Heat supply system analysis method for heat supply network unified energy path and heat energy flow modeling

Abstract

The invention provides a heat supply system analysis method for heat supply network unified energy path and heat energy flow modeling, and the method achieves the scientific selection of a modeling method through the construction of a multi-dimensional evaluation system and an AI decision mechanism: firstly, collecting and preprocessing a heat supply system description file; secondly, seven key elements such as system scale characteristics and dynamic characteristic requirements are extracted; further constructing a hierarchical evaluation system comprising 7 first-level indexes and 15 second-level indexes; and then an AI decision model is created to carry out modeling method recommendation, wherein the modeling method comprises a unified energy path method, a thermal energy flow method or a hybrid modeling strategy. According to the method, the AI technology is introduced into heat supply system modeling method selection for the first time, the blindness problem of traditional experience dependence is solved, the method selection accuracy is improved to 85% or above, the calculation efficiency is improved by 3-4 times while the precision of 95% or above is kept in a typical scene, and a key technical support is provided for intelligent design and low-carbon operation of an urban comprehensive energy system.

Description

Technical Field

[0001] This invention belongs to the field of intelligent modeling and optimization technology for heating systems, specifically relating to a heating system analysis method for unified energy path and heat flow modeling of heating networks. This method, through an AI-driven evaluation system, enables automatic optimization selection of modeling methods for complex heating networks, and is particularly suitable for the design, renovation evaluation, and operation optimization of large-scale district heating systems. Background Technology

[0002] With the rapid development of smart heating and integrated energy system construction, the scale of heating networks is constantly expanding and their structure is becoming increasingly complex, placing higher demands on the selection of modeling methods. Currently, heating system modeling mainly adopts the unified energy path method and the heat energy flow method, each with its own advantages and limitations: the unified energy path method has high accuracy in handling long-distance transmission and dynamic operating conditions, but its computational complexity is high; the heat energy flow method is efficient in handling heat exchanger coupled systems, but it is difficult to capture pipe storage and time delay effects. In engineering practice, the selection of modeling methods often relies on engineers' experience, lacking a scientific and automated decision-making mechanism. This leads to the misuse of the heat energy flow method in primary networks with drastic temperature fluctuations, resulting in temperature prediction errors exceeding 10%; the overuse of the unified energy path method in areas with dense heat exchange stations increases computation time by 3 to 5 times; and the lack of a unified standard for modeling mixed pipe networks, with inconsistent modeling methods in different areas, affects the overall simulation accuracy of the system.

[0003] Existing research largely focuses on optimizing single modeling methods, lacking a systematic framework for method selection. While some commercial software offers multiple modeling options, the selection logic remains based on simple rules, failing to consider the multi-dimensional characteristics of heating systems. With the deepening application of digital twins and AI technologies in energy systems, there is an urgent need for a technical solution that can automatically identify system characteristics and intelligently recommend the optimal modeling method. Summary of the Invention

[0004] The purpose of this invention is to address the automation and intelligentization issues in method selection for heating system modeling. By constructing a multi-dimensional evaluation system and an AI decision-making mechanism, the scientific selection of modeling methods can be achieved. This invention is the first to introduce AI technology into the selection of heating system modeling methods, solving the problem of blind reliance on experience in traditional methods. The accuracy of method selection is improved to over 85%, maintaining an accuracy of over 95% in typical scenarios while improving computational efficiency by 3 to 4 times. This provides key technical support for the intelligent design and low-carbon operation of urban integrated energy systems.

[0005] To achieve the above objectives, the present invention proposes the following technical solution: A heating system analysis method for unified energy path and heat flow modeling of heating networks includes the following steps: S1, collect and preprocess the description files of the heating system; S2, based on the description file preprocessed in step S1, extract the key elements of the heating system; S3. Based on the key elements extracted in step S2, construct an evaluation system for comparing the two modeling methods of unified energy path and thermal energy flow. S4. Based on the evaluation system constructed in step S3, create an AI decision-making model; the AI ​​decision-making model is used to determine which modeling method to use.

[0006] Users can perform modeling based on the above decision results.

[0007] In the above technical solution, further, in step S1: the description file of the heating system includes a pipeline parameter table in Excel format, a pipeline topology diagram in CAD / PNG format, an operation video recording in MP4 format, a design document in PDF format, and historical operation data in CSV format, etc.

[0008] Preprocess the collected files: First, extract data from the description files: For pipeline topology diagrams in CAD / PNG format, extract pipeline topology parameters and equipment distribution parameters using computer vision technology; for operational videos in MP4 format, extract equipment operating status parameters and dynamic operating condition change parameters using deep learning algorithms; for design documents in PDF format, extract basic pipeline design parameters and accuracy and efficiency design requirements parameters using natural language processing; for historical operational data in CSV format, extract dynamic temperature parameters and dynamic flow parameters using signal processing technology. Then, the extracted data and the data in the pipeline parameter table in Excel format are uniformly converted into a structured data format to form a digital profile of the heating system, providing a data foundation for subsequent analysis.

[0009] Furthermore, in step S2: based on the description file preprocessed in step S1, the key elements of the heating system are determined by performing multi-dimensional feature extraction and comparative analysis on the preprocessed structured data. The key elements include seven dimensions: system scale characteristics, dynamic characteristics requirements, accuracy requirements, computational efficiency requirements, primary and secondary network coupling degree, thermal inertia characteristics, and economic constraints. System scale characteristics include total pipeline length, number of nodes, number of heat exchange stations, coverage area, number of network loops, and branch complexity parameters. Dynamic characteristics requirements include temperature change frequency, maximum temperature change rate, and load fluctuation amplitude parameters. Accuracy requirements include allowable error range for temperature prediction, accuracy requirements for heat load calculation, and tolerance for dynamic response errors. Computational efficiency requirements include allowable time for a single simulation and real-time requirement levels. Primary and secondary network coupling degree includes heat exchange station density and primary and secondary network connection complexity parameters. Thermal inertia characteristics include the significance of pipe storage effect and heat wave propagation delay characteristics parameters. Economic constraints include modeling cost budget and software license limitations parameters. Each key element is quantified; for example, in the thermal inertia characteristics dimension, the total heat capacity of the working fluid within the pipeline is considered. Maximum heat transfer per unit time ratio The significance of the storage effect is quantified by the heat wave propagation delay parameter. The time delay characteristics of heat wave propagation were quantified. The quantitative extraction of these key elements provides a scientific basis for the subsequent construction of an evaluation system.

[0010] Further, in step S3: based on the seven key elements extracted in step S2, a hierarchical evaluation system is constructed, comprising seven primary indicators and 15 secondary indicators. The primary indicators include accuracy fit, efficiency fit, structural fit, dynamic fit, coupling fit, economic fit, and comprehensive robustness. The initial weight of accuracy fit is 25%, comprising two secondary indicators: temperature field accuracy fit and heat load accuracy fit, calculated based on parameters corresponding to accuracy requirements. The initial weight of efficiency fit is 20%, comprising two secondary indicators: single-time calculation time and iterative optimization efficiency, calculated based on parameters corresponding to computational efficiency requirements. The initial weight of structural fit is 15%, comprising topology complexity fit and equipment... The system has two secondary indicators for type adaptation, calculated based on parameters corresponding to system scale characteristics. Dynamic adaptability has an initial weight of 15%, including two secondary indicators: temperature fluctuation adaptability and load change adaptability, calculated based on parameters corresponding to dynamic characteristic requirements. Coupling adaptability has an initial weight of 10%, including two secondary indicators: primary and secondary network coupling adaptability and multi-energy coupling adaptability, calculated based on parameters corresponding to the degree of primary and secondary network coupling. Economic adaptability has an initial weight of 8%, including two secondary indicators: software cost and labor cost, calculated based on parameters corresponding to economic constraints. Comprehensive robustness has an initial weight of 7%, including two secondary indicators: extreme operating condition performance and parameter sensitivity, calculated based on parameters corresponding to thermal inertia characteristics.

[0011] Furthermore, the evaluation system adopts a 0-10 point scale, calculating the combined score of the unified energy path and thermal energy flow modeling methods through weighted summation. To improve the objectivity of the evaluation, an entropy weight method is introduced to dynamically adjust the index weights, and this is further refined using an expert knowledge base, ensuring that the evaluation results are consistent with objective data and incorporate domain experience. The evaluation system is validated and optimized using a historical case database to ensure its applicability to different types of heating systems.

[0012] Further, in step S4: based on the evaluation system constructed in step S3, an AI decision-making model is developed to realize intelligent recommendation of modeling methods. This model consists of three functional modules: a feature encoder, an evaluation calculation engine, and a decision optimizer, which work in tandem and execute according to the logical order of "feature extraction - suitability scoring - decision optimization". The feature encoder is used to map the seven key element data quantized in step S2 to the [0, 1] interval using the Min-Max normalization method to eliminate dimensional differences. The normalized data is input into a deep neural network containing three fully connected hidden layers and one output layer. The hidden layers of the deep neural network all use the ReLU activation function, and the output layer uses the Sigmoid activation function. The network is trained using the Adam optimizer and the mean squared error (MSE) loss function to capture the nonlinear mapping relationship between the key elements and the suitability of the two modeling methods. When the network training reaches a validation set loss below 0.001, it outputs a 16-dimensional comprehensive feature vector (i.e., a high-dimensional feature vector) that condenses the core information of the seven key elements and the coupling relationships between elements, providing high-dimensional feature support for subsequent evaluation calculations. The evaluation calculation engine first calculates the objective weights of the primary indicators based on the information entropy of the 16-dimensional comprehensive feature vector. Combined with expert subjective weights, a weighted average method is used to fuse the primary indicators to obtain their final dynamic weights, while simultaneously optimizing the weight proportions of the secondary indicators. Then, a mapping relationship is established between the 16-dimensional feature vector (high-dimensional feature vector) and 15 secondary indicators. A triangular membership function is used to transform the feature information into three categories of fuzzy semantics: "high," "medium," and "low." After fuzzy inference synthesis, the centroid method is used to defuzzify the information and obtain quantitative scores for the secondary indicators ranging from 0 to 10. Subsequently, according to the weight proportions of the secondary indicators under the primary indicators, a weighted summation formula is used to calculate the scores of the unified energy path method and the thermal energy flow method on the seven primary indicators. Finally, the scores of the seven primary indicators for both modeling methods are weighted and summed to obtain the comprehensive suitability score of the unified energy path method. Overall compatibility score with thermal energy flow method Simultaneously, it outputs the final dynamic weights of seven primary indicators and the quantitative scores of each primary indicator. The decision optimizer represents the state space with a 16-dimensional comprehensive feature vector and the action space with "selecting the unified energy path method," "selecting the thermal energy flow method," and "selecting hybrid modeling." It constructs a reward function and then trains the decision policy network using the Deep Deterministic Policy Gradient (DDPG) algorithm. Through interaction between the agent and the simulation environment, it learns the tendency of the modeling method after balancing accuracy and efficiency. Finally, it calculates the difference in the comprehensive scores of the two modeling methods. The recommendation confidence C is calculated by combining the Q-value of the reinforcement learning action value function (normalized to [0, 1]): when When ≥3, C=min (0.95+0.05×Q,1); when 1≤ When <3, C = 0.7 + 0.3 × Q; when When the score is less than 1, C = 0.5 + 0.5 × Q. Finally, combining the comprehensive score, confidence level, and engineering constraints (specifically including the recommended modeling method's accuracy must be higher than the minimum engineering accuracy requirement and the recommended modeling method's efficiency must be higher than the minimum efficiency requirement), the optimal modeling method is determined, and three types of core decision information are output: optimal modeling method recommendation (unified energy path method, thermal energy flow method, hybrid modeling strategy, and manual intervention prompts), recommendation confidence level (value range [0, 1]), and ranking of key influencing factors (ranked according to the absolute value of the contribution of the primary indicator to the decision). The optimal modeling method recommendation step is as follows: when the comprehensive suitability score of the unified energy path method is less than 1, the optimal modeling method is determined. Greater than the overall adaptability score of the thermal energy flow method Furthermore, when the confidence level is greater than or equal to 80%, the unified energy path method is recommended for modeling the entire system; when the unified energy path method has a comprehensive adaptability score... Less than or equal to the overall adaptability score of the thermal energy flow method When the confidence level is greater than or equal to 80%, the thermal energy flow method is used to model the entire system; when the confidence level is between 60% and 80%, a hybrid modeling strategy is initiated; when the confidence level is less than 60%, a manual intervention prompt is issued to the user, who can then choose the optimal modeling scheme. The absolute value of the contribution of the primary indicator to the decision is specifically obtained by multiplying the weight of the primary indicator by its score. Among them, the unified energy path method has a comprehensive adaptability score. Overall compatibility score with thermal energy flow method The calculation formula is: j=1, 2; The reward function is as follows: R = 0.6 × P - 0.3 × T + 0.1 × E; In the formula, The score represents the overall score of method j, where j=1 indicates the unified energy path method and j=2 indicates the unified energy path method. The final weight of the i-th indicator is... (j) represents the score of method j in the i-th primary indicator, R represents the reward, P represents the modeling accuracy achievement rate, T represents the computation time overrun coefficient, and E represents the economic score.

[0013] Furthermore, in the unified energy path modeling method, pipes and heat exchangers in the heating network are analogized to components such as resistors, capacitors, inductors, and voltage sources. There is an energy conservation equation in the pipes. By defining heat flow and abstracting the thermal path components, the resistor, inductor, and capacitor equations can be obtained, and the distributed parameter thermal path in the time domain can be derived. Since the heat loss and time delay phenomena that occur during the heat flow migration in the pipes cannot be ignored in the unified energy path analysis, in order to consider time-varying problems, the above-mentioned distributed parameter thermal path in the time domain needs to be transformed into a distributed parameter thermal path in the frequency domain through Fourier transform, and then integrated into a lumped parameter thermal path, so as to obtain the general thermal branch characteristic equation in the frequency domain. Then, by analogy with Kirchhoff's current law for constraint, the heat flow constraint and thermal potential constraint in the heating network are obtained, and the thermal network equation in the heating network can be obtained. Thus, the modeling model of the unified energy path method is obtained.

[0014] In the formula: c is the specific heat. Let A be the density, A be the pipe cross-sectional area, T be the excess temperature characterized by the difference between the working fluid temperature and the ambient temperature, and G be the mass flow rate. The heat dissipation coefficient of the pipe. , These are the temperatures at the beginning and end of the lumped parameter thermal circuit, respectively. , These represent the heat flows at the beginning and end of the lumped parameter thermal path, respectively. Let be the heat transfer factor of the heat flow from the beginning to the end of the pipe, denoted as . , , These are the temperatures at the beginning and end of each pipeline. and For the heat flow at the beginning and end of each pipe, Provide a temperature difference for the heating medium in the pipeline heat exchanger. These are the node-outflow and node-inflow branch correlation matrices, respectively. These are the weighted node-outflow and node-inflow branch correlation matrices, respectively.

[0015] In the thermal energy flow modeling method, taking a co-current heat exchanger as an example, the heat transfer formula is first obtained by using the logarithmic mean temperature difference as the characteristic temperature. The temperature difference and heat capacity of both the cold and hot flows are equal to the heat transfer amount Q, thus yielding the energy conservation equation. Substituting this equation into the heat transfer formula, the relationship between the cold and hot flows is obtained. A new equation is then derived through mathematical transformation. Combining this with the energy conservation equation, three identities are obtained. Finally, combining the transformed equation with the three identities, the loss thermal resistance equation is derived, where the loss thermal resistance... Characteristic temperature The thermal resistance is obtained by dividing the heat flux Q by the thermal resistance. Furthermore, it is determined based on the temperature difference between the cold and hot fluids. The control equations, from the entrance Points are collected at the exit. The integral can be obtained, and combined with the energy conservation equation, we can obtain the characteristic temperature. The standard thermal resistance R, defined by heat flux Q, can be mathematically transformed into the sum of three parts: thermal resistance due to heat loss, thermal resistance due to heat loss, and thermal resistance due to heat loss. and Therefore, the modeling model for the thermal energy flow method can be obtained.

[0016] ; ; ; ; ; ; ; ; ; In the formula, Q represents the amount of heat exchanged by the heat exchanger, K represents the heat transfer coefficient of the heat exchanger, and A represents the heat transfer area of ​​the heat exchanger. and These are the inlet and outlet temperatures of the heat exchanger, respectively. and These are the cold flow inlet and outlet temperatures of the heat exchanger, respectively. and These are the mass flow rates of cold and hot fluids, respectively. and The heat capacities of cold and hot flows are respectively. To account for thermal resistance losses, L is the heat transfer length of the heat exchanger. , These are the temperatures of the cold and hot fluids, respectively, during heat exchange.

[0017] In the hybrid modeling approach, the system automatically divides functional areas and creates functional characteristic indices. ,Will >0.45 is classified as a "long-distance primary network region" and modeled using the unified energy path method; The region is divided into "primary and secondary network coupling segment" at 0.45, and modeled using the thermal energy flow method. Functional characteristic index The calculation formula is: In the formula: i is the number of nodes, Let be the cumulative pipe length from node i to the heat source. This represents the distance from the furthest node in the entire network to the heat source. Let be the maximum rate of temperature change at node i. This represents the maximum rate of temperature change in the entire network. The number of heat exchangers within 100m around node i.

[0018] The beneficial effects of this invention are: This invention provides a heating system analysis method for unified energy path and heat flow modeling of heating networks. It is the first to introduce AI technology into the selection of heating system modeling methods, solving the problem of blind reliance on traditional experience and improving the accuracy of method selection to over 85%. This invention constructs a scientific evaluation system comprising 7 primary indicators and 15 secondary indicators, comprehensively covering the physical characteristics and engineering requirements of the heating system, providing a quantitative basis for modeling method selection. This invention breaks through the limitations of single methods, achieving regional adaptive modeling. In typical scenarios, it maintains over 95% accuracy while improving computational efficiency by 3 to 4 times. This invention supports input from multiple formats such as Excel, CAD, and video, significantly lowering the barrier to entry and enabling non-professionals to obtain professional-grade modeling solutions. In practical engineering applications, this invention significantly improves modeling efficiency and accuracy, providing key technical support for the intelligent design and low-carbon operation of urban integrated energy systems. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] Figure 1 These are the main steps of the method of the present invention.

[0021] Figure 2 This is a schematic diagram of the evaluation system constructed in this invention. Detailed Implementation

[0022] like Figure 1 As shown, the present invention provides a heating system analysis method for unified energy path and heat flow modeling of a heating network, comprising the following steps: Step S1: Collect and preprocess the description files of the heating system. These files include Excel-formatted pipe network parameter tables, CAD / PNG-formatted pipe network topology diagrams, MP4-formatted operational video recordings, PDF-formatted design documents, and CSV-formatted historical operational data. Preprocessing of the collected description files involves: First, extracting data from the description files. Image files (CAD / PNG-formatted pipe network topology diagrams) are processed using computer vision technology to identify the pipe network topology and equipment distribution; text files (PDF-formatted design documents) are processed using natural language processing to extract key parameters; time-series data files (CSV-formatted historical operational data) are processed using signal processing technology to extract dynamic characteristic parameters; and video files (MP4-formatted operational videos) are processed using deep learning algorithms to identify key operational states. Then, the extracted data and the data from the Excel-formatted pipe network parameter tables are uniformly converted into a structured data format to form a digital profile of the heating system, providing a data foundation for subsequent analysis.

[0023] Step S2: Based on the description file preprocessed in Step S1, extract the key elements of the heating system; determine the key elements of the heating system by performing multi-dimensional feature extraction and comparative analysis on the preprocessed structured data. These key elements include seven dimensions: system scale characteristics, dynamic characteristics requirements, accuracy requirements, computational efficiency requirements, primary and secondary network coupling degree, thermal inertia characteristics, and economic constraints. The system scale characteristics include total pipeline length, number of nodes, number of heat exchange stations, coverage area, number of network loops, and branch complexity parameters. The dynamic characteristics requirements include temperature change frequency, maximum temperature change rate, and load fluctuation amplitude parameters. The accuracy requirements include the allowable error range for temperature prediction, the accuracy requirements for heat load calculation, and the tolerance for dynamic response errors. The computational efficiency requirements include the allowable time for a single simulation and the level of real-time requirements. The primary and secondary network coupling degree includes the heat exchange station density and the complexity of primary and secondary network connections. The thermal inertia characteristics include the significance of pipe storage effects and the characteristics of heat wave propagation delay. The economic constraints include modeling cost budget and software licensing restrictions. Each key element is quantified; for example, in the thermal inertia characteristics dimension, the total heat capacity of the working fluid within the pipeline is considered. Maximum heat transfer per unit time ratio The significance of the storage effect is quantified by the heat wave propagation delay parameter. The time delay characteristics of heat wave propagation were quantified. The quantitative extraction of these key elements provides a scientific basis for the subsequent construction of an evaluation system.

[0024] Step S3: Based on the seven key elements extracted in Step S2, construct a hierarchical evaluation system comprising 7 primary indicators and 15 secondary indicators. The primary indicators include accuracy fit, efficiency fit, structural fit, dynamic fit, coupling fit, economic fit, and comprehensive robustness. Accuracy fit has an initial weight of 25%, comprising two secondary indicators: temperature field accuracy fit and heat load accuracy fit, calculated based on parameters corresponding to accuracy requirements. Efficiency fit has an initial weight of 20%, comprising two secondary indicators: single-time calculation time and iterative optimization efficiency, calculated based on parameters corresponding to computational efficiency requirements. Structural fit has an initial weight of 15%, comprising topology complexity fit and equipment type fit. Two secondary indicators are calculated based on parameters corresponding to the system's scale characteristics: Dynamic Adaptability (initial weight 15%), encompassing temperature fluctuation adaptation and load change adaptation, calculated based on parameters corresponding to dynamic characteristic requirements; Coupling Adaptability (initial weight 10%), encompassing primary and secondary network coupling adaptation and multi-energy coupling adaptation, calculated based on parameters corresponding to the degree of primary and secondary network coupling; Economic Adaptability (initial weight 8%), encompassing software cost and labor cost, calculated based on parameters corresponding to economic constraints; and Comprehensive Robustness (initial weight 7%), encompassing extreme operating condition performance and parameter sensitivity, calculated based on parameters corresponding to thermal inertia characteristics. The evaluation system uses a 0-10 point scale, calculating the comprehensive score from two modeling methods through weighted summation. To improve the objectivity of the evaluation, an entropy weight method is introduced to dynamically adjust the indicator weights, and adjustments are made in conjunction with an expert knowledge base to ensure that the evaluation results are consistent with objective data and incorporate domain experience. The evaluation system is validated and optimized through a historical case database to ensure its applicability to different types of heating systems.

[0025] Step S4: Based on the evaluation system constructed in Step S3, develop an AI decision-making model to achieve intelligent recommendation of modeling methods. This model consists of three main functional modules: a feature encoder, an evaluation calculation engine, and a decision optimizer, executed sequentially in the logical order of "feature extraction - suitability scoring - decision optimization." The feature encoder maps the seven key element data quantized in Step S2 to the [0, 1] interval using the Min-Max normalization method to eliminate dimensional differences. The normalized data is input into a deep neural network containing three fully connected hidden layers and one output layer. The hidden layers of the deep neural network all use the ReLU activation function, and the output layer uses the Sigmoid activation function. The network is trained using the Adam optimizer and the mean squared error (MSE) loss function to capture the nonlinear mapping relationship between the key elements and the suitability of the two modeling methods. When the network training reaches a validation set loss below 0.001, it outputs a 16-dimensional comprehensive feature vector (i.e., a high-dimensional feature vector) that condenses the core information of the seven key elements and the coupling relationships between elements, providing high-dimensional feature support for subsequent evaluation calculations. The evaluation calculation engine first calculates the objective weights of the primary indicators based on the information entropy of the 16-dimensional comprehensive feature vector. Combined with expert subjective weights, a weighted average method is used to fuse the primary indicators to obtain their final dynamic weights, while simultaneously optimizing the weight proportions of the secondary indicators. Then, a mapping relationship is established between the 16-dimensional feature vector (high-dimensional feature vector) and 15 secondary indicators. A triangular membership function is used to transform the feature information into three categories of fuzzy semantics: "high," "medium," and "low." After fuzzy inference synthesis, the centroid method is used to defuzzify the information and obtain quantitative scores for the secondary indicators ranging from 0 to 10. Subsequently, according to the weight proportions of the secondary indicators under the primary indicators, a weighted summation formula is used to calculate the scores of the unified energy path method and the thermal energy flow method on the seven primary indicators. Finally, the scores of the seven primary indicators for both modeling methods are weighted and summed to obtain the comprehensive suitability score of the unified energy path method. Overall compatibility score with thermal energy flow method Simultaneously, it outputs the final dynamic weights of seven primary indicators and the quantitative scores of each primary indicator. The decision optimizer represents the state space with a 16-dimensional comprehensive feature vector and the action space with "selecting the unified energy path method," "selecting the thermal energy flow method," and "selecting hybrid modeling." It constructs a reward function and then trains the decision policy network using the Deep Deterministic Policy Gradient (DDPG) algorithm. Through interaction between the agent and the simulation environment, it learns the tendency of the modeling method after balancing accuracy and efficiency. Finally, it calculates the difference in the comprehensive scores of the two modeling methods. The recommendation confidence C is calculated by combining the Q-value of the reinforcement learning action value function (normalized to [0, 1]): when When ≥3, C=min (0.95+0.05×Q,1); when 1≤ When <3, C = 0.7 + 0.3 × Q; when When the score is less than 1, C = 0.5 + 0.5 × Q. Finally, combining the comprehensive score, confidence level, and engineering constraints (specifically including the recommended modeling method's accuracy must be higher than the minimum engineering accuracy requirement and the recommended modeling method's efficiency must be higher than the minimum efficiency requirement), the optimal modeling method is determined, and three types of core decision information are output: optimal modeling method recommendation (unified energy path method, thermal energy flow method, hybrid modeling strategy, and manual intervention prompts), recommendation confidence level (value range [0, 1]), and ranking of key influencing factors (ranked according to the absolute value of the contribution of the primary indicator to the decision). The optimal modeling method recommendation step is as follows: when the comprehensive suitability score of the unified energy path method is less than 1, the optimal modeling method is determined. Greater than the overall adaptability score of the thermal energy flow method Furthermore, when the confidence level is greater than or equal to 80%, the unified energy path method is recommended for modeling the entire system; when the unified energy path method has a comprehensive adaptability score... Less than or equal to the overall adaptability score of the thermal energy flow method When the confidence level is greater than or equal to 80%, the thermal energy flow method is used to model the entire system; when the confidence level is between 60% and 80%, a hybrid modeling strategy is initiated; when the confidence level is less than 60%, a manual intervention prompt is issued to the user, who can then choose the optimal modeling scheme. The absolute value of the contribution of the primary indicator to the decision is specifically obtained by multiplying the weight of the primary indicator by its score. Among them, the unified energy path method has a comprehensive adaptability score. Overall compatibility score with thermal energy flow method The calculation formula is: j=1, 2; The reward function is as follows: R = 0.6 × P - 0.3 × T + 0.1 × E; In the formula, The score represents the overall score of method j, where j=1 indicates the unified energy path method and j=2 indicates the unified energy path method. The final weight of the i-th indicator is... (j) represents the score of method j in the i-th primary indicator, R represents the reward, P represents the modeling accuracy achievement rate, T represents the computation time overrun coefficient, and E represents the economic score.

[0026] Step S4: Based on the AI ​​decision model output from Step S4, automatically select and execute the optimal modeling method. When the unified energy path method is recommended and the confidence level is greater than or equal to 80%, the unified energy path method is used to model the entire system; when the thermal energy flow method is recommended and the confidence level is greater than or equal to 80%, the thermal energy flow method is used to model the entire system; when the confidence level is between 60% and 80%, a hybrid modeling strategy is initiated; when the confidence level is less than 60%, a manual intervention prompt is issued to the user, and a multi-solution comparison interface is provided.

[0027] In the unified energy path modeling method, pipes and heat exchangers in the heating network are analogized to components such as resistors, capacitors, inductors, and voltage sources. There is an energy conservation equation in the pipes. By defining heat flow and abstracting the thermal path components, the resistor, inductor, and capacitor equations can be obtained, and the distributed parameter thermal path in the time domain can be derived. Since the heat loss and time delay phenomena that occur during the heat flow migration in the pipes cannot be ignored in the unified energy path analysis, in order to consider time-varying problems, the above-mentioned distributed parameter thermal path in the time domain needs to be transformed into a distributed parameter thermal path in the frequency domain through Fourier transform, and then integrated into a lumped parameter thermal path, so as to obtain the general thermal branch characteristic equation in the frequency domain. Then, by analogy with Kirchhoff's current law for constraint, the heat flow constraint and thermal potential constraint in the heating network are obtained, and the thermal network equation in the heating network can be obtained. Thus, the modeling model of the unified energy path method is obtained.

[0028] In the formula: c is the specific heat. Let A be the density, A be the pipe cross-sectional area, T be the excess temperature characterized by the difference between the working fluid temperature and the ambient temperature, and G be the mass flow rate. The heat dissipation coefficient of the pipe. , These are the temperatures at the beginning and end of the lumped parameter thermal circuit, respectively. , These represent the heat flows at the beginning and end of the lumped parameter thermal path, respectively. Let be the heat transfer factor of the heat flow from the beginning to the end of the pipe, denoted as . , , These are the temperatures at the beginning and end of each pipeline. and For the heat flow at the beginning and end of each pipe, Provide a temperature difference for the heating medium in the pipeline heat exchanger. These are the node-outflow and node-inflow branch correlation matrices, respectively. These are the weighted node-outflow and node-inflow branch correlation matrices, respectively.

[0029] In the thermal energy flow modeling method, taking a co-current heat exchanger as an example, the heat transfer formula is first obtained by using the logarithmic mean temperature difference as the characteristic temperature. The temperature difference and heat capacity of both the cold and hot flows are equal to the heat transfer amount Q, thus yielding the energy conservation equation. Substituting this equation into the heat transfer formula, the relationship between the cold and hot flows is obtained. A new equation is then derived through mathematical transformation. Combining this with the energy conservation equation, three identities are obtained. Finally, combining the transformed equation with the three identities, the loss thermal resistance equation is derived, where the loss thermal resistance... Characteristic temperature The thermal resistance is obtained by dividing the heat flux Q by the thermal resistance. Furthermore, it is determined based on the temperature difference between the cold and hot fluids. The control equations, from the entrance Points are collected at the exit. The integral can be obtained, and combined with the energy conservation equation, we can obtain the characteristic temperature. The standard thermal resistance R, defined by heat flux Q, can be mathematically transformed into the sum of three parts: thermal resistance due to heat loss, thermal resistance due to heat loss, and thermal resistance due to heat loss. and Therefore, the modeling model for the thermal energy flow method can be obtained.

[0030] ; ; ; ; ; ; ; ; ; In the formula, Q represents the amount of heat exchanged by the heat exchanger, K represents the heat transfer coefficient of the heat exchanger, and A represents the heat transfer area of ​​the heat exchanger. and These are the inlet and outlet temperatures of the heat exchanger, respectively. and These are the cold flow inlet and outlet temperatures of the heat exchanger, respectively. and These are the mass flow rates of cold and hot fluids, respectively. and The heat capacities of cold and hot flows are respectively. To account for thermal resistance losses, L is the heat transfer length of the heat exchanger. , These are the temperatures of the cold and hot fluids, respectively, during heat exchange.

[0031] In the hybrid modeling approach, the system automatically divides functional areas and creates functional characteristic indices. ,Will >0.45 is classified as a "long-distance primary network region" and modeled using the unified energy path method; The region is divided into "primary and secondary network coupling segment" at 0.45, and modeled using the thermal energy flow method. Functional characteristic index The calculation formula is: In the formula: i is the number of nodes, Let be the cumulative pipe length from node i to the heat source. This represents the distance from the furthest node in the entire network to the heat source. Let be the maximum rate of temperature change at node i. This represents the maximum rate of temperature change in the entire network. The number of heat exchangers within 100m around node i.

[0032] This invention is the first to introduce AI technology into the selection of modeling methods for heating systems, solving the problem of blind reliance on experience in traditional methods. The accuracy of method selection is improved to over 85%, and the computational efficiency is increased by 3 to 4 times while maintaining over 95% accuracy in typical scenarios. This provides key technical support for the intelligent design and low-carbon operation of urban integrated energy systems.

Claims

1. A heating system analysis method for unified energy path and heat flow modeling of a heating network, comprising the following steps: S1. Collect and preprocess the description files of the heating system. The description files of the heating system include a pipeline parameter table in Excel format, a pipeline topology diagram in CAD / PNG format, an operation video in MP4 format, a design document in PDF format, and historical operation data in CSV format. S2, based on the description file preprocessed in step S1, extract the key elements of the heating system; S3. Based on the key elements extracted in step S2, construct an evaluation system for comparing the two modeling methods of unified energy path and thermal energy flow. S4. Based on the evaluation system constructed in step S3, an AI decision-making model is created. This AI decision-making model determines which modeling strategy to use. It scores the unified energy path modeling method and the thermal energy flow modeling method and provides the confidence level of each modeling method. Based on the scores and confidence levels of the modeling methods, a recommended modeling method is provided.

2. The heating system analysis method for unified energy path and heat flow modeling of a heating network according to claim 1, characterized in that, In step S1: the description file of the heating system is preprocessed, specifically by: First, extract the data from the description file; For pipeline topology diagrams in CAD / PNG format, computer vision technology is used to extract pipeline topology parameters and equipment distribution parameters; for operational videos in MP4 format, deep learning algorithms are used to extract equipment operating status parameters and dynamic operating condition change parameters; for design documents in PDF format, natural language processing is used to extract basic pipeline design parameters and accuracy and efficiency design requirements parameters; for historical operational data in CSV format, signal processing technology is used to extract dynamic temperature parameters and dynamic flow parameters. Then, the extracted data and the data in the pipeline parameter table in Excel format are uniformly converted into structured data format.

3. The heating system analysis method for unified energy path and heat flow modeling of a heating network according to claim 1, characterized in that, In step S2: the key elements include seven dimensions: system scale characteristics, dynamic characteristics requirements, accuracy requirements, computational efficiency requirements, primary and secondary network coupling degree, thermal inertia characteristics, and economic constraints. The system scale characteristics include total pipeline length, number of nodes, number of heat exchange stations, coverage area, number of network loops, and branch complexity parameters. The dynamic characteristics requirements include temperature change frequency, maximum temperature change rate, and load fluctuation amplitude parameters. The accuracy requirements include the allowable error range for temperature prediction, the accuracy requirements for heat load calculation, and the tolerance for dynamic response errors parameters. The computational efficiency requirements include the allowable time for a single simulation and the level of real-time requirements parameters. The primary and secondary network coupling degree includes the heat exchange station density and the complexity of primary and secondary network connections parameters. The thermal inertia characteristics include the significance of pipe storage effects and the heat wave propagation delay characteristics parameters. The economic constraints include modeling cost budget and software license restrictions parameters. Each key element is quantified.

4. The heating system analysis method for unified energy path and heat flow modeling of a heating network according to claim 3, characterized in that, The method for step S3 is as follows: Based on the seven key elements extracted in step S2, a hierarchical evaluation system is constructed, comprising seven primary indicators and 15 secondary indicators. The primary indicators include accuracy adaptability, efficiency adaptability, structural adaptability, dynamic adaptability, coupling adaptability, economic adaptability, and comprehensive robustness. Accuracy adaptability has an initial weight of 25%, comprising two secondary indicators: temperature field accuracy adaptability and heat load accuracy adaptability, calculated based on parameters corresponding to accuracy requirements. Efficiency adaptability has an initial weight of 20%, comprising two secondary indicators: single-time calculation time and iterative optimization efficiency, calculated based on parameters corresponding to computational efficiency requirements. Structural adaptability has an initial weight of 15%, comprising two secondary indicators: topology complexity adaptability and equipment type adaptability, calculated based on parameters corresponding to system scale characteristics. Dynamic adaptability has an initial weight of 15%, comprising two secondary indicators: temperature fluctuation adaptability and load change adaptability, calculated based on parameters corresponding to dynamic characteristic requirements. The initial weight of coupling adaptability is 10%, which includes two secondary indicators: primary and secondary network coupling adaptability and multi-energy coupling adaptability. It is calculated based on the parameters corresponding to the coupling degree of primary and secondary networks. The initial weight of economic adaptability is 8%, which includes two secondary indicators: software cost and human cost. It is calculated based on the parameters corresponding to economic constraints. The initial weight of comprehensive robustness is 7%, which includes two secondary indicators: extreme working condition performance and parameter sensitivity. It is calculated based on the parameters corresponding to thermal inertia characteristics. The evaluation system uses a score of 0 to 10, and calculates the comprehensive compatibility score of the two modeling methods, unified energy path and thermal energy flow, by weighted summation. The system introduces the entropy weight method to dynamically adjust the weights of each indicator, and also incorporates an expert knowledge base to correct the indicator weights. The evaluation system is then validated and optimized using a historical case database.

5. The heating system analysis method for unified energy path and heat flow modeling of a heating network according to claim 4, characterized in that, In step S4: The AI ​​decision model is composed of three functional modules: feature encoder, evaluation calculation engine and decision optimizer, which work together in sequence and are executed in the logical order of "feature extraction - suitability scoring - decision optimization". The feature encoder is used to convert key elements into high-dimensional feature vectors; The evaluation calculation engine is used to calculate the objective weights of primary indicators based on the information entropy of the high-dimensional feature vector. Based on the objective weights of the primary indicators and the subjective weights of experts, a weighted average method is used to fuse them to obtain the final dynamic weights of the primary indicators and simultaneously optimize the weight ratios of secondary indicators. A mapping relationship is established between the high-dimensional feature vector and 15 secondary indicators. A triangular membership function is used to transform the feature information into three categories of fuzzy semantics: "high," "medium," and "low." After fuzzy inference synthesis, the centroid method is used to defuzzify and obtain the quantitative scores of secondary indicators from 0 to 10. Subsequently, according to the weight ratio of the secondary indicators under the primary indicators, a weighted summation formula is used to calculate the scores of the unified energy path method and the thermal energy flow method on the seven primary indicators. Finally, the scores of the seven primary indicators of the two modeling methods are weighted and summed to obtain the comprehensive suitability score of the unified energy path method. Overall compatibility score with thermal energy flow method It also outputs the final dynamic weights of the seven primary indicators and the quantitative scores of each primary indicator. The decision optimizer represents the state space using high-dimensional feature vectors and the action space using "selecting a unified energy path method," "selecting a thermal energy flow method," or "selecting a hybrid modeling method." It constructs a reward function and trains the decision optimizer using a deep deterministic policy gradient algorithm. Through interaction between the agent and the simulation environment, it learns the tendency of the modeling method that balances accuracy and efficiency. Finally, it calculates the difference in the comprehensive fitness score between the two modeling methods. The recommendation confidence C is calculated by combining the Q-value of the reinforcement learning action value function normalized to [0, 1]. When ≥3, C=min (0.95+0.05×Q,1); when 1≤ When <3, C = 0.7 + 0.3 × Q; when When C < 1, C = 0.5 + 0.5 × Q; finally, the optimal modeling method is determined by combining the comprehensive suitability score, confidence level, and engineering constraint rules, and the core decision information is output, specifically including the optimal modeling method recommendation, recommendation confidence level, and ranking of key influencing factors; the optimal modeling method recommendation is specifically: when the comprehensive suitability score of the unified energy path method is < 1, C = 0.5 + 0.5 × Q; finally, the optimal modeling method is determined by combining the comprehensive suitability score of the unified energy path method with the confidence level, and the ranking of key influencing factors. Greater than the overall adaptability score of the thermal energy flow method Furthermore, when the confidence level is greater than or equal to 80%, the unified energy path method is recommended for modeling the entire system. When the unified energy path method is comprehensively adapted, the score is... Less than or equal to the overall adaptability score of the thermal energy flow method When the confidence level is greater than or equal to 80%, the thermal energy flow method is used to model the entire system; when the confidence level is between 60% and 80%, a hybrid modeling strategy is initiated; when the confidence level is less than 60%, a manual intervention prompt is issued to the user, who can then choose the optimal modeling scheme. The ranking of key influencing factors is specifically based on the absolute value of the contribution of the primary indicators to the decision, and the absolute value of the contribution of the primary indicators to the decision is obtained by multiplying the weight of the primary indicator by its score.

6. The heating system analysis method for unified energy path and heat flow modeling of a heating network according to claim 5, characterized in that, The hybrid modeling strategy specifically involves creating functional feature indices. Used to automatically divide the system's functional areas: The region with a value >0.45 is designated as a "long-distance primary network region" and modeled using the unified energy path method. The region is divided into "primary and secondary network coupling segment region" with a range of 0.45, and modeled using the thermal energy flow method. Functional characteristic index The calculation formula is: ； In the formula: Let be the cumulative pipe length from node i to the heat source. This represents the distance from the furthest node in the entire network to the heat source. Let be the maximum rate of temperature change at node i. This represents the maximum rate of temperature change in the entire network. The number of heat exchangers within 100m around node i.

Images