A method and device for automatic design of power converter based on multi-modal large language model system

Abstract

The present invention discloses an automatic design method for a power converter based on a multimodal large language model system, comprising the following steps: obtaining a design solution for a power converter and corresponding knowledge content, and constructing a corresponding retrieval model through a large language model framework; sending the obtained input text to the retrieval model, and the retrieval model calls the corresponding design solution process framework based on key information in the input text; visually outputting the target data-driven model and simulation results finally obtained to provide a corresponding reference for the user's subsequent power converter design. The present invention also provides an automatic design device for a power converter. The method provided by the present invention realizes the automatic design of a power converter through adaptive semantic understanding and multimodal processing.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric energy converters, and in particular relates to an automatic design method and device for a power converter based on a multi-modal large language model system. Background Art

[0002] With the rapid development of artificial intelligence technology, more and more industries are beginning to apply big language models to actual design and engineering practice. As a complex and highly specialized field, power converter design can significantly speed up the automated design process and improve design efficiency through the reasonable and effective use of big language models.

[0003] However, the use of large language models in current power converter design faces the following bottlenecks and challenges:

[0004] Insufficient domain expertise: Existing large language models are usually trained based on a broad general corpus and lack a deep understanding of the power electronics field. This makes it difficult for large language models to provide accurate technical advice and decision support when dealing with specific power converter designs.

[0005] Difficulty in processing unstructured data in the field of power electronics: Power converter design involves a large amount of unstructured data, such as time series operation waveforms, spectrum analysis, etc. Traditional large speech models have significant limitations in processing such data and cannot be effectively applied to scenarios that require deep data analysis and physical modeling.

[0006] Patent document CN109149943A designs a parameter optimization design method for LLC resonant converter based on cuckoo algorithm. The LLC resonant converter includes a square wave generator, a resonant network, a high-frequency transformer and a rectifier network connected in sequence. The method includes the following steps: S1. Using fundamental wave analysis method to obtain the equivalent circuit of LLC resonant converter circuit; S2. Determine the constraint conditions between the resonant inductance and the excitation inductance in the equivalent circuit; S3. Establish the fitness function of the cuckoo algorithm, use the cuckoo algorithm to determine the optimal set of resonant inductance and excitation inductance values, and determine the value of the resonant capacitor.

[0007] Patent CN117972524A discloses a method and system for rapid fault identification based on the timing characteristics of a power conversion module, which belongs to the field of power electronics technology and artificial intelligence, and includes a data acquisition unit, a data processing unit, a neural network model unit, and a fault classification and identification unit. The data acquisition unit is responsible for acquiring characteristic voltage and current signal data of the power conversion system under various working conditions; the data processing unit constructs a data set suitable for neural network input by extracting and analyzing the timing characteristics; the neural network model unit adopts Bi-GRU and Attention mechanisms, is trained by Adam optimizer and cross entropy loss function, and adjusts the model by semi-supervised learning; the fault classification and identification unit realizes the identification of faulty components in the power conversion system through the neural network model. The present invention can quickly and accurately identify the faulty components of the power conversion system, improve the system operation reliability, reduce the operation and maintenance costs, and has good feasibility and practical value. Summary of the invention

[0008] The purpose of the present invention is to provide a method and device for automatic design of power converters based on a multimodal large language model system. The method realizes the automatic design of power converters by applying the adaptive semantic understanding and multimodal processing capabilities of the multimodal large language model.

[0009] In order to achieve the first objective of the present invention, the following technical solution is provided: a power converter automatic design method based on a multi-modal large language model system, comprising the following steps:

[0010] Obtain the power converter design process and build the corresponding knowledge base, and build the corresponding retrieval model based on the large language model framework;

[0011] Sending the acquired input text to the retrieval model, wherein the retrieval model calls a corresponding design solution process framework based on key information in the input text, wherein the input text includes design requirements and specified fine-tuning data of a power converter, wherein the design requirements include one or more of technical objectives, simulation tests or model fine-tuning;

[0012] Based on the design solution process framework, the following workflow is executed according to the design requirements in the input text:

[0013] Aiming at the technical goal, referring to the design solution process framework, the data-driven model in the model library is used to model the target data and optimized in combination with the meta-heuristic algorithm to output circuit parameter values, modulation parameter values ​​and corresponding target data-driven models that meet the design requirements;

[0014] For simulation testing, the model is driven by the target data finally obtained, and simulation testing is performed in the pre-built simulation library to output the corresponding simulation results;

[0015] For model fine-tuning, a target data-driven model is fine-tuned based on the fine-tuning data specified by the user to output a prediction result of the fine-tuned target data-driven model;

[0016] The obtained circuit parameter values, modulation parameter values, simulation results and prediction results of the fine-tuned target data-driven model are visualized and output, and provide corresponding references for users' subsequent power converter designs based on professional knowledge in the power supply design field.

[0017] The present invention improves on the existing large language model by constructing a power converter design knowledge base and using the knowledge base to achieve retrieval enhancement generation, thereby improving the professionalism and practicality of the original large language model. At the same time, the large language model agent is integrated with the meta-heuristic algorithm, model library and simulation library into the design process, thereby greatly enhancing the flexibility and versatility of the large language model in coping with various tasks in power converter design.

[0018] Specifically, the large language model framework adopts GPT-3.5, GPT-4, Llama, Wenxin Yiyan, or iFlytek Spark.

[0019] Specifically, the retrieval model converts the input text into an embedding vector, and performs similarity matching with the power converter design solution process and the corresponding knowledge base based on the embedding vector to obtain the power converter design solution process with the highest similarity value as the design solution process framework.

[0020] Specifically, the technical objectives include performance indicators of each power converter, such as one or more of current stress, zero voltage switching range, zero current switching range, reactive power, efficiency, output voltage ripple or output current ripple.

[0021] Specifically, the model library includes a white box model, a black box model and a gray box model.

[0022] Specifically, the simulation repository includes Simulink, PLECS and PSIM for circuit simulation, and COMSOL, ANSYS and SIMSCALE for multi-physics field simulation.

[0023] Specifically, the fine-tuning data includes time series waveform data and table performance data.

[0024] In order to achieve the second purpose of the present invention, the following technical solution is provided: an automatic design device for a power converter, comprising a computer memory, a computer processor, and a computer program stored in the computer memory and executable on the computer processor, wherein the computer processor is used to execute the steps of the multi-modal generation scheme method suitable for power converter design.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The use of retrieval-augmented production technology enables the large language model to obtain domain knowledge from the provided power converter design knowledge base before generating text responses, thereby better responding to user needs. The hybrid framework consisting of a large language model agent, metaheuristic algorithm, model library, and simulation library enhances the flexibility and versatility of the large language model for various tasks in power converter design and seamlessly integrates it into the existing design workflow. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flowchart of a method for automatic design of a power converter based on a multi-modal large language model system provided in this embodiment;

[0028] Figure 2 A hybrid framework diagram of the multimodal large language model system provided for this embodiment. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] like Figure 1 As shown, a power converter automatic design method based on a multi-modal large language model system provided in this embodiment includes the following steps:

[0031] Obtain the power converter design process and build the corresponding knowledge base, and build the corresponding retrieval model based on the large language model framework;

[0032] Sending the acquired input text to the retrieval model, wherein the retrieval model calls a corresponding design solution process framework based on key information in the input text, wherein the input text includes design requirements and specified fine-tuning data of a power converter, wherein the design requirements include one or more of technical objectives, simulation tests or model fine-tuning;

[0033] Based on the design solution process framework, the following workflow is executed according to the design requirements in the input text:

[0034] Aiming at the technical goal, referring to the design solution process framework, the data-driven model in the model library is used to model the target data and optimized in combination with the meta-heuristic algorithm to output circuit parameter values, modulation parameter values ​​and corresponding target data-driven models that meet the design requirements;

[0035] For simulation testing, the model is driven by the target data finally obtained, and simulation testing is performed in the pre-built simulation library to output the corresponding simulation results;

[0036] For model fine-tuning, a target data-driven model is fine-tuned based on the fine-tuning data specified by the user to output a prediction result of the fine-tuned target data-driven model;

[0037] The obtained circuit parameter values, modulation parameter values, simulation results and prediction results of the fine-tuned target data-driven model are visualized and output, and provide corresponding references for users' subsequent power converter designs based on professional knowledge in the power supply design field.

[0038] Furthermore, in actual operation, the power converter design process is obtained and the corresponding knowledge base is constructed, and the corresponding retrieval model is constructed through the large language model framework. The power converter design includes the modulation strategy design and parameter design of the power converter.

[0039] The knowledge base includes academic papers and technical documents related to power converter topology, modulation strategy, and parameter design.

[0040] like Figure 2 As shown, the model library may include gray box models, black box models and white box models related to power converter design. A typical representative of gray box models is physical structure neural network. The simulation library may include circuit simulation related software such as MATLAB, PLECS, PSIM, and may also include multi-physics simulation related software such as COMSOL, ANSYS and SIMSCALE. The selected large language model framework may be GPT-3.5, GPT-4, Llama, Wenxin Yiyan, or iFlytek Spark.

[0041] The user can input text or upload data used to fine-tune the model in the model library according to their own purpose requirements. The text information includes the design requirements of the power converter, and the design requirements include one or more of technical goals, simulation tests or model fine-tuning;

[0042] like Figure 2 As shown, the input text obtained is converted into a vector through embedding so as to be compared with the relevant knowledge stored in the document vector database. Through similarity search, the information block most relevant to the input text is retrieved. Then the input text and information block obtained are input into the retrieval model. The large language retrieval model uses its natural language processing capabilities to perform power supply design based on key information in the text information, such as technical indicators, simulation tests on the design results, and model fine-tuning on the uploaded data, thereby calling the corresponding design solution process framework and starting the corresponding workflow.

[0043] In view of the technical goals, process 1 will be started to design using the model library. In view of the user's technical goals, such as inductor current stress, zero voltage switching range, zero current switching range, efficiency, reactive power, inductor current ripple, capacitor voltage ripple, etc., the data-driven model in the model library is used to build the data-driven model corresponding to the technical goals, and the meta-heuristic algorithm is combined to optimize the technical indicators, so as to output the circuit parameter value or modulation parameter value that meets the design requirements.

[0044] For simulation test, process 2 will be started for simulation verification. The start of process 2 must be after process 1. According to the text information input by the user and the circuit parameter value or modulation parameter value designed by process 1, the corresponding simulation software in the simulation library is started to build a corresponding simulation system for operation to output circuit voltage waveform, current waveform, thermal simulation cloud map, electromagnetic simulation cloud map and other simulation results.

[0045] For model fine-tuning, process 3, fine-tuning, will be initiated. The neurons of the data-driven model in the model library are fine-tuned according to the time series waveform data and table performance data uploaded by the user to reduce the difference between the model output and the actual results, so that the data-driven model can output more realistic prediction results.

[0046] Finally, the obtained optimization design results and simulation results are output to the user, including the optimal circuit parameter values, modulation parameter values, circuit waveforms corresponding to the optimization design goals under the optimal values, and simulation results obtained by calling related simulation software. These results are combined with the semantic understanding ability of the large language model for output, providing corresponding references for the user's subsequent power converter design.

[0047] This embodiment also provides a device for establishing a multimodal large language model system suitable for dual active bridge converters and buck converters, including a computer memory, a computer processor, and a computer program stored in the computer memory and executable on the computer processor, wherein the computer processor is used to execute the steps of the multimodal generation scheme method suitable for the design of dual active bridge converters and buck converters.

[0048] It should be understood by those skilled in the art that the embodiments of the present application can be provided as methods, systems, or computer program products. The solutions in the embodiments of the present application can be implemented in various computer languages, such as MATLAB, Python, C++, Java, etc. The simulation software used in the embodiments of the present application

[0049] Commonly used simulation software for various power converters can be used, such as MATLAB, PLECS, etc.

[0050] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0051] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0052] In summary, the method provided by the present invention can not only significantly reduce the required data set and computing resources, but also effectively quantify the uncertainty of the data and realize the transfer modeling of multiple topologies. The established model is not only highly accurate, but also has good interpretability, and has great engineering application value and broad promotion prospects.

[0053] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for automatic design of power converter based on multi-modal large language model system, characterized in that: The following steps are involved: Obtain the power converter design process and build the corresponding knowledge base, and build the corresponding retrieval model based on the large language model framework; Sending the acquired input text to the retrieval model, wherein the retrieval model calls a corresponding design solution process framework based on key information in the input text, wherein the input text includes design requirements and specified fine-tuning data of a power converter, wherein the design requirements include one or more of technical objectives, simulation tests or model fine-tuning; Based on the design solution process framework, the following workflow is executed according to the design requirements in the input text: Aiming at the technical goal, referring to the design solution process framework, the data-driven model in the model library is used to model the target data and optimized in combination with the meta-heuristic algorithm to output circuit parameter values, modulation parameter values ​​and corresponding target data-driven models that meet the design requirements; For simulation testing, the model is driven by the target data finally obtained, and simulation testing is performed in the pre-built simulation library to output the corresponding simulation results; For model fine-tuning, a target data-driven model is fine-tuned based on the fine-tuning data specified by the user to output a prediction result of the fine-tuned target data-driven model; The obtained circuit parameter values, modulation parameter values, simulation results and prediction results of the fine-tuned target data-driven model are visualized and output, and provide corresponding references for users' subsequent power converter designs based on professional knowledge in the power supply design field.

2. The power converter automatic design method based on multi-modal large language model system according to claim 1, characterized in that: The large language model framework uses GPT-3.5, GPT-4, Llama, Wenxin Yiyan, or iFlytek Spark.

3. The power converter automatic design method based on multi-modal large language model system according to claim 1, characterized in that: The retrieval model converts the input text into an embedding vector, and performs similarity matching with the power converter design solution process and the corresponding knowledge base based on the embedding vector to obtain the power converter design solution process with the highest similarity value as the design solution process framework.

4. The power converter automatic design method based on multi-modal large language model system according to claim 1, characterized in that: The technical objectives include performance indicators of each power converter.

5. The power converter automatic design method based on multi-modal large language model system according to claim 1, characterized in that: The model library includes white box models, black box models and grey box models.

6. The power converter automatic design method based on multi-modal large language model system according to claim 1, characterized in that: The simulation library includes Simulink, PLECS and PSIM for circuit simulation, and COMSOL, ANSYS and SIMSCALE for multi-physics field simulation.

7. The power converter automatic design method based on multi-modal large language model system according to claim 1, characterized in that: The fine-tuning data includes time series waveform data and tabular performance data.

8. A power converter automatic design device, comprising a computer memory, a computer processor, and a computer program stored in the computer memory and executable on the computer processor, characterized in that: The computer processor is used to execute the steps of the power converter automatic design method based on a multi-modal large language model system as described in any one of claims 1-7.

Images