LLM-based comprehensive energy model establishing and solving system

Abstract

The invention relates to an LLM-based integrated energy model establishing and solving system. The system comprises a user input module, an integrated energy knowledge base module, an LLM module and a plurality of postposition modules used for automatically establishing an integrated energy model according to an analysis result of the LLM module and solving the integrated energy model. The user input module is used for receiving a comprehensive energy planning demand described by a user in a natural language; the comprehensive energy knowledge base module is composed of a scene template base, an equipment model base, a constraint rule base and a solver strategy base. And the LLM module is used for receiving the integrated energy planning demand described by a user in a natural language and a combined text of all contents in the integrated energy knowledge base module, constructing a prompt project based on the combined text, and calling an LLM interface to generate an analysis result. According to the system, the natural language requirement of the user can be converted into a structured modeling process, the establishment of the comprehensive energy system is automatically completed, and the modeling efficiency of the comprehensive energy system is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of integrated energy system planning, in particular to an integrated energy model establishment and solution system based on LLM. BACKGROUND

[0002] With the advancement of the double carbon strategy, integrated energy systems, as an important means to improve energy utilization efficiency and optimize energy structure, have received widespread attention. Integrated energy models usually involve multiple energy forms such as electricity, heat, cold, and gas, and have characteristics such as complex energy flow, diverse participants, and multiple optimization objectives. How to efficiently and intelligently complete the model establishment and solution of integrated energy systems has become a hot topic in the cross-research of energy engineering and artificial intelligence.

[0003] Currently, the establishment and solution of energy models in integrated energy system planning and design mainly rely on expert experience or traditional optimization methods based on fixed models. These methods usually have the following problems: (1) low user participation, model construction relies on professional modeling personnel, high threshold, low efficiency; (2) rigid modeling process, difficult to flexibly adapt to customized needs of different scenarios; (3) scattered model knowledge, lack of unified energy knowledge representation and reasoning ability; (4) single solution process, difficult to cope with complex multi-constraint, multi-objective planning tasks. SUMMARY

[0004] Therefore, an integrated energy model establishment and solution system based on LLM is provided, which can convert the user's natural language requirements into a structured modeling process, automatically complete the establishment of an integrated energy system, and effectively improve the modeling efficiency of an integrated energy system.

[0005] The present application provides an integrated energy model establishment and solution system based on LLM, comprising: a user input module for receiving user's natural language description of integrated energy planning requirements; an integrated energy knowledge base module composed of a scenario template library, a device model library, a constraint rule library, and a solver strategy library; an LLM module for receiving user's natural language description of integrated energy planning requirements and combined text of all contents in the integrated energy knowledge base module, and constructing a prompt engineering based on the combined text, calling an LLM interface to generate an integrated energy system semantic configuration package, the integrated energy system semantic configuration package including a scenario semantic description dataset, a device selection and parameter configuration dataset, a target and constraint definition dataset, a data association and input interface dataset, and a solution strategy and engine configuration dataset; a scenario building module for constructing an integrated energy system scenario according to the scenario semantic description dataset; An equipment selection and parameter configuration module is configured to construct an integrated energy system model according to an equipment selection and parameter configuration dataset and an integrated energy system scenario; A target constraint setting module is configured to determine an optimization target of the integrated energy system model according to a target and constraint definition dataset; A data processing module is configured to select external data or local data according to a data association and input interface dataset; A model solving and output module is configured to determine a solver and solver parameters according to a solving strategy and engine configuration dataset, input the data selected by the data processing module into the solver, and solve the integrated energy system model based on the optimization target, and output an optimal configuration scheme and performance indicators.

[0006] In one embodiment, the external data or the local data each includes load data, meteorological data, and energy prices; The load data includes time-sharing data of cold, heat, electricity, hydrogen, and ammonia loads; The meteorological data includes time-sharing data of temperature, sunshine, and wind speed meteorological resources; The energy prices include price information of cold, heat, electricity, hydrogen, ammonia, natural gas, and diesel.

[0007] In one embodiment, the LLM-based integrated energy system model establishment and solving system further includes a semantic verification module configured to review and forward an integrated energy system semantic configuration package, and if the review is qualified, forward the integrated energy system semantic configuration package, otherwise, prompt a user intervention or trigger an automatic correction mechanism; The semantic verification module reviews the integrated energy system semantic configuration package in a manner including rule matching, boundary checking, and consistency checking.

[0008] In one embodiment, the LLM-based integrated energy system model establishment and solving system further includes a splicing module configured to splice a user's natural language description of an integrated energy planning requirement with all contents of an integrated energy knowledge base module to form a combined text.

[0009] In one embodiment, the integrated energy system scenario includes an energy form structure and an energy flow topology structure; The energy form structure is configured to determine one or more of five energy forms including cold, heat, electricity, hydrogen, and ammonia in the integrated energy system scenario; The energy flow topology structure is configured to determine energy flow topology relationships of the energy forms.

[0010] In one embodiment, the target and constraint definition dataset includes an upper limit of total investment, an emission factor limit, an energy abandonment constraint, and an energy supply ratio constraint.

[0011] In one embodiment, the optimal configuration scheme and performance indicators include: An energy configuration structure: installed capacity of each device; Each device operation strategy: time-sharing operation output of each device; Economic evaluation results: capital gain rate, static payback period, dynamic payback period, static total investment; emission results: annual carbon dust, annual carbon dioxide, annual sulfur dioxide, annual nitrogen oxides, annual PM2.5, and annual equivalent tree planting trees.

[0012] The beneficial effects of the present application are: (1) The present application adopts a large language model as a core intelligent agent, which can automatically analyze the natural language requirements input by the user, automatically complete the multi-link configuration of the modeling process, significantly reduce the dependence on professional artificial modeling, and improve the overall modeling efficiency.

[0013] (2) The large language model of the present application combines the energy knowledge base for semantic understanding and reasoning, which can automatically convert the user's unstructured language description into structured scene configuration, device parameters, constraint conditions and solving objectives, and realize highly intelligent modeling automation.

[0014] (3) The present application constructs a unified comprehensive energy knowledge base, which covers scene templates, device models, constraint rules and solver knowledge, and can be flexibly called according to different user requirements, adapting to various energy system planning scenarios.

[0015] (4) The present application also introduces a semantic verification module to review the intermediate configuration results generated by the large language model, improving the system stability and output reliability. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A schematic diagram of the framework structure of the comprehensive energy model establishment and solving system based on LLM provided in the embodiments of the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0018] Large language model (LLM, Large Language Model) has made breakthrough progress in natural language processing, knowledge reasoning, intelligent interaction and other fields, and its ability in semantic understanding, information integration and intelligent decision-making provides a new solution path for the intelligent establishment of comprehensive energy model. The large language model combined with the professional knowledge base in the energy field can convert the natural language requirements of the user into a structured modeling process, automatically complete scene construction, data processing, device parameter configuration, constraint condition setting and solving model establishment, and effectively improve the modeling efficiency and simulation accuracy.

[0019] In one embodiment, such as Figure 1 As shown, this embodiment establishes a solution system based on an LLM-based integrated energy model, consisting of a pre-module, an LLM module, and a post-module. The pre-module includes a user input module and an integrated energy knowledge base module. The post-module includes a scenario building module, a device parameter construction module, an objective constraint setting module, a data processing module, and a model solving and output module. The LLM module receives data from the pre-module, processes it, and outputs the parsed result, namely the integrated energy system semantic configuration package, which is then sent to the post-module.

[0020] The modules and their functions are described below.

[0021] The user input module receives integrated energy planning requirements described by users in natural language. These requirements include usage scenarios (e.g., industrial parks, energy bases), energy types, equipment lists, available resource conditions (e.g., whether photovoltaic power is available, land area), target preferences (e.g., low-carbon priority, cost optimization), and constraints (e.g., investment limits, carbon emission caps).

[0022] For example, taking the integrated energy planning of a science and technology park as an example, a user enters the following natural language description in the system interface, i.e., the user input module: "I want to design an integrated energy system for a science and technology innovation park covering approximately three hectares. The park has five research and office buildings and one experimental factory, with a stable annual electricity load and winter heating demand. The plan is to install a distributed photovoltaic power generation system on the factory roofs and carports, considering the configuration of a certain scale of electrochemical energy storage to achieve self-consumption of photovoltaic power and peak shaving. For winter heating, we hope to use an electric boiler solution, maintaining a backup interconnection with the existing municipal heating network. When designing the system, please prioritize the scheme with the lowest carbon emissions, while ensuring that the investment does not exceed 5 million yuan. The park is located in xx city, xx province, with good sunshine conditions. The electricity price is based on a peak-valley time-of-use system, and the expected operating period after completion is 20 years. The final output should include recommendations for installed equipment capacity, system operation strategy, annual power generation and heating, and an assessment of economic efficiency and carbon reduction effects."

[0023] The integrated energy knowledge base module consists of a scenario template library, an equipment model library, a constraint rule library, and a solver strategy library.

[0024] The scenario template library includes configuration templates for various typical application scenarios, supporting five energy forms: cold, heat, electricity, hydrogen, and ammonia. Each energy form supports an energy flow topology structure of source-grid-load-storage.

[0025] For example, a configuration template for a certain scenario is as follows: Energy forms: electricity, heat; Energy topology: Photovoltaics → Energy Storage → Electric Load / Electric Boiler → Heat Load.

[0026] The equipment model library includes investment costs, operating costs, and efficiency parameters for various equipment such as photovoltaics, wind turbines, diesel generators, electrochemical energy storage, electric boilers, gas boilers, combined heat and power, thermal energy storage systems, solar thermal heating, chillers, direct-fired turbines, lithium bromide units, and cold storage systems.

[0027] Based on the equipment list returned by the scenario template, the system extracts the technical and economic parameters of the corresponding equipment from the equipment model library. For example, the parameters of a photovoltaic system model are: Photovoltaic module efficiency: 0.18–0.21; Cost: Approximately 3800–4200 RMB / kW; Operation and maintenance costs: approximately 1% per year.

[0028] The power generation model is based on a linear power output model of time-of-use solar irradiance and component efficiency, with a rated lifespan of 25 years.

[0029] The parameters of the electrochemical energy storage system model are: Charge / discharge efficiency: 0.93; Cost: Approximately 2500 RMB / kWh; State of charge (SOC) range: [0.1, 0.9]; Lifespan: Approximately 5,000 cycles.

[0030] The parameters of the electric boiler model are: Thermal efficiency: 0.98; Investment cost: 800 yuan / kW; Adjustable operating power range: 30%–100%.

[0031] The constraint rule library includes constraint rule templates for equipment installed capacity, investment amount, carbon emissions, and equipment output.

[0032] The constraints can be: the total investment is capped at xx yuan; the installed capacity of a certain equipment is within the range of [0, x]; and the curtailment rate of new energy is less than or equal to x.

[0033] The solver strategy library includes configuration parameters for optimizing the solver COPT, which include preprocessing configuration, maximum computation time, feasibility tolerance, integer tolerance, or heuristic strategy parameters.

[0034] For example, one use case of the solver strategy library includes: Optimization type: Linear multi-objective programming; Objective function: Minimize carbon emissions + suboptimal economics; Solver type: COPT; Parameter settings: Maximum calculation time 300 s, feasibility tolerance 1×10 -6Heuristic search is enabled.

[0035] The LLM module receives a combined text of the user's integrated energy planning requirements described in natural language and all content from the integrated energy knowledge base module. Based on this combined text, it constructs a prompt project and calls the LLM interface to generate an integrated energy system semantic configuration package. The integrated energy system semantic configuration package includes a scenario semantic description dataset, an equipment selection and parameter configuration dataset, an objective and constraint definition dataset, a data association and input interface dataset, and a solution strategy and engine configuration dataset. The integrated energy system semantic configuration package guides the execution of tasks in subsequent modeling modules.

[0036] The integrated energy system semantic configuration package, a structured semantic dataset output by a large language model, represents the semantic mapping between user needs and energy system modeling. This package contains key information such as scenario description, energy type, equipment composition, parameter constraints, optimization objectives, and solution strategies, serving as an intermediate layer connecting natural language input and the mathematical modeling process. For the upper-layer input module, it receives the user's natural language description and knowledge base references; for the lower-layer modules, it provides clear and structured input for the scenario building module, equipment parameter model construction module, and objective constraint module.

[0037] Specifically, the scenario semantic description dataset includes the following data subsets: scenario category (e.g., park, factory area, commercial complex, energy base); energy type (electricity, heat, cooling, gas, hydrogen, ammonia, etc.); scenario scale parameters (building area, land area, number of users, load level); operating cycle and typical time period (year, season, typical day); energy flow topology description (source-grid-load-storage relationship).

[0038] The equipment selection and parameter configuration dataset includes the following subsets: a list of recommended equipment types (such as photovoltaic, electric energy storage, electric boilers, combined heat and power, and chiller units); and key technical parameters for each type of equipment (rated power, efficiency, investment cost, lifespan, and operating range).

[0039] The target and constraint definition dataset includes the following data subsets: optimization objectives (such as minimizing carbon emissions, optimizing economics, maximizing energy efficiency, or a combination of multiple objectives); symbolic templates for constraint equations such as investment, emissions, equipment operation, and energy balance; and priority weights or multi-objective weight allocation coefficients.

[0040] The data association and input interface dataset includes the following data subsets: external data sources and their formats required to describe the model, including: load data file paths and types (time-of-use data such as electricity, heat, cold, and hydrogen); meteorological data sources (solar irradiance, wind speed, and temperature); and energy price inputs (time-of-use electricity prices, gas prices, and heating prices).

[0041] The solution strategy and engine configuration dataset includes the following subsets: solver type (e.g., COPT, Gurobi); solution strategy (single-objective / multi-objective, linear / nonlinear); computational parameters (maximum computation time, tolerance, heuristic settings, number of parallel threads).

[0042] The scenario building module is used to construct integrated energy system scenarios that meet the needs of integrated energy planning based on the scenario semantic description dataset.

[0043] Specifically, integrated energy system scenarios include energy form structure and energy flow topology.

[0044] Energy form structure is used to determine whether an energy system scenario includes one or more of the five energy forms: cold, heat, electricity, hydrogen, and ammonia; energy flow topology structure is used to determine the energy flow topology relationship of each energy form.

[0045] The equipment parameter construction module is used to build a comprehensive energy model based on equipment selection and parameter configuration datasets and integrated energy system scenarios. The comprehensive energy model is a mathematical model of the equipment that includes physical parameters, operating logic, capacity upper and lower limits, investment costs, and operating costs.

[0046] The objective and constraint setting module is used to determine the optimization objective of the integrated energy model based on the objective and constraint definition dataset. The objective and constraint definition dataset includes an upper limit on total investment, emission factor limits, energy curtailment constraints, and energy supply ratio constraints. The optimization objective of the integrated energy model can be one of three types: economical, environmentally friendly, or high-efficiency.

[0047] The data processing module is used to select external or local data based on data association and the input interface dataset. Specifically, both external and local data include load data, meteorological data, and energy prices.

[0048] Load data includes time-of-use data for cooling, heating, electricity, hydrogen, and ammonia loads; meteorological data includes time-of-use data for temperature, sunshine duration, and wind speed; energy prices include price information for cooling, heating, electricity, hydrogen, ammonia, natural gas, and diesel.

[0049] The model solving and output module is used to determine the solver and solver parameters based on the solving strategy and engine configuration dataset, and input the data selected by the data processing module into the solver to solve the integrated energy model based on the optimization objective, and output the optimal configuration scheme and performance indicators.

[0050] In one embodiment, the LLM-based integrated energy model solution system also includes a semantic verification module, which is used to review and forward the semantic configuration package of the integrated energy system. If the review is successful, the package is forwarded; otherwise, the user is prompted to intervene or an automatic correction mechanism is triggered. The semantic verification module reviews the semantic configuration package of the integrated energy system through rule matching, boundary checks, and consistency checks.

[0051] Rule matching utilizes various templates from the integrated energy knowledge base, such as "constraint rule templates" and "equipment parameter templates," to perform field-level matching and pattern comparison on the integrated energy planning semantic configuration package output by the LLM. When anomalies such as missing fields, mismatched types, or inconsistent units are detected, the system automatically triggers a correction mechanism or prompts the user for intervention.

[0052] Boundary checks are physical and empirical verifications of the numerical parameters and engineering variables in intermediate results to ensure that the generated equipment parameters, capacity settings and operating indicators are within reasonable ranges and to avoid configurations that are not engineering feasible.

[0053] Consistency checking and verification involves a logical-level global review of intermediate results to ensure logical consistency in energy flow relationships, constraint dependencies, and objective function settings within the model. Based on the intermediate results, an energy topology and constraint network model is constructed, and consistent reasoning is performed on energy balance, logical conditions, and the solution objective to determine whether any logical conflicts or dependency contradictions exist.

[0054] In one embodiment, the LLM-based integrated energy model building and solving system also includes a splicing module for splicing together the integrated energy planning requirements described by the user in natural language with all the contents of the integrated energy knowledge base module to form a combined text.

[0055] In one embodiment, the optimal configuration and performance metrics include: Energy configuration structure: installed capacity of each piece of equipment; Equipment operation strategy: Time-sharing output of each piece of equipment; Economic evaluation results: return on equity, static payback period, dynamic payback period, and total static investment; Emissions results: annual carbon dust, annual carbon dioxide, annual sulfur dioxide, annual nitrogen oxides, annual PM2.5, and annual equivalent of planting trees.

[0056] The overall operation flow of the LLM-based integrated energy model building and solving system in this embodiment is as follows: Users input their planning requirements via natural language. Combined with content from the integrated energy planning and design knowledge base, a prompt for the large language model is constructed. The large language model module performs semantic understanding, generating a series of intermediate modeling configurations for the planning and design phase. Then, each module collaboratively completes the scenario construction, data processing, equipment configuration, constraint setting, and model solving process, ultimately outputting a complete integrated energy optimal configuration scheme and performance indicators. This system enables intelligent modeling, automatic solving, and result interpretation driven by natural language, exhibiting a high degree of intelligence and applicability to various energy system planning scenarios.

[0057] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. An LLM-based integrated energy model building and solving system, characterized in that, The application comprises the following steps: a user input module for receiving user's natural language description of integrated energy planning requirements; an integrated energy knowledge base module composed of a scenario template library, a device model library, a constraint rule library, and a solver strategy library; an LLM module for receiving user's natural language description of integrated energy planning requirements and all contents in the integrated energy knowledge base module to form a combined text, and constructing a prompt engineering based on the combined text, calling an LLM interface to generate an integrated energy system semantic configuration package, which includes a scenario semantic description dataset, a device selection and parameter configuration dataset, a target and constraint definition dataset, a data association and input interface dataset, and a solving strategy and engine configuration dataset; a scenario building module for building an integrated energy system scenario according to the scenario semantic description dataset; a device parameter type building module for building an integrated energy model according to the device selection and parameter configuration dataset and the integrated energy system scenario; a target constraint setting module for determining the optimization target of the integrated energy model according to the target and constraint definition dataset; a data processing module for selecting external data or local data according to the data association and input interface dataset; a model solving and output module for determining a solver and solver parameters according to the solving strategy and engine configuration dataset, inputting the data selected by the data processing module into the solver to solve the integrated energy model based on the optimization target, and outputting an optimal configuration scheme and performance indicators.

2. The LLM-based integrated energy model building and solving system of claim 1, wherein, The external data or local data includes load data, meteorological data, and energy prices; The load data includes time-sharing data of cold, heat, electricity, hydrogen, and ammonia loads; The meteorological data includes time-sharing data of temperature, sunshine, and wind speed meteorological resources; The energy prices include price information of cold, heat, electricity, hydrogen, ammonia, natural gas, and diesel.

3. The LLM-based integrated energy model building and solving system of claim 2, wherein, It also includes a semantic verification module for reviewing and forwarding the integrated energy system semantic configuration package, and forwarding it after passing the review, otherwise prompting user intervention or triggering an automatic correction mechanism; The semantic verification module reviews the integrated energy system semantic configuration package in the following ways: rule matching, boundary checking, and consistency checking.

4. The LLM-based integrated energy model building and solving system of claim 3, wherein, It also includes a splicing module for splicing user's natural language description of integrated energy planning requirements and all contents in the integrated energy knowledge base module to form a combined text.

5. The LLM-based integrated energy model building and solving system of claim 1, wherein, The integrated energy system scenario includes an energy form structure and an energy flow topology structure; The energy form structure is used to determine one or more of the five energy forms of cold, heat, electricity, hydrogen, and ammonia in the energy system scenario; The energy flow topology structure is used to determine the energy flow topology relationship of each energy form.

6. The LLM-based integrated energy model building and solving system of claim 1, wherein, The target and constraint definition dataset includes an upper limit of total investment, emission factor restrictions, energy abandonment constraints, and energy supply ratio constraints.

7. The LLM-based integrated energy model building and solving system of claim 1, wherein, The optimal configuration scheme and performance indicators include: energy configuration structure: installed capacity of each device; device operation strategy: time-sharing operation output of each device; economic evaluation results: capital gain rate, static payback period, dynamic payback period, and static total investment; Emission results: annual carbon dust, annual carbon dioxide, annual sulfur dioxide, annual nitrogen oxides, annual PM2.5, and annual equivalent trees planted.

Images