Power equipment fault diagnosis method and system, electronic equipment and readable storage medium
By optimizing prompt words through differential evolution algorithm and multi-model collaboration mechanism, and fine-tuning pre-trained model by combining instruction learning method, the problem of low efficiency of traditional power equipment fault diagnosis method is solved, and efficient multi-scenario adaptability and accurate fault diagnosis are achieved.
Patent Information
- Application Number
- CN202510931455.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-07
- Publication Date
- 2025-11-28
AI Technical Summary
Traditional power equipment fault diagnosis methods rely on human experience, which is inefficient and difficult to adapt to the comprehensive judgment of various types of equipment and faults. Furthermore, expert systems are unable to cope with complex and new fault scenarios, and the variable operating environment of equipment leads to large differences in the manifestation of similar faults.
By combining differential evolution algorithm and multi-model collaboration mechanism to optimize prompt words, a multi-scenario fault diagnosis dataset is constructed, and the pre-trained model is fine-tuned through instruction learning method to achieve fault diagnosis of power equipment.
It reduces the false positive and false negative rates, adapts the model to different working environments and device types, reduces dependence on specific scenario data, improves system performance, avoids performance bottlenecks, and reduces reliance on human experts.
Smart Images
Figure CN121027642A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of fault diagnosis, in particular to a power equipment fault diagnosis method and system, an electronic device and a readable storage medium. BACKGROUND
[0002] There are various types of power equipment, including transformers, circuit breakers, cables, mutual inductors, etc. The working environment is complex and variable, the fault types are various, and the correlation is strong. The traditional power equipment fault diagnosis method mainly relies on manual experience judgment, single equipment monitoring system or expert system, and faces the following problems:
[0003] Manual judgment depends on expert experience, is inefficient and difficult to inherit; a single fault detection system is difficult to adapt to the comprehensive judgment of multiple types of equipment and faults; a rule-based expert system is difficult to cope with complex and new fault scenarios; the equipment operating environment is variable, and the same type of fault has different manifestations in different scenarios.
[0004] Therefore, there is an urgent need for a new power equipment fault diagnosis scheme to solve the above problems. SUMMARY
[0005] Therefore, the purpose of the present application is to provide a power equipment fault diagnosis method and system, an electronic device and a readable storage medium to solve the above technical problems.
[0006] To achieve the above purpose, the first aspect of the present application provides a power equipment fault diagnosis method, comprising:
[0007] Optimizing the prompt words required for the power equipment fault diagnosis task by combining the differential evolution algorithm and the multi-model collaboration mechanism;
[0008] Based on the optimized prompt words, combining the historical operation data of the power equipment, a fault diagnosis data set covering multiple scenarios is constructed;
[0009] Based on the fault diagnosis data set, the power equipment fault diagnosis pre-training model is fine-tuned using the instruction learning method;
[0010] The fine-tuned power equipment fault diagnosis pre-training model is used for fault diagnosis of the power equipment.
[0011] In one of the embodiments, the optimization of the prompt words required for the power equipment fault diagnosis task by combining the differential evolution algorithm and the multi-model collaboration mechanism comprises:
[0012] Generating a plurality of initial prompt words using a prompt word generator;
[0013] Performing semantic variation on the plurality of initial prompt words using a prompt word variator to generate variant prompt words;
[0014] cross the variation prompt word with the initial prompt word to generate a trial prompt word;
[0015] evaluate the quality of the trial prompt word by using a prompt word evaluator, and replace the corresponding initial prompt word with the trial prompt word with a higher score.
[0016] In one embodiment, the semantic variation of the plurality of initial prompt words by the prompt word variation device to generate a variation prompt word comprises:
[0017] The prompt word variation device selects one of the initial prompt words as a reference prompt word, and on the basis of the reference prompt word, integrates the advantages of the initial prompt words other than the reference prompt word to generate the variation prompt word.
[0018] In one embodiment, the cross of the variation prompt word with the initial prompt word to generate a trial prompt word comprises:
[0019] The variation prompt word and the initial prompt word are fused according to a cross probability, the structure of the initial prompt word is retained, and the innovation points of the variation prompt word are integrated to generate the trial prompt word; wherein the cross probability determines the proportion of the content retained from the initial prompt word.
[0020] In one embodiment, the fault diagnosis data set includes instructions and responses, the instructions include basic instructions, combined instructions, scene adaptive instructions and decision support instructions; wherein the basic instructions are basic detection instructions for a single device and a single fault type; the combined instructions are instructions for integrating multiple data sources or analyzing multiple fault possibilities; the scene adaptive instructions are comprehensive analysis instructions that introduce working environment factors; the decision support instructions are advanced instructions for fault handling and prevention.
[0021] In one embodiment, the fine-tuning of the power equipment fault diagnosis pre-training model by using an instruction learning method based on the fault diagnosis data set comprises:
[0022] Based on the fault diagnosis data set, the optimized prompt word and the historical operation data of the power equipment are input, the target response is output, and the parameters of the power equipment fault diagnosis pre-training model are continuously adjusted by minimizing the difference between the generated result and the target response, so as to fine-tune the power equipment fault diagnosis pre-training model; wherein in the fine-tuning process, the power equipment fault diagnosis pre-training model extracts the key information of the task from the optimized prompt word, and establishes an effective mapping relationship from the instruction to the response.
[0023] In one embodiment, after fine-tuning the pre-trained power equipment fault diagnosis model using the instruction learning method based on the fault diagnosis dataset; before performing fault diagnosis on the power equipment using the fine-tuned pre-trained power equipment fault diagnosis model, the method further includes:
[0024] The performance of the fine-tuned pre-trained model for fault diagnosis of power equipment is evaluated.
[0025] Based on the same inventive concept, a second aspect of this application provides a power equipment fault diagnosis system, which includes:
[0026] The prompt word optimization module is used to optimize the prompt words required for power equipment fault diagnosis tasks by combining differential evolution algorithm and multi-model collaboration mechanism;
[0027] The dataset construction module is used to build a fault diagnosis dataset covering multiple scenarios based on optimized prompt words and combined with historical operating data of power equipment;
[0028] The model fine-tuning module is used to fine-tune the pre-trained model for power equipment fault diagnosis based on the fault diagnosis dataset using the instruction learning method.
[0029] The fault diagnosis module is used to diagnose faults in power equipment using a finely tuned pre-trained model for power equipment fault diagnosis.
[0030] In one embodiment, the prompt word optimization module includes:
[0031] The initial prompt word generation submodule is used to generate multiple initial prompt words using a prompt word generator.
[0032] The variant prompt word generation submodule is used to perform semantic mutations on multiple initial prompt words using a prompt word mutant to generate variant prompt words;
[0033] The test prompt word generation submodule is used to perform semantic cross-referencing between the variant prompt word and the initial prompt word to generate test prompt words;
[0034] The quality assessment submodule is used to evaluate the quality of the test prompts using a prompt word evaluator and replace the corresponding initial prompts with the test prompts that have a higher score.
[0035] In one embodiment, the step of using a cue word mutant to semantically mutate multiple initial cue words to generate mutated cue words includes:
[0036] The prompt word mutant selects one of the initial prompt words as a reference prompt word, and on the basis of the reference prompt word, incorporates the advantages of the initial prompt words other than the reference prompt word to generate the mutant prompt word.
[0037] In one embodiment, the step of semantically crossing the variant cue word with the initial cue word to generate a test cue word includes:
[0038] The mutated prompt word is fused with the initial prompt word according to the cross-probability, retaining the structure of the initial prompt word and incorporating the innovative points of the mutated prompt word to generate the experimental prompt word; wherein, the cross-probability determines the proportion of content retained from the initial prompt word.
[0039] In one embodiment, the fault diagnosis dataset includes instructions and responses. The instructions include basic instructions, combined instructions, scenario-adaptive instructions, and decision support instructions. The basic instructions are basic detection instructions for a single device and a single fault type. The combined instructions are instructions that integrate multiple data sources or analyze multiple fault possibilities. The scenario-adaptive instructions are comprehensive analysis instructions that incorporate working environment factors. The decision support instructions are advanced instructions for fault handling and prevention.
[0040] In one embodiment, the step of fine-tuning the pre-trained power equipment fault diagnosis model using an instruction learning method based on the fault diagnosis dataset includes:
[0041] Based on the fault diagnosis dataset, the optimized prompt words and historical operating data of the power equipment are used as inputs, and the target response is used as the output. By minimizing the difference between the generated result and the target response, the parameters of the power equipment fault diagnosis pre-training model are continuously adjusted to achieve fine-tuning of the power equipment fault diagnosis pre-training model. In the fine-tuning process, the power equipment fault diagnosis pre-training model extracts key information of the task from the optimized prompt words and establishes an effective mapping relationship from instructions to responses.
[0042] In one embodiment, it further includes:
[0043] The performance evaluation module is used to evaluate the performance of the fine-tuned pre-trained model for fault diagnosis of power equipment.
[0044] Based on the same inventive concept, a third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the power equipment fault diagnosis method described in the first aspect above.
[0045] Based on the same inventive concept, a fourth aspect of this application provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the power equipment fault diagnosis method described in the first aspect.
[0046] As can be seen from the above, this application guides the model to focus on key information through optimized prompt words, thereby reducing the false positive and false negative rates. Through multi-scenario instruction learning and knowledge transfer mechanisms, the model can adapt to different working environments and device types, reducing dependence on specific scenario data. Simultaneously, it innovatively combines differential evolution algorithms with a multi-model collaboration mechanism to achieve systematic optimization of high-quality prompt words, solving the problem that traditional numerical optimization algorithms cannot be directly applied to text optimization. Furthermore, it integrates expert knowledge into the model through instruction learning, reducing reliance on human experts, especially during off-peak hours and in areas with scarce expert resources. Moreover, through model collaboration and differential evolution mechanisms, it achieves continuous improvement in system performance, avoiding performance bottlenecks. Attached Figure Description
[0047] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0048] Figure 1 A flowchart of a power equipment fault diagnosis method provided in an embodiment of this application;
[0049] Figure 2 A schematic diagram of the architecture of a power equipment fault diagnosis method provided in another embodiment of this application;
[0050] Figure 3 A flowchart of a power equipment fault diagnosis method provided in another embodiment of this application;
[0051] Figure 4 A flowchart of a power equipment fault diagnosis method provided in another embodiment of this application;
[0052] Figure 5 A schematic diagram of a power equipment fault diagnosis system provided in another embodiment of this application;
[0053] Figure 6 This is a schematic diagram of an electronic device according to another embodiment of this application. Detailed Implementation
[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.
[0055] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0056] In recent years, generative large models have begun to be applied in the field of industrial fault detection due to their powerful contextual understanding and knowledge integration capabilities. Instruction learning, as a crucial technique for fine-tuning large models, enables models to learn the ability to execute natural language instructions, allowing them to provide customized outputs based on user needs. However, effectively applying instruction learning techniques to fine-tune pre-trained large models to adapt to the actual business scenarios of power equipment fault diagnosis, especially how to construct high-quality "instruction-response" data pairs, still faces significant challenges.
[0057] Based on this, this application provides a power equipment fault diagnosis solution to solve the above problems.
[0058] Reference Figure 1 and 2 As shown, one embodiment of this application provides a method for diagnosing faults in power equipment, which includes the following steps:
[0059] Step S10: Optimize the prompt words required for power equipment fault diagnosis tasks by combining differential evolution algorithm and multi-model collaboration mechanism;
[0060] Step S20: Based on the optimized prompt words and combined with the historical operating data of power equipment, construct a fault diagnosis dataset covering multiple scenarios;
[0061] Step S30: Based on the fault diagnosis dataset, fine-tune the pre-trained model for power equipment fault diagnosis using the instruction learning method;
[0062] Step S40: Use the finely tuned power equipment fault diagnosis pre-trained model to diagnose power equipment faults.
[0063] This application guides the model to focus on key information through optimized prompts, thereby reducing false positive and false negative rates. Through multi-scenario instruction learning and knowledge transfer mechanisms, the model can adapt to different working environments and device types, reducing dependence on specific scenario data. Simultaneously, it innovatively combines differential evolution algorithms with a multi-model collaboration mechanism to achieve systematic optimization of high-quality prompts, solving the problem that traditional numerical optimization algorithms cannot be directly applied to text optimization. Furthermore, expert knowledge is integrated into the model through instruction learning, reducing reliance on human experts, especially during off-peak hours and in areas with scarce expert resources. Moreover, through model collaboration and differential evolution mechanisms, continuous improvement in system performance is achieved, avoiding performance bottlenecks.
[0064] Differential Evolution (DE) is a population-based stochastic optimization algorithm that searches for the global optimum by performing differential mutation, crossover, and selection operations on individual vectors. Specifically, the algorithm first randomly initializes a set of candidate solutions in the parameter space. Then, in each generation, for each target individual, three different individuals are randomly selected from the population. The difference vector between these three individuals is calculated, multiplied by a scaling factor, and added to the baseline vector to generate a mutation vector. Next, a crossover strategy is used to combine the mutation vector with the target vector to generate an experimental vector. Finally, based on the objective function value, a greedy selection is performed between the experimental and target vectors, retaining the better one for the next generation. Due to its simple implementation, few parameters, and low requirements for the continuity or differentiability of the objective function, DE has been widely used in engineering optimization, machine learning model parameter tuning, and other fields.
[0065] Differential evolution algorithms were originally designed for numerical optimization, with population initialization, mutation, crossover, and selection operations all targeting vector data. However, natural language has a complex semantic structure and, unlike vector data, cannot be directly manipulated numerically. To address this challenge, this application defines "semantic mutation" and "semantic crossover" of prompt words as text-content-based modification operations to simulate the vector-to-vector operators in differential evolution. Specifically, Model 1 (the prompt word generator) generates the initial prompt word; Model 2 (the prompt word mutant) generates mutated prompt words based on the initial prompt word; and Model 3 (the prompt word evaluator) evaluates the quality of the prompt words.
[0066] Reference Figure 3 As shown, in one embodiment, step S10, which optimizes the prompts required for the power equipment fault diagnosis task by combining differential evolution algorithm and multi-model cooperation mechanism, includes:
[0067] Step S11: Generate multiple initial prompts using a prompt word generator, such as... Figure 3The initial population of prompt words in the text.
[0068] Specifically, based on information in the database, such as equipment type (transformer), fault characteristics (overheating), data description (temperature data), and analysis requirements (determining the cause of the fault and making suggestions), Model 1 extracts key components and generates N initial candidate prompts, each containing these key components. Example initial prompt: "You are a power equipment fault diagnosis expert. Please analyze the following transformer temperature data, determine if an overheating fault exists, and provide possible causes and handling suggestions. The data is as follows:..."
[0069] Step S12: Use a cue word mutant to semantically mutate multiple initial cue words to generate mutated cue words, such as... Figure 3 The prompt word offspring population. Optionally, the prompt word mutant selects one of the initial prompt words as a reference prompt word, and on the basis of the reference prompt word, incorporates the advantages of the initial prompt words other than the reference prompt word to generate a mutated prompt word.
[0070] For example, three different candidate cue words r1, r2, and r3 are selected, and semantic mutation is performed using Model 2 to generate the mutated cue word: V i =Model 2("Refer to prompt word r1, incorporate the advantages of r2 but avoid the disadvantages of r3, and generate new mutation prompt words"), where the mutation intensity is controlled by specifying the retention / mutation ratio parameter to control the degree of mutation.
[0071] Step S13: Semantically cross-reference the mutated prompt words with the initial prompt words to generate experimental prompt words. Specifically, the mutated prompt words and the initial prompt words are fused according to the cross-reference probability, retaining the structure of the initial prompt words and incorporating the innovative points of the mutated prompt words to generate experimental prompt words; wherein, the cross-reference probability determines the proportion of content retained from the initial prompt words.
[0072] Step S14: Use the cue word evaluator to evaluate the quality of the test cue words and replace the corresponding initial cue words with test cue words that have higher scores.
[0073] Specifically, Model 3 was used to evaluate the experimental cue word U. i mass f(U) i Model 3("Evaluate the effectiveness of the following prompts in power equipment fault detection scenarios, considering accuracy, clarity, completeness, and professionalism, and give a score of 0-100: {U i}”), if f(U i )>f(T i If so, then use U. i Replace T i Otherwise, keep T. i .
[0074] Optionally, the prompt words in step S10 are derived from a domain knowledge base, and the following steps are included before step S10:
[0075] Construct a multi-scenario domain knowledge base that includes various equipment types, working environments, and fault types. This multi-scenario domain knowledge base includes:
[0076] Data classification system: by equipment type (transformer, circuit breaker, cable, instrument transformer, busbar, etc.); by fault type (electrical fault, mechanical fault, thermal fault, material aging, etc.); by working environment (indoor / outdoor, high humidity / dry, high temperature / low temperature, severe / minor pollution, etc.).
[0077] Multimodal data acquisition: sensor data (temperature, humidity, vibration, sound, oil chromatography, etc.); image data (infrared thermograms, visible light images, partial discharge images, etc.); text data (historical maintenance records, fault reports, expert diagnostic opinions, etc.).
[0078] Complete labeling information: fault diagnosis results, fault severity, fault development trend, handling suggestions and priorities, and expert explanations.
[0079] The multi-scenario domain knowledge base includes specialized terminology in the field of power equipment fault diagnosis, such as "overheating fault," "insulation aging," and "short circuit." During the generation and mutation of prompt words, the model references the multi-scenario domain knowledge base to ensure that the generated prompt words contain an appropriate amount of specialized terminology. By introducing specialized terminology, the generated prompt words are made more aligned with the needs of the power equipment fault diagnosis field.
[0080] In one embodiment, to make the fault diagnosis dataset cover multiple scenarios, adaptive optimization for multiple scenarios is performed, including:
[0081] a) Scene identifier design: Introduce clear scene identifiers in the prompts, such as: "[Scenario: Outdoor high temperature and high humidity] You are an expert in power equipment fault diagnosis...".
[0082] b) Diversity maintenance mechanism: Maintain the diversity of cue words for different scenarios and fault types in the population, and design special elimination rules to avoid a single type of cue word dominating the entire population.
[0083] In one embodiment, in step S20, the fault diagnosis dataset includes instructions and responses, and the instructions include basic instructions, combined instructions, scenario adaptation instructions, and decision support instructions.
[0084] Among them, basic instructions are basic detection instructions for a single device or a single fault type, such as: "Analyze the transformer's temperature data to determine if there is an overheating problem"; combined instructions are instructions that integrate multiple data sources or analyze multiple fault possibilities, such as: "Based on temperature and vibration data, analyze the correlation between transformer overheating and mechanical damage"; scenario adaptation instructions are comprehensive analysis instructions that incorporate working environment factors, such as: "Considering high temperature and high humidity environmental factors, analyze the severity and handling priority of transformer overheating"; and decision support instructions are advanced instructions for fault handling and prevention, such as: "Based on the transformer overheating diagnosis results, generate detailed cause analysis and maintenance plan suggestions".
[0085] Among them, the pre-trained model learns rich language representations and contextual association capabilities through self-supervised learning on massive general corpora. In the field of industrial fault detection, the pre-trained model can be used as a foundation, leveraging its powerful semantic understanding and knowledge transfer capabilities to quickly adapt to power system terminology and fault diagnosis scenarios, thereby significantly improving the model's inference performance with limited labeled data.
[0086] In one embodiment, reference is made to Figure 2 As shown, step S20 also includes preprocessing historical operating data of power equipment and constructing a multi-scenario power equipment fault database. Then, based on the optimized prompt words and the multi-scenario power equipment fault database, a fault diagnosis dataset covering multiple scenarios is constructed.
[0087] In one embodiment, step S30, fine-tuning the pre-trained model for power equipment fault diagnosis using an instruction learning method based on the fault diagnosis dataset, includes:
[0088] Based on a fault diagnosis dataset, this method takes optimized prompts and historical operating data of power equipment as input, and the target response as output. By minimizing the difference between the generated result and the target response (calculated using the cross-entropy loss function), the parameters of the pre-trained power equipment fault diagnosis model are continuously adjusted, achieving fine-tuning of the model. During fine-tuning, the pre-trained model extracts key task information from the optimized prompts, establishing an effective mapping relationship from instructions to responses. This ensures that even when faced with entirely new instructions, the model can accurately understand their intent and generate the expected response. To improve the model's generalization ability and inference depth, this method introduces diverse training strategies: starting with basic, concise instructions and gradually increasing their complexity; constructing positive and negative example comparison samples to enhance the model's sensitivity to subtle differences; and designing a sparse reward function for complex inference tasks to guide the model to form a complete and clear inference chain.
[0089] Specifically, step S40 includes: inputting the actual operating data of the equipment and the prompts optimized through differential processing into the pre-trained model after prompt learning and fine-tuning. The model can then utilize its generalization ability and understanding of specific domain knowledge gained during pre-training and fine-tuning to comprehensively analyze the input information. Finally, the model returns the fault diagnosis results of the power equipment, which may cover information such as the cause of the fault, the type of fault, and the severity, providing a reference for subsequent maintenance and handling.
[0090] Reference Figure 4 As shown, in one embodiment, after step S30, fine-tuning the pre-trained power equipment fault diagnosis model using the instruction learning method based on the fault diagnosis dataset; and before step S40, performing fault diagnosis on the power equipment using the fine-tuned pre-trained power equipment fault diagnosis model, the following steps are also included:
[0091] Step S31: Evaluate the performance of the fine-tuned power equipment fault diagnosis pre-training model. Ensure that the performance of the fine-tuned power equipment fault diagnosis pre-training model meets the requirements.
[0092] It should be noted that the method in this embodiment can be executed by a single device, such as a computer or server. The method can also be applied in a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the method.
[0093] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0094] Based on the same inventive concept, corresponding to any of the above embodiments, this application also provides a power equipment fault diagnosis system, referring to... Figure 5 As shown, the system includes the following modules:
[0095] The prompt word optimization module is used to optimize the prompt words required for power equipment fault diagnosis tasks by combining differential evolution algorithm and multi-model collaboration mechanism;
[0096] The dataset construction module is used to build a fault diagnosis dataset covering multiple scenarios based on optimized prompt words and combined with historical operating data of power equipment;
[0097] The model fine-tuning module is used to fine-tune the pre-trained model for power equipment fault diagnosis based on the fault diagnosis dataset and using the instruction learning method.
[0098] The fault diagnosis module is used to diagnose faults in power equipment using a finely tuned pre-trained model for power equipment fault diagnosis.
[0099] This application guides the model to focus on key information through optimized prompt words, thereby reducing false positive and false negative rates. Through multi-scenario instruction learning and knowledge transfer mechanisms, the model can adapt to different working environments and device types, reducing dependence on specific scenario data. Simultaneously, it innovatively combines differential evolution algorithms with a multi-model collaboration mechanism to achieve systematic optimization of high-quality prompt words, solving the problem that traditional numerical optimization algorithms cannot be directly applied to text optimization. Furthermore, it integrates expert knowledge into the model through instruction learning, reducing reliance on human experts, especially during off-peak hours and in areas with scarce expert resources. Finally, through model collaboration and differential evolution mechanisms, it achieves continuous improvement in system performance, avoiding performance bottlenecks.
[0100] In one embodiment, the prompt word optimization module includes the following sub-modules:
[0101] The initial prompt word generation submodule is used to generate multiple initial prompt words using a prompt word generator; for example... Figure 3 The initial population of prompt words in the text.
[0102] The variant prompt word generation submodule is used to semantically mutate multiple initial prompt words using a prompt word mutant to generate variant prompt words; for example... Figure 3 The prompt word in the offspring population.
[0103] The test prompt word generation submodule is used to perform semantic cross-referencing between the variant prompt words and the initial prompt words to generate test prompt words;
[0104] The quality assessment submodule is used to evaluate the quality of the test prompts using the prompt word evaluator and replace the corresponding initial prompts with test prompts that have higher scores.
[0105] In one embodiment, a cue word mutant is used to semantically mutate multiple initial cue words to generate mutated cue words, including:
[0106] The cue word mutant selects one of the initial cue words as a reference cue word. Based on the reference cue word, it incorporates the advantages of the initial cue words other than the reference cue word to generate a mutant cue word.
[0107] In one embodiment, the above-mentioned semantic cross-referencing of the variant cue word with the initial cue word to generate the test cue word includes:
[0108] The mutated prompts and the initial prompts are fused according to the cross-probability, retaining the structure of the initial prompts and incorporating the innovative points of the mutated prompts to generate experimental prompts; where the cross-probability determines the proportion of content retained from the initial prompts.
[0109] In one embodiment, the aforementioned fault diagnosis dataset includes instructions and responses. The instructions include basic instructions, combined instructions, scenario-adaptive instructions, and decision support instructions. The basic instructions are basic detection instructions for a single device or a single fault type. The combined instructions are instructions that integrate multiple data sources or analyze multiple fault possibilities. The scenario-adaptive instructions are comprehensive analysis instructions that incorporate working environment factors. The decision support instructions are advanced instructions for fault handling and prevention.
[0110] In one embodiment, the above-mentioned fine-tuning of the pre-trained model for power equipment fault diagnosis based on the fault diagnosis dataset using the instruction learning method includes:
[0111] Based on the fault diagnosis dataset, the optimized prompt words and historical operating data of power equipment are used as inputs, and the target response is used as the output. By minimizing the difference between the generated result and the target response, the parameters of the power equipment fault diagnosis pre-training model are continuously adjusted to achieve fine-tuning of the power equipment fault diagnosis pre-training model. In the fine-tuning process, the power equipment fault diagnosis pre-training model extracts key information of the task from the optimized prompt words and establishes an effective mapping relationship from instructions to responses.
[0112] In one embodiment, the power equipment fault diagnosis system further includes the following modules:
[0113] The performance evaluation module is used to evaluate the performance of the fine-tuned pre-trained model for fault diagnosis of power equipment.
[0114] The power equipment fault diagnosis system in this embodiment has the beneficial effects of the above-described method embodiments, which will not be repeated here.
[0115] Based on the same inventive concept, corresponding to any of the above embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the power equipment fault diagnosis method of any of the above embodiments.
[0116] Figure 6This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1101, a memory 1102, an input / output interface 1103, a communication interface 1104, and a bus 1105. The processor 1101, memory 1102, input / output interface 1103, and communication interface 1104 are interconnected internally via the bus 1105.
[0117] The processor 1101 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.
[0118] The memory 1102 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1102 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1102 and is called and executed by the processor 1101.
[0119] Input / output interface 1103 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.
[0120] Communication interface 1104 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).
[0121] Bus 1105 includes a pathway for transmitting information between various components of the device, such as processor 1101, memory 1102, input / output interface 1103, and communication interface 1104.
[0122] It should be noted that although the above-described device only shows the processor 1101, memory 1102, input / output interface 1103, communication interface 1104, and bus 1105, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.
[0123] The electronic devices described above are used to implement the corresponding power equipment fault diagnosis methods in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0124] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to execute the power equipment fault diagnosis method as described in any of the above embodiments.
[0125] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.
[0126] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the power equipment fault diagnosis method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0127] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.
[0128] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.
[0129] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.
[0130] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.
Claims
1. A method for diagnosing faults in power equipment, characterized in that, include: The prompt words required for power equipment fault diagnosis tasks are optimized by combining differential evolution algorithm and multi-model collaboration mechanism; Based on the optimized prompts and combined with historical operating data of power equipment, a fault diagnosis dataset covering multiple scenarios is constructed. Based on the aforementioned fault diagnosis dataset, the pre-trained model for power equipment fault diagnosis is fine-tuned using the instruction learning method. The finely tuned pre-trained model for power equipment fault diagnosis is used to diagnose faults in power equipment.
2. The power equipment fault diagnosis method according to claim 1, characterized in that, The optimization of prompt words required for power equipment fault diagnosis tasks by combining differential evolution algorithm and multi-model cooperation mechanism includes: Use a prompt word generator to generate multiple initial prompt words; A cue word mutant is used to semantically mutate multiple initial cue words to generate mutated cue words; The variant prompt words are semantically cross-referenced with the initial prompt words to generate experimental prompt words; The quality of the test cue words is evaluated using a cue word evaluator, and the corresponding initial cue words are replaced with the test cue words that have a higher score.
3. The power equipment fault diagnosis method according to claim 2, characterized in that, The step of using a cue word mutant to semantically mutate multiple initial cue words to generate mutated cue words includes: The prompt word mutant selects one of the initial prompt words as a reference prompt word, and on the basis of the reference prompt word, incorporates the advantages of the initial prompt words other than the reference prompt word to generate the mutant prompt word.
4. The power equipment fault diagnosis method according to claim 2, characterized in that, The step of semantically crossing the mutated cue word with the initial cue word to generate experimental cue words includes: The mutated prompt word is fused with the initial prompt word according to the cross-probability, retaining the structure of the initial prompt word and incorporating the innovative points of the mutated prompt word to generate the experimental prompt word; wherein, the cross-probability determines the proportion of content retained from the initial prompt word.
5. The power equipment fault diagnosis method according to claim 1, characterized in that, The fault diagnosis dataset includes instructions and responses. The instructions include basic instructions, combined instructions, scenario-adaptive instructions, and decision support instructions. The basic instructions are basic detection instructions for a single device and a single fault type. The combined instructions are instructions that integrate multiple data sources or analyze multiple fault possibilities. The scenario-adaptive instructions are comprehensive analysis instructions that incorporate working environment factors. The decision support instructions are advanced instructions for fault handling and prevention.
6. The power equipment fault diagnosis method according to claim 5, characterized in that, The step of fine-tuning the pre-trained model for power equipment fault diagnosis using the instruction learning method based on the fault diagnosis dataset includes: Based on the fault diagnosis dataset, the optimized prompt words and historical operating data of the power equipment are used as inputs, and the target response is used as the output. By minimizing the difference between the generated result and the target response, the parameters of the power equipment fault diagnosis pre-training model are continuously adjusted to achieve fine-tuning of the power equipment fault diagnosis pre-training model. In the fine-tuning process, the power equipment fault diagnosis pre-training model extracts key information of the task from the optimized prompt words and establishes an effective mapping relationship from instructions to responses.
7. The power equipment fault diagnosis method according to claim 1, characterized in that, After fine-tuning the pre-trained model for power equipment fault diagnosis using the instruction learning method based on the fault diagnosis dataset; Before using the fine-tuned power equipment fault diagnosis pre-trained model to diagnose power equipment faults, the following steps are also included: The performance of the fine-tuned pre-trained model for fault diagnosis of power equipment is evaluated.
8. A fault diagnosis system for power equipment, characterized in that, include: The prompt word optimization module is used to optimize the prompt words required for power equipment fault diagnosis tasks by combining differential evolution algorithm and multi-model collaboration mechanism; The dataset construction module is used to build a fault diagnosis dataset covering multiple scenarios based on optimized prompt words and combined with historical operating data of power equipment; The model fine-tuning module is used to fine-tune the pre-trained model for power equipment fault diagnosis based on the fault diagnosis dataset using the instruction learning method. The fault diagnosis module is used to diagnose faults in power equipment using a finely tuned pre-trained model for power equipment fault diagnosis.
9. The power equipment fault diagnosis system according to claim 8, characterized in that, The prompt word optimization module includes: The initial prompt word generation submodule is used to generate multiple initial prompt words using a prompt word generator. The variant prompt word generation submodule is used to perform semantic mutations on multiple initial prompt words using a prompt word mutant to generate variant prompt words; The test prompt word generation submodule is used to perform semantic cross-referencing between the variant prompt word and the initial prompt word to generate test prompt words; The quality assessment submodule is used to evaluate the quality of the test prompts using a prompt word evaluator and replace the corresponding initial prompts with the test prompts that have a higher score.
10. The power equipment fault diagnosis system according to claim 9, characterized in that, The step of using a cue word mutant to semantically mutate multiple initial cue words to generate mutated cue words includes: The prompt word mutant selects one of the initial prompt words as a reference prompt word, and on the basis of the reference prompt word, incorporates the advantages of the initial prompt words other than the reference prompt word to generate the mutant prompt word.
11. The power equipment fault diagnosis system according to claim 9, characterized in that, The step of semantically crossing the mutated cue word with the initial cue word to generate experimental cue words includes: The mutated prompt word is fused with the initial prompt word according to the cross-probability, retaining the structure of the initial prompt word and incorporating the innovative points of the mutated prompt word to generate the experimental prompt word; wherein, the cross-probability determines the proportion of content retained from the initial prompt word.
12. The power equipment fault diagnosis system according to claim 8, characterized in that, The fault diagnosis dataset includes instructions and responses. The instructions include basic instructions, combined instructions, scenario-adaptive instructions, and decision support instructions. The basic instructions are basic detection instructions for a single device and a single fault type. The combined instructions are instructions that integrate multiple data sources or analyze multiple fault possibilities. The scenario-adaptive instructions are comprehensive analysis instructions that incorporate working environment factors. The decision support instructions are advanced instructions for fault handling and prevention.
13. The power equipment fault diagnosis system according to claim 12, characterized in that, The step of fine-tuning the pre-trained model for power equipment fault diagnosis using the instruction learning method based on the fault diagnosis dataset includes: Based on the fault diagnosis dataset, the optimized prompt words and historical operating data of the power equipment are used as inputs, and the target response is used as the output. By minimizing the difference between the generated result and the target response, the parameters of the power equipment fault diagnosis pre-training model are continuously adjusted to achieve fine-tuning of the power equipment fault diagnosis pre-training model. In the fine-tuning process, the power equipment fault diagnosis pre-training model extracts key information of the task from the optimized prompt words and establishes an effective mapping relationship from instructions to responses.
14. The power equipment fault diagnosis system according to claim 8, characterized in that, Also includes: The performance evaluation module is used to evaluate the performance of the fine-tuned pre-trained model for fault diagnosis of power equipment.
15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the power equipment fault diagnosis method as described in any one of claims 1 to 7.
16. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the power equipment fault diagnosis method according to any one of claims 1 to 7.