Intelligent selection and optimization method of building composite phase change materials based on large language model

Abstract

The present invention discloses a method for intelligent selection and optimization of building composite phase-change materials based on a large language model, belonging to the technical field of building energy conservation and intelligent material design. The method collects basic thermal property data, building environmental parameter data, and user demand data; utilizes deep reinforcement learning technology to construct an intelligent model based on a deep Q-network strategy, and integrates the generated data samples into a thermal property database; uses the large language model to output candidate material combination recommendations; uses semantic label vectors, and a ranking engine evaluates the candidate material combination recommendations to obtain performance evaluation results; generates a performance evaluation report based on the simulation model; and uses a multi-agent negotiation algorithm to optimize cross-region thermal control performance, generate multi-dimensional performance comparison charts and tuning suggestions, and record designer correction information, optimizing and adaptively providing feedback. The present invention screens building composite phase-change material combinations, optimizes cross-region thermal control, outputs adaptation solutions and suggestions, and enhances the system's adaptive optimization capabilities.

Description

Technical Field

[0001] The present invention belongs to the technical field of building energy conservation and intelligent material design, and specifically provides an intelligent selection and optimization method for building composite phase change materials based on a large language model. Background Art

[0002] With the deepening implementation of the "dual carbon" strategy, building energy conservation is increasingly dependent on phase change energy storage materials. Composite phase change materials, due to their unique performance in thermal comfort regulation and energy consumption peak reduction, have become a key component of new energy-saving envelope structures. However, the current selection of PCM materials relies primarily on expert experience or manual parameter adjustment, which leads to problems such as inconsistent standards, long design cycles, and highly subjective combination schemes. In addition, the multidimensional and complex variables of the building environment require material combinations that can adapt to the coupled characteristics of different regions, climates, and building functions. Traditional methods have difficulty achieving accurate and efficient matching.

[0003] In recent years, artificial intelligence (AI) technologies have emerged as a key enabler of material design and building energy prediction. The rise of reinforcement learning and large language models (LLMs) in particular has provided a novel approach to material property modeling, combination evaluation, and solution generation. While some initial attempts have been made to employ AI to aid material prediction, these efforts often remain at the data mining level, lacking systematic combination design capabilities, semantic parsing mechanisms, and real-time feedback and adjustment capabilities. Furthermore, the current PCM selection process generally lacks intelligent support systems such as modularization, closed-loop optimization, and visual recommendations, making it difficult to meet the increasing thermal control demands of high-performance buildings. Summary of the Invention

[0004] In response to the shortcomings in the background technology, the purpose of the present invention is to propose an intelligent selection and optimization method for building composite phase change materials based on a large language model. Through systematic data collection and deep reinforcement learning technology, it can automatically expand the thermal property data set and improve the diversity and credibility of the data. At the same time, it uses a large language model for semantic analysis to convert user needs into structured labels, realize accurate modeling of material properties and identification of optimal combination paths, introduce a computational fluid dynamics model for thermal response simulation, and ensure that the dynamic thermal response capability and adaptability of the material combination are quantitatively evaluated. In addition, based on the multi-agent negotiation algorithm, a personalized material recommendation list is generated, which not only optimizes the cross-regional thermal control performance, but also provides real-time feedback and tuning suggestions, thereby improving the intelligence, modularization, closed-loop optimization and visual recommendation capabilities of building energy-saving material selection, providing a scientific and effective solution to the thermal control needs of high-performance buildings and promoting the sustainable development of the construction industry.

[0005] The technical solution adopted by the present invention is:

[0006] A method for intelligent selection and optimization of building composite phase change materials based on a large language model includes the following steps:

[0007] S1. Collect basic thermal properties data of building composite phase change materials, building environment parameter data and user demand data;

[0008] S2. Based on the data obtained in S1, using deep reinforcement learning technology, build an intelligent model based on deep Q network strategy, set the state space, action space and reward function, automatically expand the basic thermal property data, simulate the potential distribution of thermal property data under different building environmental conditions, generate diverse and highly reliable thermal property data samples, and integrate the generated data samples into the thermal property database;

[0009] S3. Using a large language model, based on building environment parameter data, user demand data, and a database, the building type, climate parameters, and energy-saving demand information entered by the user are converted into a structured semantic label vector. This is combined with the building climate knowledge graph and the material performance ontology to perform semantic modeling, identify the optimal material combination path, and output a candidate material combination recommendation scheme containing a semantic label vector.

[0010] S4. Utilize semantic label vectors and a combination effectiveness ranking engine to evaluate the multi-dimensional performance indicators of different PCM combinations in the candidate material combination recommendation scheme, obtaining performance evaluation results. Based on a computational fluid dynamics simulation model, perform thermal response simulation on the material combination, output temperature fields and energy fluxes, and quantify the dynamic thermal response capability and scenario adaptability of the material combination to generate a performance evaluation report.

[0011] S5. Generate a personalized material recommendation list and candidate material combination recommendation plan based on the performance evaluation results. Use a multi-agent negotiation algorithm to simulate material combination suggestions for different building areas and optimize cross-regional thermal control performance. At the same time, generate multi-dimensional performance comparison charts, applicable scenario descriptions, and material tuning suggestions, and record the designer's correction information to optimize and adaptively provide feedback.

[0012] Preferably, in step S2, the state space formula is as follows:

[0013]

[0014] In the formula, represents the state vector, represents the specific heat capacity, represents thermal conductivity, represents the latent heat of phase change, Indicates the ambient temperature, represents the internal heat load of the building, Indicates the target temperature.

[0015] Preferably, in step S2, the formula of the action space is as follows:

[0016]

[0017] In the formula, represents the action vector, represents the specific heat capacity adjustment step, represents the thermal conductivity adjustment step, Represents the adjustment step of phase change latent heat.

[0018] Preferably, in step S2, the formula of the reward function is as follows:

[0019]

[0020] In the formula, Indicates time The reward value, 、 、 represents the weight coefficient, represents the change in thermal potential, Indicates the temperature stability index, Preferably, in step S3, the formula for identifying the optimal material combination path through the multi-index decision-making model function is as follows:

[0021]

[0022] In the formula, Indicates the overall target value; Represents the total number of performance indicator functions; Indicates the The weight of each performance indicator; Indicates index subscript; Represents the corresponding performance indicator function; represents the material parameter vector; Represents the total number of performance indicator functions.

[0023] Preferably, in step S4, a combination efficiency ranking engine is used to evaluate the multi-dimensional performance indicators of different PCM combinations in the candidate material combination. The evaluation of the multi-dimensional performance indicators is achieved by the following formula:

[0024]

[0025] In the formula, Represents the comprehensive performance score, i.e., the performance evaluation result, represents the corresponding weight coefficient, Indicates the performance indicators, Indicates the total number of performance indicators.

[0026] Preferably, in step S4, the calculation formula of the temperature field is as follows:

[0027]

[0028] In the formula, is the material density, is the specific heat capacity, is the temperature field, For time, is the thermal conductivity, is the temperature change rate; It is an internal heat source or absorption source.

[0029] Preferably, in step S4, the calculation formula of energy flux is as follows:

[0030]

[0031] In the formula, represents the energy flux, represents thermal conductivity, represents the normal temperature gradient.

[0032] Preferably, in step S5, the cross-region thermal control performance is optimized by the following objective function:

[0033]

[0034] In the formula, Material configuration for each area, For the The actual energy consumption of each region, is the target energy consumption, It is the coordination consistency index of the material combination scheme. To adjust the penalty coefficient of the weights between the balance indicators, Indicates the total number of regions, Indicates a counting subscript.

[0035] Preferably, in step S5, the correction information is updated by the following formula:

[0036]

[0037] in, Represents the parameter correction value based on the negotiation results and feedback adjustment, Represents the original material parameters, Represents the updated parameters.

[0038] Compared with the prior art, the present invention proposes an intelligent selection and optimization method for building composite phase change materials based on a large language model. The advantages of this method are:

[0039] Through systematic data collection and deep reinforcement learning technology, the present invention can automatically expand the thermal property data set and improve the diversity and credibility of the data. At the same time, it uses a large language model for semantic analysis to convert user needs into structured labels, realize accurate modeling of material properties and identification of optimal combination paths, introduce computational fluid dynamics models for thermal response simulation, and ensure that the dynamic thermal response capability and adaptability of the material combination are quantitatively evaluated. In addition, based on the multi-agent negotiation algorithm, a personalized material recommendation list is generated, which not only optimizes cross-regional thermal control performance, but also provides real-time feedback and tuning suggestions, thereby improving the intelligence, modularity, closed-loop optimization and visual recommendation capabilities of building energy-saving material selection, providing a scientific and effective solution to the thermal control needs of high-performance buildings and promoting the sustainable development of the construction industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the steps of the method of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of this application to further clearly and completely describe the technical solutions in the embodiments of this application. It should be noted that the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making any creative work shall fall within the scope of protection of this application.

[0042] In order to make the invention objectives, technical solutions and advantages of this application clearer, the embodiments of this application are further described in detail in conjunction with the drawings in the specification: In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the advantages of the present invention will be further illustrated by comparing the embodiments in conjunction with the drawings and specific implementation methods.

[0043] The present invention proposes an intelligent selection and optimization method for building composite phase change materials based on a large language model. Figure 1 As shown, the steps of the method are described in detail:

[0044] S1. Collect basic thermal properties data of building composite phase change materials, building environment parameter data and user demand data;

[0045] Specifically, in order to realize the intelligent selection and optimization of building composite phase change materials, it is necessary to first systematically collect relevant basic data through a variety of efficient technical means; specifically, using advanced sensor technology and data acquisition systems, multifunctional thermophysical sensors are arranged at key locations of different types of buildings to monitor and collect core thermophysical parameters such as thermal conductivity, specific heat capacity, latent heat capacity, etc. of phase change materials in real time, that is, basic thermophysical data, to ensure high accuracy and timeliness of data; at the same time, drones and laser scanning technology are used to obtain geometric information of building structure and outer skin, providing a spatial basis for thermal performance analysis; in order to enrich and improve the thermophysical database, laboratory simulation tests and on-site sample analysis are also combined, and differential scanning calorimeters (DSC), laser thermal analyzers and other equipment are used to characterize the thermal performance and test parameters of material samples; in terms of building environmental parameters, physical The Internet of Things (IoT) platform integrates multi-source data collection equipment such as weather stations, indoor and outdoor temperature and humidity sensors, wind speed and solar radiation meters, etc., to achieve continuous, all-weather, multi-dimensional monitoring of climate parameters. At the same time, it uses environmental simulation systems (such as programmable air flow and temperature control boxes) to simulate different climate conditions, enabling the rapid collection of multi-scene and multi-scenario environmental data, namely building environmental parameter data. To obtain user demand data, structured surveys and data entry are conducted through electronic means such as smart questionnaires, mobile applications, and building management systems (BMS). Users' energy-saving goals, comfort requirements, usage habits, and design preferences, namely user demand data, are collected. In addition, big data analysis and machine learning technologies are used to process and analyze the collected large-scale multi-source data, thus providing a solid data foundation for subsequent deep learning training and expansion of thermal physical property datasets.

[0046] S2. Based on the data obtained in S1, using deep reinforcement learning technology, build an intelligent model based on deep Q network strategy, set the state space, action space and reward function, automatically expand the basic thermal property data, simulate the potential distribution of thermal property data under different building environmental conditions, generate diverse and highly reliable data samples, and integrate the generated data samples into the basic thermal property database;

[0047] Specifically, using Deep Reinforcement Learning (DRL) technology, we built an intelligent model based on the Deep Q-Network (DQN) strategy to effectively achieve automatic expansion of thermal property data and potential distribution simulation;

[0048] Specifically, the state space is defined to ensure that the intelligent model can fully reflect the thermal performance information of the material and the environment. The definition formula of the state space is as follows:

[0049]

[0050] In the formula, represents the state vector, represents the specific heat capacity, represents thermal conductivity, represents the latent heat of phase change, Indicates the ambient temperature, represents the internal heat load of the building, Indicates the target temperature;

[0051] The action space corresponds to the adjustment strategy of the thermophysical property data, which can explore the potential effects of different parameter changes. The setting formula of the action space is as follows:

[0052]

[0053] In the formula, represents the action vector, represents the specific heat capacity adjustment step, represents the thermal conductivity adjustment step, represents the adjustment step of phase change latent heat;

[0054] The reward function is used to guide the intelligent model to learn the optimal strategy. The definition formula of the reward function is:

[0055]

[0056] In the formula, Indicates time The reward value, 、 、 represents the weight coefficient, represents the change in thermal potential, Indicates the temperature stability index, represents the simulation error;

[0057] This function combines the cumulative thermal potential change , temperature stability index and simulation error , by adjusting the weight coefficient 、 、 , ensuring that the intelligent model takes into account both data credibility and simulation accuracy when exploring the potential thermal performance space. This strategy enables the intelligent model to automatically simulate the potential distribution of thermal physical property data under different environmental conditions, thereby generating diverse and highly reliable data samples;

[0058] The generated data samples not only enrich the existing data, but also improve the intelligent model's predictive ability for extreme and under-covered scenarios. After each simulated sample is quality screened, the highly credible data samples are integrated to form a dynamically updated, highly adaptable thermal property database. The core advantage of this method lies in its automation, intelligence, and efficiency. It not only saves a lot of experimental and manual research costs, but also significantly improves the accuracy and reliability of modeling. Through this technical means of combining deep reinforcement learning with physical models, it is possible to continuously learn and optimize the potential distribution of thermal performance parameters, providing strong theoretical support and data guarantee for the precise selection and performance prediction of building materials.

[0059] S3. Using a large language model, based on building environment parameter data, user demand data, and a database, the building type, climate parameters, and energy-saving demand information entered by the user are converted into a structured semantic label vector. This is combined with the building climate knowledge graph and the material performance ontology to perform semantic modeling, identify the optimal material combination path, and output a candidate material combination recommendation scheme containing a semantic label vector.

[0060] Specifically, in this step, advanced large language models (such as the GPT series) are used to deeply understand and semantically extract the unstructured text information such as building type, climate parameters, and energy-saving requirements input by the user. In specific operations, the natural language processing capabilities of the large language model are used to convert these unstructured text information into high-dimensional structured semantic label vectors (such as a set of keywords or semantic features in the vector space), thereby realizing the digitization and standardization of information. Next, these semantic label vectors are integrated with the building climate knowledge graph and material performance ontology for semantic modeling, and a structured knowledge system is established for the relationship between different material properties. Through graph reasoning and ontology inference, the optimal material combination path that best meets user needs can be automatically identified, that is, an optimal set of materials is found to achieve the best match in terms of thermal performance, environmental adaptability, and energy-saving effect. Finally, the output candidate material combination recommendation plan not only includes specific material information, but also embeds rich semantic label vectors to facilitate subsequent multi-dimensional performance evaluation and scenario matching.

[0061] Use the multi-index decision model to identify the optimal material combination path, and multiply the performance indicators of each material combination by the corresponding performance function results according to the weight to obtain the sum , the formula is as follows:

[0062]

[0063] in, Indicates the overall target value; Represents the total number of performance indicator functions; Indicates the The weights of the performance indicators reflect the importance of different indicators, such as thermal performance, cost efficiency or environmental adaptability; Indicates index subscript; Represents the corresponding performance index function to evaluate a given material combination In the performance in various aspects; represents the material parameter vector;

[0064] By optimizing It can screen and sort candidate material combination recommendations in a clear, scientific and reasonable manner, greatly improving the intelligence, personalization and scientific nature of material selection. It not only improves design efficiency, but also ensures that the recommended material combination recommendations are highly adaptable and credible in practical applications.

[0065] S4. Utilize semantic label vectors and a combination effectiveness ranking engine to evaluate the multi-dimensional performance indicators of different PCM combinations in the candidate material combination recommendation scheme, obtaining performance evaluation results. Based on a computational fluid dynamics simulation model, perform thermal response simulation on the material combination, output temperature fields and energy fluxes, and quantify the dynamic thermal response capability and scenario adaptability of the material combination to generate a performance evaluation report.

[0066] Specifically, the semantic label vectors obtained through semantic parsing using a large language model are combined with the evaluation of multi-dimensional performance indicators, and a combination effectiveness ranking engine is used to conduct a comprehensive evaluation of the various performance aspects of different PCM (phase change material) combinations in the candidate material combination recommendation scheme.

[0067] Specifically, the ranking engine evaluates the performance by introducing a weighted scoring model. The formula is as follows:

[0068]

[0069] in, Indicates the performance indicators (such as thermal response speed, energy storage efficiency, temperature regulation ability, environmental adaptability, etc.), Indicates the corresponding weight coefficient, reflecting the importance of the indicator; Indicates the total number of performance indicators;

[0070] By assigning weights to different performance indicators and summing their scores, we can balance various performance requirements and select material combinations with good comprehensive performance in multiple aspects such as thermal response and energy storage, avoiding focusing on only a single performance.

[0071] Next, to achieve dynamic simulation of the thermal performance of the material combination (i.e., thermal response simulation), the system introduces a simulation model based on computational fluid dynamics (CFD). By solving the heat conduction equation, the temperature field and energy flux of the material under different boundary conditions and environmental scenarios are simulated. The temperature field calculation formula is as follows:

[0072]

[0073] in, is the material density, is the specific heat capacity, is the temperature field, For time, is the thermal conductivity, is the temperature change rate; As an internal heat source or absorption source, numerical simulation can be used to quantify in detail the dynamic thermal response capabilities of each material combination in different scenarios, namely temperature changes, response time, and energy transfer efficiency, thereby evaluating its scenario adaptability;

[0074] The energy flux is calculated as follows:

[0075]

[0076] In the formula, represents the energy flux, represents thermal conductivity, represents the normal temperature gradient;

[0077] Ultimately, the simulated temperature field distribution and energy flux data are combined with multiple performance indicators to form a complete performance evaluation report. This process greatly enhances the scientific nature and reliability of material performance, ensuring that the design plan is not only based on static parameters but also includes dynamic thermal response characteristics, ensuring that the recommended material combination has excellent practical application performance and adaptability. At the same time, the use of a multi-indicator weighted scoring method ensures the balance and optimization of various performance aspects, providing a scientific basis for energy saving, thermal regulation, and comfort of building systems.

[0078] S5. Generate personalized material recommendation lists and candidate material combination recommendations based on performance evaluation results. Using a multi-agent negotiation algorithm, simulate material combination recommendations for different building areas to optimize cross-regional thermal control performance. Simultaneously, generate multi-dimensional performance comparison charts, applicable scenario descriptions, and material tuning recommendations. The system also records designer corrections, optimizes, and adaptively provides feedback.

[0079] Specifically, in the process of generating personalized material recommendation lists and candidate material combination recommendation schemes based on performance evaluation results, we first screen out candidate material combinations that meet the specific needs of each region based on the performance evaluation indicators of each region to ensure that each region can achieve the optimal match of its thermal performance, energy-saving goals and environmental adaptability; then, we introduce a multi-agent negotiation algorithm to The algorithm simulates the collaboration and game-playing among multiple agents (representing different areas or design objectives). Each agent represents a building area or subsystem, with specific parameters, performance targets, and preferences. In the first round, each agent proposes preliminary material combination recommendations based on local performance indicators and the current recommended solution. Subsequently, through an information exchange mechanism, each agent gradually shares its own solutions, performance results, and feedback on solutions from other areas. Through negotiation algorithms (such as negotiation protocols in game theory, online weight adjustment, and dynamic compromise), agents continuously adjust their candidate material combination recommendations to achieve global performance optimization. For example, a particular area may prioritize maximizing thermal stability. While maintaining this goal, the agent will refer to the recommendations of other areas and gradually adjust the material combination ratio to reduce overall energy consumption or improve cross-area thermal synergy.

[0080] During the negotiation process, the cross-region thermal control performance is optimized using the optimization objective function, as follows:

[0081]

[0082] in, For the The actual energy consumption of each region, is the target energy consumption, It is the coordination consistency index of the material combination scheme. In order to adjust the penalty coefficient of the weights between the balance indicators, by continuously optimizing the objective function, it is possible to improve the thermal control efficiency and coordination of the building system as a whole while meeting the needs of individual areas;

[0083] During the optimization process, the correction information is updated according to:

[0084]

[0085] in, Represents the parameter correction value based on the negotiation results and feedback adjustment, Represents the original material parameters, Represents the updated parameters;

[0086] After each round of optimization, a multi-dimensional performance comparison chart is automatically generated, which intuitively shows the performance differences of different solutions in terms of thermal performance, energy consumption, comfort, etc., while providing scenario applicability instructions and material tuning suggestions, giving designers a scientific and detailed reference basis; in addition, all designer corrections and preference information are recorded and archived for system self-learning and iterative optimization, gradually forming a highly adaptable and intelligent building material selection and tuning model; this technical solution integrates multi-agent negotiation, dynamic optimization and visual analysis, providing a comprehensive, systematic and innovative solution for building energy-saving design, effectively improving the collaborative efficiency of cross-regional thermal control.

[0087] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0088] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. An intelligent selection and optimization method for building composite phase change materials based on a large language model, characterized in that: include: S1. Collect basic thermal properties data of building composite phase change materials, building environment parameter data and user demand data; S2. Based on the data obtained in S1, using deep reinforcement learning technology, build an intelligent model based on deep Q network strategy, set the state space, action space and reward function, automatically expand the basic thermal property data, simulate the potential distribution of thermal property data under different building environmental conditions, generate diverse and highly reliable thermal property data samples, and integrate the generated data samples into the thermal property database; The formula of the action space is as follows: ； In the formula, represents the action vector, represents the specific heat capacity adjustment step, represents the thermal conductivity adjustment step, represents the adjustment step of phase change latent heat; S3. Using a large language model, based on building environment parameter data, user demand data, and a database, the building type, climate parameters, and energy-saving demand information entered by the user are converted into a structured semantic label vector. This is combined with the building climate knowledge graph and the material performance ontology to perform semantic modeling, identify the optimal material combination path, and output a candidate material combination recommendation scheme containing a semantic label vector. S4. Utilize semantic label vectors and a combination effectiveness ranking engine to evaluate the multi-dimensional performance indicators of different PCM combinations in the candidate material combination recommendation scheme, obtaining performance evaluation results. Based on a computational fluid dynamics simulation model, perform thermal response simulation on the material combination, output temperature fields and energy fluxes, and quantify the dynamic thermal response capability and scenario adaptability of the material combination to generate a performance evaluation report. S5. Generate personalized material recommendation lists and candidate material combination recommendations based on performance evaluation results. Using a multi-agent negotiation algorithm, simulate material combination recommendations for different building areas to optimize cross-regional thermal control performance. Simultaneously, generate multi-dimensional performance comparison charts, applicable scenario descriptions, and material tuning recommendations. The system also records designer corrections, optimizes, and adaptively provides feedback. The cross-region thermal control performance is optimized by the following objective function: ； In the formula, Material configuration for each area, For the The actual energy consumption of each region, is the target energy consumption, It is the coordination consistency index of the material combination scheme. To adjust the penalty coefficient of the weights between the balance indicators, Indicates the total number of regions, Indicates a counting subscript.

2. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S2, the formula of the state space is as follows: ； In the formula, represents the state vector, represents the specific heat capacity, represents thermal conductivity, represents the latent heat of phase change, Indicates the ambient temperature, represents the internal heat load of the building, Indicates the target temperature.

3. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S2, the formula of the reward function is as follows: ； In the formula, Indicates time The reward value, represents the weight coefficient, represents the change in thermal potential, Indicates the temperature stability index, Represents the simulation error.

4. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S3, the formula for identifying the optimal material combination path through the multi-index decision-making model function is as follows: ； In the formula, Indicates the overall target value; Represents the total number of performance indicator functions; Indicates the The weight of each performance indicator; Indicates index subscript; Represents the corresponding performance indicator function; represents the material parameter vector; Represents the total number of performance indicator functions.

5. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S4, the combination efficiency ranking engine is used to evaluate the multi-dimensional performance indicators of different PCM combinations in the candidate material combination. The evaluation of the multi-dimensional performance indicators is achieved by the following formula: ； In the formula, Represents the comprehensive performance score, i.e., the performance evaluation result, represents the corresponding weight coefficient, Indicates the performance indicators, Indicates the total number of performance indicators.

6. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S4, the calculation formula of the temperature field is as follows: ； In the formula, is the material density, is the specific heat capacity, is the temperature field, For time, is the thermal conductivity, is the temperature change rate; is an internal heat source or absorption source, It is the partial derivative of temperature with respect to time, reflecting how fast the temperature changes with time.

7. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S4, the calculation formula of energy flux is as follows: ； In the formula, represents the energy flux, represents thermal conductivity, represents the normal temperature gradient.

8. The intelligent selection and optimization method for building composite phase change materials based on a large language model according to claim 1 is characterized in that: In step S5, the correction information is updated using the following formula: ； in, Represents the parameter correction value based on the negotiation results and feedback adjustment, Represents the original material parameters, Represents the updated parameters.

Images