sensitivity of the change in system cost by 1℃ change in pipe segment temperature.

[0031] The impact analysis of parameter measurement error, taking the measurement error of temperature sensor as an example, is explained as follows: Error simulation: Introduce a random error of ±0.1℃ into the sensor monitoring data to generate a noisy dataset; Embedding robustness analysis: Compare the difference between the original embedding and the noisy embedding to identify the nodes most sensitive to error (such as heat pump nodes that rely on high-precision temperature); Propagation path tracing: Through the topology of heterogeneous networks (such as the edge of sensor → pipe segment → heat source), trace the propagation link of error from sensor to heat source output decision, and quantify the degree of error amplification / attenuation at each step.

[0032] The chain reaction analysis of heat source state parameters is illustrated below, taking the output fluctuation of heat source S1 as an example: Fluctuation simulation: Set the output fluctuation of heat source S1 to ±10% to generate multi-scenario datasets; Embedded vector difference comparison: Calculate the embedding vector differences of heat source S1, pipe segment T1, and user D1 under different fluctuation scenarios; Chain reaction quantification: Combine the edge weights of the heterogeneous network (such as the heating weight of S1→D1 and the heat transfer weight of T1→D1) to construct the chain reaction chain of heat source output fluctuation → pipe segment heat loss → user temperature fluctuation, and output the influence coefficient of each link.

[0033] In this embodiment, the static attribute meta-path includes at least: static meta-path for heat load demand forecasting, static meta-path for heat source heat supply forecasting, meta-path for heat load change affecting source-network planning, meta-path related to external transmission pipeline network, meta-path related to heat source decommissioning / new pipeline and pipeline scheme, meta-path for large temperature difference unit / thermal storage new scheme, and meta-path related to multi-heat source routing change scheme; the dynamic attribute meta-path includes at least: dynamic meta-path for heat load demand forecasting, dynamic meta-path for heat source heat supply forecasting, meta-path for heat load change affecting source-network optimization, meta-path related to renewable energy and heat pump operation, meta-path related to thermal storage and multi-heat source flow direction, meta-path related to grid electricity price and water pump operation, and meta-path related to user demand and multi-heat source response.

[0034] It should be noted that the static metapath of heat load demand forecasting describes the fixed factors that affect heat load demand. These factors determine the baseline level and inherent patterns of heat load, which do not change with intraday / seasonal fluctuations. These factors include: user type → building type → building envelope performance → baseline heat load, user scale → heating area → heat index per unit area → total baseline heat load, geographical region → climate zone → outdoor design temperature → minimum heat load demand.

[0035] Static metapath for heat source heat supply prediction: This describes the inherent and stable supply capacity characteristics of the heat source itself. These characteristics determine the maximum supply ceiling and basic operating parameters of the heat source, and are not affected by short-term operating condition fluctuations. These include: heat source type → rated output → fuel type → maximum supply capacity, number of heat sources → equipment model → efficiency range → basic supply efficiency, geographical location of heat source → pipeline connection method → ​​transmission loss coefficient → effective supply capacity.

[0036] The impact of heat load changes on source-network planning: This describes the long-term, fixed correlation rules between heat load and source-network planning, i.e., the inherent constraint logic of source-network planning, which does not adjust with short-term load changes. This includes: upper limit of heat load growth → heat source expansion threshold → upper limit of pipeline capacity → scale of source-network planning and load distribution density → pipeline topology → heat source layout principles → source-network matching scheme and load type (stable / fluctuating) → heat source type selection → pipeline design standards → source-network planning constraints.

[0037] The related meta-paths of the extended transmission pipeline network focus on the planning constraints and impacts of the long-distance transmission network, characterizing the relationship between long-distance access, network topology, and user coverage. The long-distance heat source-external pipeline segment-regional node-user model depicts the long-term layout logic of the extended heat source-long-distance pipeline segment-regional topology node-user coverage, used to analyze the impact of the long-distance pipeline network construction sequence on regional user heat supply. The external pipeline segment-diameter parameter-transmission efficiency-benefit node model characterizes the relationship between external pipeline segment diameter, transmission efficiency, and system economic cost, supporting the economic optimization of long-distance pipeline network diameter selection.

[0038] The meta-paths related to heat source decommissioning / new construction and pipeline schemes, centered around the planning logic of heat source iteration and pipeline matching, depict the relationship between heat source decommissioning / new construction → pipeline adaptation → system energy efficiency. Existing heat source - decommissioning sequence - pipeline reuse - new heat source: depicts the sequence logic of existing heat source decommissioning time → compatibility of reused pipe sections → new heat source access, used to identify the transition window between heat source decommissioning and new construction. New heat source - pipeline scheme - heat medium parameters - large temperature difference unit: depicts the relationship between new heat source → matching pipeline → heat medium parameters → large temperature difference unit energy efficiency, supporting the collaborative planning of new heat sources and large temperature difference technology.

[0039] The relevant meta-paths for large temperature difference (DTF) generator units / thermal storage solutions focus on the planning value of new technologies such as large temperature difference and thermal storage, depicting the relationship between technology selection, parameter matching, and efficiency improvement. DTF Unit - Heat Medium Parameters - Thermal Storage Unit - Efficiency Node: This section depicts the relationship between DTF unit → heat medium parameters → thermal storage unit adaptation → system carbon emission reduction efficiency, supporting the joint planning of large temperature difference and thermal storage technologies. Thermal Storage Unit - Charge / Discharge Sequence - Heat Source Output - Economic Cost: This section depicts the relationship between the thermal storage unit's charge / discharge sequence → heat source output adjustment → system economic cost, used for long-term economic scheduling planning of thermal storage systems.

[0040] The meta-paths related to multi-heat source / routing change schemes revolve around the planning logic of multi-heat source collaboration and routing optimization, characterizing the relationship between multi-heat source layout, routing selection, and system reliability. Multi-heat source-routing scheme-pipeline redundancy-safety node: characterizes the relationship between multi-heat source layout, routing scheme, pipeline redundancy, and system safety factor, supporting reliability optimization of multi-heat source routing. Heat source routing-geographical node-construction difficulty-investment cost: characterizes the relationship between heat source routing, geographical nodes, construction difficulty, and investment cost, used for economic comparison of multi-heat source routing schemes.

[0041] Dynamic metapath for heat load demand forecasting: Characterizes the short-term changing factors that affect heat load demand. These factors cause the heat load to fluctuate above the static baseline and directly determine the real-time demand value of the load. These factors include: intraday time period → user heating behavior → real-time heat load, short-term weather changes → outdoor temperature / solar radiation → load fluctuation, holidays / production shifts → user activity intensity → load change.

[0042] Dynamic meta-path for heat source heat supply forecasting: This describes the short-term dynamic changes in heat source supply capacity. These changes originate from fluctuations in operating conditions and external conditions, and determine the real-time supply capacity of the heat source, including: fuel supply pressure → heat source combustion efficiency → real-time output, equipment operating time → temporary impact of aging → supply efficiency, renewable energy resource quantity → output fluctuations → supply supplementation.

[0043] The dynamic meta-path of heat load change affecting source network optimization: It describes the real-time correlation rules between dynamic heat load changes and source network optimization, that is, how load fluctuations trigger dynamic adjustments of source network operating parameters to ensure that the optimization scheme adapts to short-term changes, including: load fluctuation amplitude → heat source output adjustment speed → pipeline flow distribution and load change frequency → heat storage charging and releasing rhythm → water pump operating frequency and load area distribution changes → heat source start-stop combination → pipeline pressure regulation.

[0044] The meta-path related to renewable energy and heat pump operation focuses on the dynamic correlation between renewable energy output, heat pump operation, and heating network parameters, supporting the optimization of collaborative heating with new energy sources. Renewable Energy-Heat Pump-Heating Network Segment-Parameters: This describes the dynamic feedback of operating parameters from renewable energy output (e.g., solar, wind power) to heat pump operating status, heating network segment temperature supply, and is used for real-time adjustment of the output distribution between the heat pump and conventional heat sources. Heat Pump-Electricity Market-Operating Cost-Benefit: This describes the correlation between heat pump operation, electricity market price fluctuations, system operating costs, and carbon emission reduction benefits, supporting the economical operation and scheduling of heat pumps during periods of low electricity prices.

[0045] The meta-path related to thermal storage and multi-heat source flow revolves around the dynamic logic of thermal storage strategy → multi-heat source flow → heating network heat storage, optimizing the coordinated operation of multiple heat sources and thermal storage. Thermal storage unit - multiple heat sources - flow - heating network heat storage: This describes the dynamic correlation between the thermal storage unit's heat charging / discharging strategy → multi-heat source output flow → heating network heat storage status → heat storage parameters, used for joint scheduling optimization of multiple heat sources and thermal storage. Multi-heat source flow - heating network heat storage - pump - start / stop: This describes the dynamic feedback of multi-heat source flow adjustment → heating network heat storage changes → pump operating status → pump start / stop plan, supporting intelligent start / stop of pumps according to the heating network heat storage demand.

[0046] This paper combines the economic logic of grid electricity price fluctuations → water pump operation → heating network energy consumption to achieve cost-optimal scheduling of water pump operation. The graph depicts the relationship between grid electricity price fluctuations, water pump operating power, heating network energy consumption, and system operating costs, supporting full-load operation of water pumps during off-peak hours and on-demand start-up and shutdown during peak hours. The graph also depicts the dynamic relationship between water pump operation, heating network heat storage demand, renewable energy output response, and renewable energy output adjustment, used for the coordinated optimization of heating network and renewable energy consumption.

[0047] The meta-path related to user demand and multi-heat source response focuses on the closed-loop logic of dynamic user demand → rapid multi-heat source response → operational strategy, improving the quality of heating services. User-Dynamic Heat Load-Multi-Heat Source-Output: Characterizes the dynamic changes in user heat load → multi-heat source output response → heat source output adjustment → dynamic feedback of operating parameters, used for real-time tracking and scheduling of user load by multiple heat sources. Multi-Heat Source-Output-Heat Pump-Operation: Characterizes the correlation between multi-heat source output allocation → heat pump operating status → heat pump parameters → operational optimization, supporting complementary operation strategies between conventional heat sources and heat pumps.

[0048] In this embodiment, S3 specifically includes: S301, selecting from multiple types of static and dynamic attribute meta-paths: static meta-path for heat load demand prediction, dynamic meta-path for heat load demand prediction, static meta-path for heat source heat supply prediction, and dynamic meta-path for heat source heat supply prediction; S302, extracting inherent characteristics of heat users based on the static meta-path for heat load demand prediction to calculate the baseline load value for each heat user, and then capturing the time-series change characteristics of heat load based on the dynamic meta-path for heat load demand prediction to correct the baseline load value; S303, using the inherent characteristics of heat users and the time-series change characteristics of heat load as joint inputs, and substituting them into the first machine... The machine learning algorithm performs collaborative learning of two types of features and outputs the heat load demand prediction results of the heating system under different cycles; S304, the inherent features of the heat source are extracted based on the static meta-path of heat source heat supply prediction to clarify the upper limit of the heat supply capacity of the heat source equipment and calculate the heat supply benchmark value of the heat source equipment. Then, the heat source heat supply prediction dynamic meta-path captures the heat fluctuation time series change features and dynamically corrects the supply amount; S305, the inherent features of the heat source and the heat fluctuation time series change features are used as joint inputs and substituted into the second machine learning algorithm to perform collaborative learning of the two types of features and output the heat supply prediction results of the heating system under different cycles.

[0049] In this embodiment, S4 specifically includes: S401, dividing the heating system transformation and construction planning cycle into multiple cycles based on the technological iteration of the heating system, urban planning development, and heat load growth; technological iteration includes the gradual application of heat pumps and thermal storage technologies; urban planning development includes the timing of new urban area construction and old urban area renovation; and heat load growth includes the gradual expansion of the number of users and building scale. S402, based on the focus of heating system planning in each cycle, setting node weights for each cycle of heating system planning, and obtaining multi-source data and corresponding heat load demand prediction results and heat supply prediction results under multiple static attribute meta-paths corresponding to the heating system in each cycle, as input to the planning model; the planning focus of each cycle is different, such as the early cycle focusing on heat source transformation and the later cycle focusing on the deployment of new technologies, therefore it is necessary to set weights for the planning nodes of each cycle to reflect the resource allocation of different cycles, such as setting high weights for heat source transformation nodes and prioritizing their allocation. Funding allocation; S403. Based on the static attribute element path, clarify the constraint information of the edges between each node, including at least physical constraints, supply and demand balance constraints, transformation and construction sequence, and economic constraints; physical constraints clarify the physical limits of the system, such as the maximum output of the heat source, the maximum transmission flow of the pipeline network, and the operating temperature range of the equipment; supply and demand balance constraints require that the heat load demand forecast results match the heat supply forecast results, that is, the total heat supply in each cycle should be as equal as possible to the total heat load demand; transformation and construction sequence constraints combine the timeline of urban planning and technology iteration to clarify the implementation order of each renovation / new construction project, such as the heat source must be renovated before the pipeline network can be extended, and new thermal storage technologies must be put into use after the heat source renovation is completed; economic constraints require that the construction cost of each cycle be within the budget and that the time value of money be considered; S404. With the goal of minimizing the transformation and construction cost in each cycle, construct an optimization planning model for the heating system, expressed as: ; The cost of heat source modification, construction, and decommissioning in the t-th cycle; Let t represent the cost of pipeline modification and new construction in the t-th cycle. The cost of putting into operation the large temperature difference unit and new thermal storage technology in the t-th cycle; Let n be the system operation and maintenance cost for the t-th cycle; n is the number of cycles; S405, solve the heating system optimization planning model and output the planning scheme for each cycle of transformation and construction.

[0050] Static attribute meta-paths describe long-term, fixed, and time-invariant attribute association paths between various elements of a heating system (heat sources, pipe networks, users, technical equipment, etc.). Their core value lies in providing systematic feature constraints and data association logic for planning models, transforming previously fragmented multi-source data (such as heat source parameters, user building information, and pipe network design drawings) into a model-understandable network. Meta-paths integrate multi-source data through attribute association logic, making the predicted heat load and heat supply more realistic (e.g., accurately capturing the differences in heat load benchmarks for different building types), avoiding prediction biases caused by data fragmentation, and ultimately ensuring that the planning scheme accurately matches actual needs. The attribute association logic of meta-paths is transparent (e.g., the association between heat source type and rated output is a clear technical rule), therefore, the constraints and optimization models generated based on meta-paths possess clear interpretability. This means that the planning scheme is not only cost-effective but also understandable to technical personnel and decision-makers (e.g., why a certain type of heat source should be prioritized for renovation in this cycle can be explained through the association logic of the meta-path).

[0051] Static attribute metapath is the nerve center of this heating system optimization planning technology. It is responsible for integrating multi-source data and generating accurate inputs, as well as defining system constraints and ensuring technical feasibility. Ultimately, it upgrades the planning scheme from simply cost optimization to a precise, systematic, and interpretable global optimal solution. It is a key technical support for realizing the scientific planning and efficient implementation of heating system transformation and construction.

[0052] In this embodiment, the planning schemes for each cycle of the transformation and construction include: variables for the renovation / decommissioning of existing heat sources on the heating network side, variables for the type and capacity of new heat sources, variables for the commissioning of large temperature difference units and thermal storage; variables for the expansion / new construction of external transmission pipelines on the pipeline network side, variables for the renovation / new construction of regional pipelines, variables for the adjustment of multiple heat source routes; and the implementation time of each planning decision variable.

[0053] In this embodiment, S5 specifically includes: S501, acquiring multi-source data and corresponding heat load demand prediction results and heat supply prediction results under multiple dynamic attribute meta-paths corresponding to the heating system in each cycle, and using the planning schemes for each cycle of transformation and construction as input to the planning model; S502, constructing an optimized operation model for the heating system with the goal of minimizing operating costs and maximizing safety in each cycle of the heating system, expressed as: ; S503. Set the constraints on each node and the relationships between nodes during the operation of the heating system, including: supply and demand balance constraints, dynamic output constraints, economic constraints, safety constraints, and planning connection constraints; S504. Solve the optimized operation model of the heating system and output the operation scheduling scheme for each cycle of the transformation and construction.

[0054] The role of dynamic metapaths is to provide dynamic calculation basis for cost targets and dynamic constraint boundaries for safety targets. Dynamic output constraints mean that the output of equipment such as heat sources and water pumps must be within a dynamically adjustable range; economic constraints mean that operating costs must be within budget and low-cost energy should be used preferentially; safety constraints mean that pipeline pressure and flow must not exceed the design upper limit, and the rate of change of heat source output must not exceed the equipment response limit; planning coordination constraints mean that the operating strategy must match the planning scheme for the current cycle.

[0055] Solving the optimal planning model and optimal operation model of the heating system can employ heuristic algorithms (such as reinforcement learning and genetic algorithms) or mathematical programming solvers. Solving the optimal planning model outputs planning schemes for each cycle of transformation and construction, while solving the optimal operation model outputs operation scheduling schemes for each cycle. The dynamic attribute meta-path is the dynamic engine of operation optimization. It is responsible for capturing real-time fluctuations in the system and defining the dynamic constraints and their association with the objectives. Ultimately, it upgrades the operation scheduling scheme from statically feasible to dynamically optimal, cost-effective, and safe real-time strategies, providing key technical support for achieving on-demand adjustment and efficient operation of the heating system.

[0056] In this embodiment, the operation scheduling scheme for each cycle of the transformation and construction includes: the output plan of each heat source equipment on the heat source side, the charging and discharging heat power of the thermal storage system; the operating time and frequency of the water pumps on the pipeline side, the allocation of multiple heat source flow directions; and the implementation time of each optimization decision.

[0057] In this embodiment, in step S6, based on the network learning layer, the impact of the planning scheme and operation scheduling scheme of the previous cycle on the optimization of the next cycle is analyzed. Specifically, this includes: after simulating and implementing the initial planning scheme and operation scheduling scheme of the first cycle, collecting actual operation data and simulated operation data for the entire cycle, and using the network learning layer to quantify the impact of the first cycle scheme on the optimization of the second cycle based on the heterogeneous information network, outputting the influence coefficient matrix, and repeating the iteration until the optimization of all cycles is completed.

[0058] It should be noted that the core iterative layers are the optimization planning layer (determining hardware), the load supply and demand layer (determining trends), and the optimization operation layer (determining software), combined with the network learning layer (feedback on historical impact), forming a closed loop of scheme implementation → effect evaluation → feedback optimization → re-implementation. The technical process of each stage includes: 1) First cycle: Scheme simulation and implementation Simulation: Based on the initial planning scheme (such as new heat source, pipeline renovation) and operation scheduling scheme (such as heat source output sequence, pipeline flow allocation) of the first cycle, the simulation model is used to simulate the full cycle operation scenario and output simulated operation data (such as simulated cost, simulated safety risk, supply and demand matching degree); Implementation: The scheme verified by simulation is actually implemented, and the actual operation data of the whole cycle (such as actual construction cost, actual heat loss, user complaint rate) is collected.

[0059] 2) Network Learning Layer: Quantifying the Impact of the Scheme. The network learning layer analyzes the correlation between the planning scheme, operation scheme, and actual effect in the first cycle through heterogeneous information networks and deep learning, quantifies its impact on the optimization of the second cycle, and outputs an impact coefficient matrix, including: Economic impact coefficient: for example, if the actual cost of newly built heat pumps in the first cycle exceeds the budget by 10%, the heat pump construction cost coefficient will be increased by 10% in the next cycle; Safety impact coefficient: if the pipeline network experiences peak pressure in the first cycle, the upper limit coefficient of the pipeline network will be reduced by 8% in the next cycle; Supply and demand matching impact coefficient: if the actual heat load is 5% higher than the predicted value, the correction coefficient of the load prediction model will be increased by 5% in the next cycle; System coupling impact coefficient: if the planned thermal storage capacity is insufficient, resulting in limited operation scheduling, the minimum threshold coefficient of thermal storage configuration capacity will be increased by 15% in the next cycle. The impact coefficient matrix and the analysis of historical schemes are fed back to the optimization planning layer, optimization operation layer, and load supply and demand layer as the core basis for adjusting the scheme in the next cycle.

[0060] 3) Iterative Optimization: Based on feedback from the network learning layer, the three core layers work together to optimize the second-cycle scheme, repeating the simulation-implementation-learning-optimization cycle. Load Supply and Demand Layer: Based on the supply and demand matching influence coefficient, the prediction model is corrected (e.g., adjusting the dynamic weights of heat load prediction), outputting more accurate heat load and heat supply predictions; Optimization Planning Layer: Combining economic / safety influence coefficients, the planning model is adjusted (e.g., correcting heat source construction costs and pipeline transmission limits), outputting the second-cycle optimized planning scheme; Optimization Operation Layer: Based on the new planning scheme and prediction results, the operation model is adjusted (e.g., optimizing thermal storage scheduling strategies), combining system coupling influence coefficients, outputting the second-cycle optimized operation scheduling scheme; Cyclic Convergence: The second-cycle scheme is simulated and implemented, repeating the above process until all cycles' schemes meet the convergence conditions.

[0061] Through feedback from the network learning layer, the solution can inherit historical experience and avoid past shortcomings. For example, cost overruns in the first cycle will be automatically corrected in the planning of the next cycle. The three-layer iteration ensures global coordination of hardware planning (optimization planning layer), supply and demand forecasting (load supply and demand layer), and software scheduling (optimization operation layer), avoiding local optima. The node-edge association of the heterogeneous information network makes the source of the influence coefficient traceable (such as the reason for cost overruns can be located to the planning decision of a certain type of equipment), improving the credibility of the solution.

[0062] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0063] Furthermore, the functional modules in the various embodiments of this invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. If the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0064] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for optimization planning and operation of a heating system based on a heterogeneous information network, characterized in that, It includes: S1. In the heterogeneous information network construction layer, identify and select multiple types of core nodes in the planning and operation process of the heating system, define the relevant attributes of the nodes, analyze the relationship between the nodes to form a relationship matrix, and then construct the heterogeneous information network of the heating system. S2. At the network learning layer, the constructed heterogeneous information network of the heating system is combined with historical operation data to perform embedding learning on each core node. The parameter changes of a single typical pipe section, the impact of parameter measurement error on the optimization result parameters of the pipe section and the propagation path are analyzed. The chain effect analysis of the fluctuation of heat source and heat user status parameters is also carried out. Through the topology integration of the heterogeneous information network, multiple static attribute meta-paths and multiple dynamic attribute meta-paths of the heating system are formed. S3. At the load supply and demand layer, based on the multiple static and dynamic attribute meta-paths in the heterogeneous information network of the heating system, select meta-paths related to the heat load prediction type and meta-paths related to the heat supply prediction, integrate the multi-source data of the heating system source network load, and use machine learning algorithms to establish a long-term heat load demand prediction model and a heat supply prediction model in the heating system. S4. In the optimization planning layer, the heating system transformation and construction planning cycle is divided into multiple cycles. Based on the multi-class static attribute meta-path and prediction model of the heating system, a set of planning schemes for each transformation and construction cycle is set. With the goal of minimizing the transformation and construction cost of each cycle, the heating system optimization planning model is constructed by combining the weights and constraints of each node, and the planning schemes for each transformation and construction cycle are output. S5. At the optimized operation layer, based on the multi-type dynamic attribute meta-path and prediction model of the heating system, combined with the planning schemes of each cycle of transformation and construction, the economic and safety objectives are integrated to construct an optimized operation model of the heating system and output the operation scheduling schemes of each cycle of transformation and construction. S6. Through iterative optimization of the planning layer, load supply and demand layer, and operation layer, and based on the network learning layer, the impact of the planning scheme and operation scheduling scheme of the previous cycle on the optimization of the next cycle is analyzed, and finally the optimal planning scheme and operation scheduling scheme for each cycle of transformation and construction are formed.

2. The heat supply system optimization planning and operation optimization method according to claim 1, characterized in that, S1 specifically includes: S101. Identify and select multiple types of core nodes in the planning and operation process of the heating system, including at least: physical entity nodes, parameter attribute nodes, benefit target nodes, and decision behavior nodes; the physical entity nodes include heat sources, heating networks, heat users, heating network equipment, and sensors; the parameter attribute nodes include heating parameters and economic parameters; the benefit target nodes include economic costs and carbon emissions; and the decision behavior nodes include planning decisions and operational decisions. S102. Define the relevant attributes of each node, including at least: the output scale of the heat source, energy type and operating efficiency, the heat load scale of the heat user, the heat usage period and temperature demand, the diameter and length of the heating network pipes, the adjustment accuracy and operating parameters of the heating network equipment, the measurement accuracy and response time of the sensors, the geographical coordinates of each physical entity node and the network topology; and the supply and return water temperature, pressure, flow rate, heat load and heat transfer rate in the heating parameters, the energy price in the economic parameters, the system transformation construction cost and operating cost in the economic costs, the carbon emission amount and carbon emission intensity in the carbon emission, the construction sequence, transformation cycle and equipment selection constraints in the planning decision, and the equipment scheduling strategy in the operation decision. S103. Analyze the relationships between nodes, including at least: the relationship between heat source and heat network, the relationship between heat network and heat user, the relationship between heat network equipment and parameter attribute nodes, and the relationship between parameter attribute nodes and benefit target nodes. Based on the number of nodes, node attributes, and the relationship between edges, consider the edge representation value and edge direction to form a relationship matrix. S104, constructing a heterogeneous information network of the heat supply system according to the multiple types of core nodes, the related attributes of each node, and the inter-node relationship matrix , for each core node set, for each attribute set owned by each node, for each node relationship set in the network.

3. The method for optimizing the planning and operation of a heating system according to claim 1, characterized in that, S2 specifically includes: S201. In the network learning layer, a graph embedding algorithm is used to transform the nodes in the heterogeneous information network of the heating system into low-dimensional vector representations, capturing the structural and attribute features of the nodes: Extract historical operational data for each core node, including at least: heat source output, supply and return water temperature, pressure, flow rate, heat user load, sensor monitoring values, economic costs, and time-series data on carbon emissions; Connecting heterogeneous information networks includes mapping historical operational data to nodes and edges in the network to form a weighted heterogeneous information network. A graph embedding algorithm is used to adapt to the multi-relationship characteristics of heterogeneous information networks, and the embedding vectors of each core node are output. S202. Set the disturbance of the heating parameters of the pipe segment, generate the disturbance dataset, and re-input the disturbance dataset into the heterogeneous information network to obtain the new embedding vectors of the pipe segment and associated nodes; calculate the vector difference between the original embedding and the embedding after disturbance, and combine the edge weights of the heterogeneous information network to identify the propagation path with the greatest influence; and associate the changes in pipe segment parameters with the benefit target nodes to quantify the impact of the changes in the benefit target nodes. S203. Introduce random errors into sensor monitoring data to generate a noisy dataset; compare the differences between the original embedding vector and the noisy embedding vector to identify the nodes most sensitive to errors; trace the propagation path of errors from sensor nodes to other nodes through the topology of the heterogeneous information network. S204. Set fluctuations in heat source output and heat source equipment renovation, and fluctuations in heat user status parameters to generate multi-scenario datasets; calculate the differences in embedding vectors of heat source, pipe section, and heat user under different fluctuation scenarios; combine the edge weights of the heterogeneous information network to construct a chain of influence between heat source, pipe section, and heat user. S205. Based on the topology and embedding learning results of heterogeneous information networks, extract multiple static attribute meta-paths and multiple dynamic attribute meta-paths.

4. The method for optimizing the planning and operation of a heating system according to claim 3, characterized in that, The static attribute meta-path includes at least the following: static meta-path for heat load demand forecasting, static meta-path for heat source heat supply forecasting, meta-path for heat load change affecting source network planning, meta-path related to external transmission pipeline network, meta-path related to heat source decommissioning / new pipeline and pipeline scheme, meta-path for large temperature difference unit / new thermal storage scheme, and meta-path related to multi-heat source routing change scheme. The dynamic attribute metapaths include at least: dynamic metapaths for heat load demand forecasting, dynamic metapaths for heat source heat supply forecasting, metapaths for heat load changes affecting source-grid optimization, metapaths related to renewable energy and heat pump operation, metapaths related to thermal storage and multi-heat source flow, metapaths related to grid electricity price and water pump operation, and metapaths related to user demand and multi-heat source response.

5. The method for optimizing the planning and operation of a heating system according to claim 3, characterized in that, S3 specifically includes: S301. Select the static element path for heat load demand forecasting, the dynamic element path for heat load demand forecasting, the static element path for heat source heat supply forecasting, and the dynamic element path for heat source heat supply forecasting from multiple static and dynamic attribute element paths. S302. Based on the static meta-path of heat load demand forecast, the inherent characteristics of heat users are extracted and used to calculate the baseline load value of each heat user. Then, based on the dynamic meta-path of heat load demand forecast, the time-series change characteristics of heat load are captured and used to correct the baseline load value. S303. Take the inherent characteristics of heat users and the time-series variation characteristics of heat load as joint inputs, and substitute them into the first machine learning algorithm to perform collaborative learning of the two types of features, and output the heat load demand prediction results of the heating system under different periods. S304. Based on the static meta-path of heat source heat supply prediction, the inherent characteristics of the heat source are extracted to clarify the upper limit of the heat supply capacity of the heat source equipment and calculate the heat supply benchmark value of the heat source equipment. Then, based on the dynamic meta-path of heat source heat supply prediction, the time-series change characteristics of heat fluctuation are captured and the supply is dynamically corrected. S305. The inherent characteristics of the heat source and the temporal variation characteristics of heat fluctuations are used as joint inputs and substituted into the second machine learning algorithm to perform collaborative learning of the two types of features, and output the heat supply prediction results of the heating system under different cycles.

6. The method for optimizing the planning and operation of a heating system according to claim 1, characterized in that, S4 specifically includes: S401. Based on the technological iteration of the heating system, urban planning and development, and the growth of heat load, the planning cycle for the transformation and construction of the heating system is divided into multiple cycles. S402. Based on the key points of the heating system planning for each cycle, set the node weights for each cycle of the heating system planning, and obtain multi-source data and corresponding heat load demand prediction results and heat supply prediction results under the multi-type static attribute meta-paths corresponding to the heating system for each cycle, as inputs to the planning model. S403. Based on the static attribute meta-path, clarify the constraint information of the edges between each node, including at least physical constraints, supply and demand balance constraints, transformation and construction sequence and economic constraints; S404. With the goal of minimizing the transformation and construction costs in each cycle, an optimal planning model for the heating system is constructed, expressed as: ； The cost of heat source modification, construction, and decommissioning in the t-th cycle; Let t represent the cost of pipeline modification and new construction in the t-th cycle. The cost of putting into operation the large temperature difference unit and new thermal storage technology in the t-th cycle; Let n be the system maintenance cost for the t-th cycle; n is the number of cycles. S405. Solve the heating system optimization planning model and output the planning schemes for each cycle of transformation and construction.

7. The method for optimizing the planning and operation of a heating system according to claim 6, characterized in that, The planning schemes for each phase of the transformation and construction include: variables for the renovation / decommissioning of existing heat sources on the heating network side, variables for the type and capacity of new heat sources, variables for the commissioning of large temperature difference units and thermal storage units; variables for the expansion / new construction of external transmission pipelines on the pipeline network side, variables for the renovation / new construction of regional pipelines, variables for the adjustment of multiple heat source routes; and the implementation time of each planning decision variable.

8. The method for optimizing the planning and operation of a heating system according to claim 1, characterized in that, S5 specifically includes: S501. Obtain multi-source data and corresponding heat load demand forecast results and heat supply forecast results under the multi-type dynamic attribute meta-paths corresponding to the heating system in each cycle, and combine them with the planning schemes of each cycle of transformation and construction as inputs to the planning model. S502. With the goal of minimizing operating costs and maximizing safety in each cycle of the heating system, an optimized operation model for the heating system is constructed, expressed as: ； ； S503. Set constraints on the relationships between nodes during the operation of the heating system, including: supply and demand balance constraints, dynamic output constraints, economic constraints, safety constraints, and planning coordination constraints. S504 Solve the optimized operation model of the heating system and output the operation and scheduling schemes for each cycle of the transformation and construction.

9. The method for optimizing the planning and operation of a heating system according to claim 8, characterized in that, The operational scheduling schemes for each phase of the transformation and construction include: the output plan of each heat source equipment on the heat source side, the charging and discharging power of the thermal storage system; the operating time and frequency of the water pumps on the pipeline side, the allocation of multiple heat source flows; and the implementation time of each optimization decision.

10. The method for optimizing the planning and operation of a heating system according to claim 1, characterized in that, In step S6, based on the network learning layer, the impact of the planning scheme and operation scheduling scheme of the previous cycle on the optimization of the next cycle is analyzed, specifically including: After simulating and implementing the initial planning and operation scheduling schemes for the first cycle, actual and simulated operation data for the entire cycle are collected. The network learning layer is used to quantify the impact of the first cycle scheme on the optimization of the second cycle based on heterogeneous information networks, and the impact coefficient matrix is ​​output. This process is repeated iteratively until the optimization of all cycles is completed.

Images