Large model and knowledge graph layered collaborative reasoning wind power project report generation method, system and device and storage medium

Abstract

The invention provides a large model and knowledge graph layered collaborative reasoning wind power project report generation method, system and device, and a storage medium. The method comprises the following steps: S1, constructing a knowledge graph; s2, reasoning the knowledge graph; s3, generating a hierarchical embedding vector according to a reasoning result; s4, embedding the embedding vector into the large model; s5, dynamically adjusting the hierarchy and the weight of the large model; s6, generating an initial wind power project report; s7, optimizing a feedback mechanism and reinforcement learning; and S8, generating a final wind power project report. Through a layered embedding mechanism, it is ensured that the large model can accurately capture key information in a reasoning result, and a generated report better conforms to domain logic and professional judgment; through embedding of a basic layer and a high-level layer, appropriate reasoning information can be quoted in different stages of report generation, and the continuity and logic consistency of the report are ensured; the large model can dynamically adjust the generation strategy according to the report, and the generation strategy not only comprises technical suggestions in the field, but also can be adjusted among different hierarchies.

Description

Technical Field

[0001] The present invention relates to the fields of natural language processing and intelligent management of photovoltaic projects, and specifically relates to a method, system, device and storage medium for generating a wind power project report through hierarchical collaborative reasoning of a large model and a knowledge graph. Background Art

[0002] Wind power project reports play an important role in aspects such as the construction planning of wind farms, equipment selection, construction progress prediction, and risk assessment. Traditional generation methods rely on manual writing by experts, which is inefficient and easily affected by human factors. With the development of artificial intelligence technology, report generation methods driven by large models have made certain progress. However, due to the lack of professional domain knowledge in the general corpus used for large model pre-training, when used for downstream tasks in specific fields, they cannot generate high-quality reports that conform to professional logic. Therefore, how to effectively combine the structured knowledge in the professional domain with the natural language content generated by the large model to generate a wind power project report with high logic and professionalism has become an urgent technical problem to be solved.

[0003] In the prior art, in the multi-task learning method, interference between different tasks may lead to performance degradation, and its training process is complex, requiring reasonable design of tasks and loss functions. The policy optimization method requires high professional knowledge and engineering implementation, and is too dependent on rules, restricting the flexibility and adaptability of the model. The external knowledge base enhancement method requires complex knowledge retrieval and integration mechanisms, increasing the complexity of the system, and the quality and accuracy of the external knowledge base directly affect the inference effect of the model. The implementation process of the adaptive inference method is relatively complex, requiring an accurate evaluation mechanism to judge the complexity of the input, and usually at the cost of increasing the inference time. Summary of the Invention

[0004] The first object of the present invention is to provide a method for generating a wind power project report through hierarchical collaborative reasoning of a large model and a knowledge graph in view of the above-mentioned problems.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for generating a wind power project report through hierarchical collaborative reasoning of a large model and a knowledge graph, comprising the following steps:

[0007] S1: Knowledge graph construction;

[0008] S2: Knowledge graph reasoning;

[0009] S3: Generating a hierarchical embedding vector from the inference result;

[0010] S4: Embedding the embedding vector into the large model;

[0011] S5: Dynamically adjust the levels and weights of the large model;

[0012] S6: Generate an initial wind power project report;

[0013] S7: Feedback mechanism and reinforcement learning optimization;

[0014] S8: Generate a final wind power project report that meets requirements in multiple dimensions.

[0015] While adopting the above technical solutions, the present invention can also adopt or combine the following technical solutions:

[0016] As a preferred technical solution of the present invention: In step S1, the knowledge graph includes nodes of entities in the fields of wind power equipment, construction conditions, environmental factors, and geological data, as well as the relationships between the nodes.

[0017] As a preferred technical solution of the present invention: In step S2, the inference results include equipment selection suggestions, construction progress prediction, and risk assessment.

[0018] As a preferred technical solution of the present invention: In step S3, the hierarchical embedding includes basic layer embedding and advanced layer embedding.

[0019] As a preferred technical solution of the present invention: Step S7 specifically means that the initial wind power project report is reviewed through feedback from experts or historical data, and the weights of the hierarchical embedding are automatically adjusted and the content of the generation strategy optimization report is generated.

[0020] As a preferred technical solution of the present invention: In step S8, the final wind power project report includes analyses in multiple dimensions, such as site selection feasibility, equipment selection, construction plan, economic benefit assessment, policy support analysis, equipment selection suggestions, construction progress prediction, and risk assessment.

[0021] The second object of the present invention is to provide a wind power project report generation system for hierarchical collaborative inference of a large model and a knowledge graph.

[0022] To achieve this, the above object of the present invention is realized through the following technical solutions:

[0023] A knowledge graph construction module, which is used to construct a knowledge graph reflecting professional knowledge in the wind power field;

[0024] A knowledge graph inference module, which is used to enable the knowledge graph to perform logical inference based on the input wind farm project data and generate inference results by defining the inference rules of domain experts;

[0025] Hierarchical embedding module, which is used to output the inference results in a structured form, including information such as node relationships and inference chains, and generate two layers of embedding vectors: basic layer embedding and advanced layer embedding;

[0026] Large model embedding module, which is used to embed the basic layer embedding and advanced layer embedding vectors into the large model;

[0027] Large model dynamic adjustment module, which is used to dynamically combine the vector information of the basic layer and the advanced layer embedding during the generation process to ensure that the report generation can refer to specific data and logical inference chains at the same time. In different report parts, the system will automatically adjust the reference and weight allocation of different levels of embedding;

[0028] Initial wind power project report generation module, which is used to generate an initial wind power project report;

[0029] Feedback and optimization module, which is used to conduct reviews through expert or historical data feedback, automatically adjust the hierarchical embedding weights and generation strategies, optimize the report content generated by the large model, and gradually improve the logic and professionalism of the report through multiple rounds of optimization of reinforcement learning;

[0030] Final wind power project report generation module, which is used to generate a final wind power project report that meets requirements in multiple dimensions.

[0031] The third object of the present invention is to provide an electronic device.

[0032] For this reason, the above object of the present invention is achieved by the following technical solutions:

[0033] An electronic device includes a memory and a processor. An executable program is stored in the memory, and the processor is configured to run the executable program to execute the steps of the wind power project report generation method for hierarchical collaborative inference of a large model and a knowledge graph as described above.

[0034] The fourth object of the present invention is to provide a non-volatile storage medium.

[0035] For this reason, the above object of the present invention is achieved by the following technical solutions:

[0036] A non-volatile storage medium stores an executable program. When the executable program is executed by a processor, it realizes the steps of the wind power project report generation method for hierarchical collaborative inference of a large model and a knowledge graph as described above.

[0037] The present invention provides a wind power project report generation method, system, device and storage medium for hierarchical collaborative reasoning between a large model and a knowledge graph, which has the following beneficial effects: the present invention ensures that the large model can accurately capture the key information in the knowledge graph reasoning results through a hierarchical embedding mechanism, and the generated report content is more in line with the domain logic and professional judgment; the present invention embeds the basic layer and the advanced layer, and the system can reference appropriate reasoning information at different stages of report generation to ensure the coherence and logical consistency of the report; the large model can dynamically adjust the generation strategy according to the report content, and the generated report not only includes technical suggestions in the field, but also can be adaptively adjusted between different levels (technology, management). BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of the wind power project report generation method based on hierarchical collaborative reasoning of a large model and a knowledge graph provided by the present invention. DETAILED DESCRIPTION

[0039] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0040] like Figure 1 As shown, a method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph includes the following steps:

[0041] S1: Knowledge graph construction;

[0042] Node and relationship definition: First, a knowledge graph that reflects professional knowledge in the wind power field is constructed, which includes nodes of entities in the fields of wind power equipment, construction conditions, environmental factors, geological data, etc., as well as relationships between nodes, such as "equipment is suitable for a certain environment", "geological conditions affect construction progress", etc.

[0043] S2: Knowledge graph reasoning;

[0044] Reasoning rule setting: By defining the reasoning rules of domain experts, the knowledge graph can perform logical reasoning based on the input wind farm project data (such as construction environment, equipment parameters, etc.) and generate reasoning results. These results include equipment selection recommendations, construction progress forecasts, risk assessments, etc., which can provide strong structured support for report generation.

[0045] Reasoning result output: The knowledge graph reasoning results are output in a structured form, including information such as node relationships and reasoning chains, and two layers of embedding vectors are generated: basic layer embedding and advanced layer embedding. The basic layer contains basic node relationships, such as the matching of equipment and environment; the advanced layer embedding contains more complex logical chains, such as the logical basis for equipment selection and risk assessment.

[0046] S3: Generate hierarchical embedding vectors from the inference results;

[0047] Base layer embedding: Convert the node and relationship information in the knowledge graph reasoning result (such as "Device A is applicable to Environment B") into embedding vectors and embed them into the large model. This layer of embedding is mainly used to directly reference specific data during the generation process, such as device parameters and applicable conditions.

[0048] Advanced layer embedding: The advanced layer embedding reflects the logical path and professional judgment of the reasoning chain, such as "Due to geological condition X, it is recommended to use Device Y, and the construction progress Z will be affected by the climate". The embedding vectors of the advanced layer are used to ensure logical coherence and the reasoning rationality of the report content during the generation process.

[0049] S4: Embed the embedding vectors into the large model;

[0050] Context alignment and dynamic embedding: When generating a wind power project report, the large model dynamically introduces hierarchical embedding information through a context alignment mechanism according to different parts of the generated content (such as equipment selection, construction progress analysis, etc.). Specifically, when generating the equipment selection paragraph, the model preferentially uses the equipment-related data in the base layer embedding, and relies on the reasoning chain in the advanced layer embedding to generate predictions in the construction progress analysis section.

[0051] Optimization of the attention mechanism: During the generation process, the large model uses a multi-head attention mechanism to enhance the capture of the knowledge graph reasoning result. By assigning higher attention weights to the key information in the reasoning chain, it ensures that the report content accurately reflects the logic of the reasoning result.

[0052] S5: The large model dynamically adjusts the levels and weights;

[0053] Collaborative processing of embedding levels: The large model dynamically combines the vector information of the base layer and the advanced layer embedding during the generation process to ensure that the report generation can refer to specific data and the logical reasoning chain simultaneously. In different parts of the report, the system automatically adjusts the reference and weight allocation of different levels of embedding.

[0054] S6: Generate an initial wind power project report;

[0055] S7: Feedback mechanism and reinforcement learning optimization;

[0056] The generated wind power project report is reviewed through expert feedback or historical data feedback. Based on the feedback, the system automatically adjusts the hierarchical embedding weights and generation strategies to optimize the report content generated by the large model. Through multiple rounds of optimization by reinforcement learning, the logic and professionalism of the report are gradually improved.

[0057] S8: Generate a final wind power project report that meets requirements in multiple dimensions.

[0058] The finally generated wind power project report includes analyses from multiple dimensions, such as equipment selection suggestions, construction schedule prediction, risk assessment, etc. These reports not only cover the natural language content generated by the large model but also include structured data and charts generated by knowledge graph reasoning, making the report content rich and professional.

[0059] Specifically, the method for generating a wind power project report through hierarchical collaborative reasoning of the above-mentioned large model and knowledge graph is implemented as follows:

[0060] S1: The knowledge graph constructs multiple core dimensions of the wind power project, covering nodes such as wind farm site selection, equipment configuration, project budget, technical feasibility, etc., and defines the relationships between these nodes. For example, inference chains such as "Equipment A is suitable for specific wind speed conditions" and "Region B is suitable for building a wind farm".

[0061] S2: After the basic data of the project (such as geographical location, meteorological conditions, investment budget, etc.) is input into the system, the knowledge graph inference engine generates relevant inference results based on these data, such as: feasibility analysis of project site selection (based on geographical conditions, meteorological data, etc.), equipment selection and configuration suggestions (according to wind speed, budget, etc.), prediction of project economic benefits (combining power generation estimation and market electricity price prediction), and the inference results are divided into basic layer embeddings (such as basic data like project location, wind speed, budget, etc.) and advanced layer embeddings (such as site selection feasibility, logical chains behind equipment selection, economic benefit prediction, etc.).

[0062] S3: When the large model generates a project proposal, it first extracts the basic information of the project, such as geographical location, meteorological conditions, equipment configuration, etc., from the basic layer embedding to generate a basic description. For example: "This project is located in the coastal area, with an average annual wind speed of 8.5 m / s, which is suitable for building a wind farm."

[0063] Then the large model generates more detailed project proposal content based on the inference chains in the advanced layer embedding. For example, based on the inference chain "Due to the high wind speed and sufficient project budget, it is recommended to select wind turbine model Y, which can ensure the power generation efficiency and cost-effectiveness of the project.", suggestions for equipment recommendation and budget analysis are generated.

[0064] At the same time, when generating the economic benefit assessment part of the project, the model combines the economic inference chains in the knowledge graph (such as power generation prediction and electricity price trend) to generate suggestions on the economic return of the project: "It is estimated that the annual power generation is XX GWh. Based on the current market electricity price, the project can recover the cost and generate continuous income within X years."

[0065] S4 - S6: Project Site Selection Part: Based on the reasoning results of the knowledge graph, the large model extracts geographical information (such as wind speed, geological conditions) from the basic layer and uses the high - level layer embedding to generate a feasibility analysis for site selection: "Since the project is located in a high - wind - speed coastal area with firm geology, it has excellent construction conditions."

[0066] Equipment Selection and Technical Solution Part: The model extracts equipment model and technical parameter information from the basic layer embedding and uses the high - level layer embedding to generate the detailed logic and technical solutions for equipment recommendation: "According to the wind speed characteristics of the wind farm, it is recommended to use equipment of model Y with a rated power of XX MW, which is suitable for efficient operation under such wind speed conditions."

[0067] Economic Benefit Analysis Part: Based on the reasoning chain of economic benefit prediction, the large model extracts power generation and electricity price prediction data from the basic layer embedding and combines with the high - level reasoning logic to generate a prediction report on economic returns: "Based on the prediction of market electricity price trends, it is expected that this project will break even in X years."

[0068] S7: After the initial generated proposal is reviewed by experts in the wind power field, the system finds that the economic return analysis in some areas is slightly conservative and fails to fully consider local policy preferences. The system conducts a review through expert or historical data feedback, adjusts the weight of policy preferences in the high - level layer embedding, strengthens this reasoning result, and the content of the final generated proposal is more comprehensive, adding an analysis of local policy support.

[0069] S8: The finally generated wind power project proposal contains analysis content in multiple dimensions, including project site selection feasibility, equipment selection, construction plan, economic benefit evaluation, and policy support analysis. The report structure is clear and the content is comprehensive, ensuring that all parties involved in the project (such as investors, technical teams, government departments) can obtain complete project proposal information.

[0070] A wind power project report generation system with hierarchical collaborative reasoning of a large model and a knowledge graph. The system includes the following modules:

[0071] Knowledge Graph Construction Module, which is used to construct a knowledge graph reflecting the professional knowledge in the wind power field;

[0072] Knowledge Graph Reasoning Module, which is used to define the reasoning rules of domain experts, enabling the knowledge graph to perform logical reasoning based on the input wind farm project data and generate reasoning results;

[0073] Hierarchical Embedding Module, which is used to output the reasoning results in a structured form, including information such as node relationships and reasoning chains, and generate two - layer embedding vectors: basic layer embedding and high - level layer embedding;

[0074] The large model embedding module is used to embed the basic layer embedding and the advanced layer embedding vectors into the large model;

[0075] The large model dynamic adjustment module is used to dynamically combine the vector information of the basic layer and the advanced layer embedding during the generation process, ensuring that the report generation can refer to specific data and logical reasoning chains simultaneously. In different report parts, the system will automatically adjust the reference and weight allocation of different-level embeddings;

[0076] The initial wind power project report generation module is used to generate the initial wind power project report;

[0077] The feedback and optimization module is used to conduct reviews through expert or historical data feedback, automatically adjust the hierarchical embedding weights and generation strategies, optimize the report content generated by the large model, and gradually improve the logic and professionalism of the report through multiple rounds of optimization of reinforcement learning;

[0078] The final wind power project report generation module is used to generate the final wind power project report that meets requirements in multiple dimensions.

[0079] The present invention also provides an electronic device, including a processor and a memory for storing processor-executable instructions. Among them, when the processor is set to execute the executable instructions, the method steps of generating a wind power project report through hierarchical collaborative reasoning of the large model and the knowledge graph as described above are implemented.

[0080] The present invention also provides a non-volatile storage medium, in which an executable program is stored. When the executable program is executed by a processor, the method steps of generating a wind power project report through hierarchical collaborative reasoning of the large model and the knowledge graph as described above are implemented.

[0081] The above specific embodiments are used to explain the present invention, and are only the preferred embodiments of the present invention, rather than limiting the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and scope of the claims of the present invention fall within the protection scope of the present invention.

Claims

1. A wind power project report generation method based on hierarchical collaborative reasoning of a large model and a knowledge graph, characterized in that: The steps include: S1: Knowledge graph construction; S2: Knowledge graph reasoning; S3: Generate hierarchical embedding vectors from the inference results; S4: embedding the embedding vector into the large model; S5: Large models dynamically adjust levels and weights; S6: Generate an initial wind power project report; S7: Feedback mechanism and reinforcement learning optimization; S8: Generate a final wind power project report that meets the requirements in multiple dimensions.

2. The method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph according to claim 1 is characterized in that: In step S1, the knowledge graph includes nodes of entities in the fields of wind power equipment, construction conditions, environmental factors, and geological data, as well as the relationships between the nodes.

3. The method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph according to claim 1 is characterized in that: In step S2, the reasoning results include equipment selection recommendations, construction progress prediction, and risk assessment.

4. The method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph according to claim 1 is characterized in that: In step S3, the hierarchical embedding includes base layer embedding and high-level layer embedding.

5. The method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph according to claim 1 is characterized in that: Step S7 specifically involves reviewing the initial wind power project report through experts or historical data feedback, automatically adjusting the layered embedding weights and generating strategies to optimize the report content.

6. The method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph according to claim 1 is characterized in that: In step S8, the final wind power project report includes analysis in multiple dimensions, such as site feasibility, equipment selection, construction plan, economic benefit evaluation, policy support analysis, equipment selection recommendations, construction progress forecast, and risk assessment.

7. A wind power project report generation system based on hierarchical collaborative reasoning of a large model and knowledge graph, characterized by: The system includes the following modules: A knowledge graph construction module, wherein the knowledge graph construction module is used to construct a knowledge graph reflecting professional knowledge in the field of wind power; A knowledge graph reasoning module, which is used to define the reasoning rules of domain experts so that the knowledge graph can perform logical reasoning based on the input wind farm project data and generate reasoning results; A hierarchical embedding module, which is used to output the reasoning results in a structured form, including information such as node relationships and reasoning chains, and generate two layers of embedding vectors: a basic layer embedding and a high-level layer embedding; A large model embedding module, the large model embedding module is used to embed the base layer embedding and the high-level layer embedding vectors into the large model; Large model dynamic adjustment module, which is used to dynamically combine the vector information embedded in the basic layer and the advanced layer during the generation process to ensure that the report can refer to both specific data and logical reasoning chains at the same time. In different report parts, the system will automatically adjust the references and weight distribution of embeddings at different levels; An initial wind power project report generating module, wherein the initial wind power project report generating module is used to generate an initial wind power project report; Feedback and optimization module, which is used to review through expert or historical data feedback, automatically adjust the layered embedding weights and generation strategies, optimize the report content generated by the large model, and gradually improve the logic and professionalism of the report through multiple rounds of optimization of reinforcement learning; The final wind power project report generation module is used to generate a final wind power project report that meets requirements in multiple dimensions.

8. An electronic device, comprising a memory and a processor, characterized in that: An executable program is stored in the memory, and the processor is configured to run the executable program to execute the steps of a method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph as described in any one of claims 1 to 6.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores an executable program, and when the executable program is executed by the processor, the steps of the method for generating a wind power project report by hierarchical collaborative reasoning of a large model and a knowledge graph as described in any one of claims 1 to 6 are implemented.

Images