A method, apparatus, terminal equipment, and storage medium for assessing the condition of power equipment.

Abstract

This invention discloses a method, apparatus, terminal equipment, and storage medium for power equipment condition assessment, belonging to the technical field of power equipment operation assessment. The method involves: acquiring structured and unstructured data of the power equipment; wherein the structured data includes: insulation resistance, oil gas concentration, dielectric loss factor, and DC resistance imbalance coefficient; and the unstructured data includes: infrared thermographic images and ultraviolet discharge images; generating a first membership degree of the structured data to each condition category based on the structured data; generating a second membership degree of the unstructured data to each condition category based on the unstructured data; calculating a comprehensive membership degree for each condition category based on the first and second membership degrees; and taking the condition category with the highest comprehensive membership degree as the power equipment condition assessment result. This invention, based on structured and unstructured data, improves the accuracy of power equipment condition assessment.

Description

Technical Field

[0001] This invention relates to the field of power equipment operation assessment technology, and in particular to a power equipment condition assessment method, apparatus, terminal equipment, and storage medium. Background Technology

[0002] As smart grids evolve towards digitalization, networking, and intelligence, condition assessment of power equipment has become a core element in ensuring the safety and stability of the power grid. However, existing assessment technologies have significant shortcomings in areas such as data utilization, weight allocation, and uncertainty handling. Existing technologies suffer from the problem of fragmented value of multi-source heterogeneous data at the data utilization level. Traditional assessment methods rely on single-dimensional structural data (such as dissolved gas concentration in oil) and ignore the synergistic relationship between unstructured and structural data, which can easily lead to inaccurate assessment of the condition of power equipment. Summary of the Invention

[0003] This invention provides a method, apparatus, terminal equipment, and storage medium for assessing the condition of power equipment, which can solve the problem of inaccurate assessment of the condition of power equipment in the prior art.

[0004] An embodiment of the present invention provides a method for assessing the condition of power equipment, comprising: Acquire structured and unstructured data of power equipment; wherein the structured data includes the following structured indicators: insulation resistance, oil gas concentration, dielectric loss factor, and DC resistance imbalance coefficient; the unstructured data includes: infrared thermographic images and ultraviolet discharge images; The structured data is normalized. The unstructured data is preprocessed; Based on the normalized structured data, the first membership degree of the structured data to each state category is generated; wherein, the state categories include: good, normal, attention, abnormal, and fault. Based on the preprocessed unstructured data, generate the second membership degree of the unstructured data to each state category; Calculate the comprehensive membership degree of each state category based on the first and second membership degrees corresponding to each state category; The state category with the highest comprehensive membership degree is taken as the state assessment result of the power equipment.

[0005] Further, the step of generating the first membership degree of the structured data to each state category based on the normalized structured data includes: For each structured index, calculate the third membership degree of the structured index to each state category based on the structured index. For each state category, the first membership degree of the structured data to the state category is calculated based on all the third membership degrees corresponding to the state category and the weight of each structured indicator. The formula for calculating the third membership degree is as follows: ; In the formula, Indicates the first The structured index for the first The third membership degree of each state category; Indicates the first One structured indicator; Indicates the first The structured indicator belongs to the first... Expectations for each state category; Indicates the first The structured indicator belongs to the first... The entropy of each state category; Indicates the first The hyperentropy of a structured index; N(0,1) is a random number sampled from a standard normal distribution.

[0006] The formula for calculating the first membership degree is: ; In the formula, Representing structured data for the first The first membership degree of each state category; This represents the total number of structured indicators; Indicates the first The weights of each structured indicator.

[0007] Further, the step of generating a second membership degree of the unstructured data to each state category based on the preprocessed unstructured data includes: The preprocessed unstructured data is input into the trained power equipment state prediction model so that the power equipment state prediction model can predict the probability value of the power equipment belonging to each state category based on the preprocessed unstructured data, which is used as the second membership degree. The training process of the power equipment prediction model includes: Obtain several historical unstructured data samples of the power equipment that have been labeled with status level tags; The historical unstructured data samples are preprocessed; Each preprocessed historical unstructured data sample is input into the power equipment state prediction model, so that the power equipment state prediction model predicts the probability of the historical unstructured data sample belonging to each state level based on the preprocessed historical unstructured data sample, and outputs the prediction result corresponding to each state level; a loss value is calculated based on the prediction result and the state level label corresponding to the historical unstructured data sample; and the power equipment state prediction model is adjusted based on the loss value.

[0008] Further, the step of calculating the comprehensive membership degree of each state category based on the first and second membership degrees corresponding to each state category includes: Obtain the subjective weight of structured data, the objective weight of structured data, the subjective weight of unstructured data, and the fusion ratio of the objective weight and subjective weight of unstructured data; The comprehensive weight of structured data is calculated based on the subjective weight of structured data, the objective weight of structured data, and the fusion ratio of subjective weight. The comprehensive weight of unstructured data is calculated based on the subjective weight of unstructured data, the objective weight of unstructured data, and the fusion ratio of subjective weight. Calculate the comprehensive membership degree corresponding to each state category based on the first membership degree, the second membership degree, the comprehensive weight of structured data, and the comprehensive weight of unstructured data. The formula for calculating the comprehensive weight of the structured data is as follows: ; In the formula, This represents the overall weight of structured data; Represents subjective weights in structured data; Represents the objective weights of structured data; Indicates the proportion of subjective weighting; The formula for calculating the comprehensive weight of the unstructured data is as follows: ; In the formula, This represents the overall weight of unstructured data; Represents the subjective weights of unstructured data; Represents the objective weights of unstructured data; The formula for calculating the comprehensive membership degree is: ; In the formula, Indicates the first The comprehensive membership degree corresponding to each state category; Representing structured data for the first The first membership degree of each state category; Representing unstructured data to the first The second membership degree of each state category.

[0009] Furthermore, obtain the subjective weights for structured data and unstructured data, including: Obtain importance scores from experts based on several dimensions for structured data; Based on the importance score corresponding to each dimension, a fuzzy judgment matrix is ​​constructed; Calculate the subjective weights of the structured data based on the fuzzy judgment matrix; Calculate the subjective weight of the unstructured data based on the subjective weight of the structured data; The formula for calculating the subjective weight of the unstructured data is as follows: .

[0010] Furthermore, obtaining the objective weights of structured data and unstructured data includes: Obtain the marginal probabilities of structured data for each state category and the marginal probabilities of unstructured data for each state category; The first information entropy of the structured data state evaluation result is calculated based on the marginal probabilities of the structured data for each state category. The second information entropy of the unstructured data state evaluation results is calculated based on the marginal probabilities of the unstructured data for each state category. Based on the marginal probabilities of structured data for each state category and unstructured data for each state category, calculate the redundancy between the state evaluation results of structured data and the state evaluation results of unstructured data. Calculate the objective weight of structured data and the objective weight of unstructured data based on the first information entropy, the second information entropy, and the redundancy. The formula for calculating the objective weight of the structured data is as follows: ; In the formula, The first information entropy represents the state assessment result of structured data; The second information entropy represents the state assessment result of unstructured data; This indicates the redundancy between the state assessment results of structured data and the state assessment results of unstructured data. The formula for calculating the objective weight of the unstructured data is as follows: .

[0011] Furthermore, the subjective weight fusion ratio is obtained, including: Based on the subjective weight of the structured data, the objective weight of the structured data, the subjective weight of the unstructured data, the objective weight of the unstructured data, the first membership degree corresponding to each state category, and the second membership degree corresponding to each state category, a subjective-objective weight deviation function with the subjective weight adjustment coefficient as the variable is constructed. With the goal of minimizing the total difference between subjective and objective weights, the subjective and objective weight deviation function is solved to obtain the target value of the subjective weight adjustment coefficient, which is used as the subjective weight fusion ratio.

[0012] Another embodiment of the present invention provides a power equipment condition assessment device, including: a data acquisition module, a preprocessing module and a condition assessment module; The data acquisition module is used to acquire structured and unstructured data of power equipment; wherein, the structured data includes the following structured indicators: insulation resistance, oil gas concentration, dielectric loss factor, and DC resistance imbalance coefficient; the unstructured data includes: infrared thermographic images and ultraviolet discharge images; The preprocessing module is used to normalize the structured data and preprocess the unstructured data. The status assessment module is used to generate a first membership degree of the structured data to each status category based on the normalized structured data; generate a second membership degree of the unstructured data to each status category based on the preprocessed unstructured data; calculate the comprehensive membership degree of each status category based on the first and second membership degrees; and take the status category with the highest comprehensive membership degree as the status assessment result of the power equipment; wherein, the status categories include: good, normal, attention, abnormal, and fault.

[0013] Another embodiment of the present invention provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the steps of the power equipment condition assessment method of the present invention.

[0014] Another embodiment of the present invention provides a computer-readable storage medium item, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to perform steps as described in the power equipment condition assessment method of the present invention.

[0015] The following benefits can be obtained by implementing the present invention: This invention considers both structured and unstructured data. It calculates the membership degree to state categories based on both structured and unstructured data, and then integrates these two membership degrees to obtain the power equipment condition assessment result. This invention's power equipment condition assessment method based on both structured and unstructured data improves the accuracy of power equipment condition assessment. Attached Figure Description

[0016] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of a power equipment condition assessment method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a power equipment condition assessment device provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the term "comprising" and any variations thereof in the specification, claims and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0020] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "several" means two or more, unless otherwise explicitly defined.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] See Figure 1 To address the problem of inaccurate power equipment condition assessment in existing technologies, an embodiment of the present invention provides a power equipment condition assessment method, comprising: S1. Acquire structured and unstructured data of power equipment; wherein the structured data includes the following structured indicators: insulation resistance, oil gas concentration, dielectric loss factor and DC resistance imbalance coefficient; the unstructured data includes: infrared thermographic images and ultraviolet discharge images.

[0023] In step S1, structured data reflecting the operating status of the equipment is acquired in real time through a sensor network and monitoring system. Specifically, the following structured indicators are included: ① Electrical test parameters: such as insulation resistance (measured by a megohmmeter, range), dielectric loss factor (acquired by a dielectric loss tester according to GB / T5654-2017 standard, accuracy), and DC resistance imbalance coefficient (measured by a DC bridge, resolution); ② Gas data in oil: gas concentration (detection limit) is acquired by an oil chromatograph.

[0024] Industrial cameras and specialized imaging equipment are used to acquire images of the equipment's appearance and status as unstructured data. Specifically, this includes: ① Infrared thermal imaging: using an infrared thermal imager (640×512 resolution, temperature measurement range, accuracy) to capture the surface temperature distribution of the equipment, focusing on easily heated parts such as transformer bushings, switch cabinet contacts, and cable joints; ② Ultraviolet discharge images: using an ultraviolet imager (detection band) to capture the distribution of ultraviolet photons generated by corona discharge and record the location and intensity of the discharge area.

[0025] S2. Normalize the structured data.

[0026] In step S2, the maximum-minimum normalization method is used to perform dimensionless transformation and outlier handling for different types of structured indicators to ensure data comparability.

[0027] S3. Preprocess the unstructured data.

[0028] In step S3, the image data is standardized to account for differences in size, noise, and illumination, while retaining key features: ①Size standardization: Infrared thermal images ( ), ultraviolet images ( ), uniformly reset the size to For each pixel, a bicubic interpolation algorithm is used to ensure edge sharpness, ensuring that critical areas of the device (such as the top of the sleeve and the contact position) account for ≥30% of the image center area; ②Pixel value calibration: In infrared thermal imaging, temperature values ​​are mapped to grayscale values ​​of 0-255 (e.g., ...). , ), through formula Implementation, in which , The operating temperature range of the equipment; Z-score normalization is performed on the RGB three channels for ultraviolet discharge images. ),in , To train the mean and standard deviation of this channel, the influence of light intensity differences is eliminated.

[0029] It should be noted that through the above preprocessing process, unstructured data retains key state characteristics, providing high-quality input for subsequent single-source assessment and decision-making integration. At the same time, the original data and preprocessing parameters are stored synchronously, supporting traceability and re-inspection.

[0030] S4. Based on the normalized structured data, generate the first membership degree of the structured data to each state category; wherein, the state categories include: good, normal, attention, abnormal and fault.

[0031] It should be noted that, based on the equipment type (such as transformer) and indicator characteristics, the range of each structured indicator for each level is defined. For example: H2 concentration in oil: Good [0,30]ppm, Normal (30,50]ppm, Caution (50,80]ppm, Abnormal (80,120]ppm, Severe Fault (120,150]ppm.

[0032] In a preferred embodiment, generating a first membership degree of the structured data to each state category based on the normalized structured data includes: For each structured index, calculate the third membership degree of the structured index to each state category based on the structured index. For each state category, the first membership degree of the structured data to the state category is calculated based on all the third membership degrees corresponding to the state category and the weight of each structured indicator. The formula for calculating the third membership degree is as follows: ; In the formula, Indicates the first The structured index for the first The third membership degree of each state category; Indicates the first One structured indicator; Indicates the first The structured indicator belongs to the first... Expectations for each state category; Indicates the first The structured indicator belongs to the first... The entropy of each state category; Indicates the first Hyperentropy of a structured indicator; The formula for calculating the first membership degree is: ; In the formula, Representing structured data for the first The first membership degree of each state category; This represents the total number of structured indicators; Indicates the first The weights of each structured indicator.

[0033] It should be noted that the formula for calculating the expected value is: ; In the formula, Indicates the first The structured indicator belongs to the first... The minimum value of the interval range of each state category; Indicates the first The structured indicator belongs to the first... The maximum value of the interval range for each state category; The formula for calculating entropy is: .

[0034] In the formula for calculating the third membership degree, By using a forward cloud generator and injecting Gaussian noise to simulate measurement errors, and taking into account the dynamic entropy value after measurement errors, the sensor accuracy is incorporated into the cloud model calculation, making the membership degree closer to the actual measurement scenario.

[0035] The process of determining the weights of structured indicators includes: In terms of weighting based on a combination of subjective and objective factors, the Analytic Hierarchy Process (AHP) is used to construct the judgment matrix, and the weights are calculated using the eigenvector method. A consistency check was performed. Simultaneously, the improved CRITIC method was used to calculate the objective weights. Objective weighting The calculation steps are as follows: ① Data standardization: for the first The first indicator Original values ​​of each sample Performing max-min normalization, we obtain ; ② Calculate the mean: ( (sample size) ③ Calculate the standard deviation: ; ④ Calculate the coefficient of variation: The information entropy and relevance of the indicators are calculated based on the CRITIC method with improved coefficient of variation, and the weights are obtained. : in, The coefficient of variation is a relative statistic used to quantify the dispersion of individual indicator values, objectively reflecting the intensity of fluctuations in indicator data. It is an indicator and Mutual information entropy normalization (i.e., the mutual information entropy mentioned above) through (Result after interval normalization) To achieve comprehensive information entropy, an optimization model is constructed by minimizing the bias between subjective and objective weights to ensure the combined weight coefficients... and satisfy The weights of the structured indicators are obtained. .

[0036] S5. Based on the preprocessed unstructured data, generate the second membership degree of the unstructured data to each state category.

[0037] In a preferred embodiment, generating a second membership degree of the unstructured data to each state category based on the preprocessed unstructured data includes: The preprocessed unstructured data is input into the trained power equipment state prediction model so that the power equipment state prediction model can predict the probability value of the power equipment belonging to each state category based on the preprocessed unstructured data, which is used as the second membership degree. The training process of the power equipment prediction model includes: Obtain several historical unstructured data samples of the power equipment that have been labeled with status level tags; The historical unstructured data samples are preprocessed; Each preprocessed historical unstructured data sample is input into the power equipment state prediction model, so that the power equipment state prediction model predicts the probability of the historical unstructured data sample belonging to each state level based on the preprocessed historical unstructured data sample, and outputs the prediction result corresponding to each state level; a loss value is calculated based on the prediction result and the state level label corresponding to the historical unstructured data sample; and the power equipment state prediction model is adjusted based on the loss value.

[0038] In this embodiment, unstructured data is directly output as a probability distribution for five state levels through an improved CNN.

[0039] The key modules of CNN design include: ① The ESA (Enhanced Spatial Attention) module, as the basic unit of TESA, is designed according to the approach of "feature compression - feature fusion - attention generation - feature enhancement". In the feature compression part, the input features are first processed... Convolution (stride 1, number of output channels equal to the number of input channels) Dimensionality reduction is performed, followed by spatial dimension compression using Global Average Pooling (GAP), and finally... Convolution (stride 1) increases the dimensionality to the original number of channels, forming a compressed feature map. In the feature fusion part, the original input features are added element-wise to the compressed feature map to obtain fused features, enhancing the contextual relevance of the features. In the attention generation part, the fused features are processed... After adjusting the channels using a 3-convolution (stride 1), an attention map in the range of 0-1 is generated using the Sigmoid activation function, where fault regions (such as infrared hotspots and ultraviolet spots) have significantly higher weights than the background. For feature enhancement, the input features are multiplied element-wise with the attention map to increase the response intensity of fault features. This reduces irrelevant background elements (such as device casing textures and ambient lighting).

[0040] ② The PHTCB (Parallel Hybrid Transformer CNN Block) module captures global correlations and local details of power images through parallel Transformer and CNN branches, respectively. It then enhances fault region features through an attention mechanism, ultimately achieving complementary fusion of multi-scale features. Specifically, each PHTCB block consists of a TESA (Triple Enhanced Spatial Attention) block, two parallel TCN (Transformer CNN) blocks, and... The convolutional fusion layer consists of: the TESA block, which uses triple attention iteration to focus on key regions, addressing the issue of small fault area proportions (10%-15%) in power images; and the parallel TCN block, where TCN1 employs a Swing Transformer Layer (…). Window, 4-head attention) modeling global dependencies, TCN2 adopts Convolution (ReLU activation) extracts local features, and the outputs of both are added together and then processed. Convolution compresses channels, balancing computational cost and feature representation capability.

[0041] b) Detailed design of each module: Input layer reception For standardized images, infrared thermal images first need to map temperature values ​​to [0, 255] grayscale and replicate them into 3 channels, then undergo Z-score normalization. Ultraviolet images, on the other hand, have their RGB channels normalized separately to unify image distribution and enhance fault feature contrast. The initial feature extraction layer... Convolution (output) This approach compresses redundant channels to quickly capture basic features such as device outlines and temperature distribution trends, with 64 output channels to retain more power image details. The U-shaped backbone, as the core feature extraction module, contains five PHTCB (Parallel Hybrid Transformer CNN Blocks), designed based on a "global-local feature dual perspective." The U-shaped backbone employs a "3-layer downsampling + 2-layer upsampling" structure with skip connections: in the downsampling segment, PHTCB1 transfers the input from... Processed as (Downsampling is achieved using 3×3 convolutions with a stride of 2, and TESA blocks focus on the device region), PHTCB2 is further processed into (The number of channels was doubled to enhance temperature gradient expression, and the TESA block initially identified the anomalous region.) PHTCB3 processing was performed as follows: (Focusing on localized faults, the TESA block increases the weight of the fault region to 0.8-0.9); In the upsampling segment, PHTCB4 is recovered to [value missing] via bilinear interpolation. By incorporating the low-dimensional features of PHTCB2, PHTCB5 is restored to... Furthermore, it supplements the global temperature distribution characteristics of PHTCB1, and the skip connection design effectively transmits multi-scale information. The classification head first uses global average pooling to... Compress to The Dropout layer with rate=0.2 suppresses overfitting, and then the 128-dimensional features are compressed into 64-dimensional features through a fully connected layer from 128 to 64. The ReLU activation function enhances the model's ability to express complex features. Finally, the 64-dimensional features are mapped into a 5-dimensional vector and the logits value is output through a fully connected layer from 64 to 5. The Softmax layer transforms it into a 5-dimensional probability distribution.

[0042] c) Model Training and Inference: ① Training Data and Labels: Using... ① Group labeled unstructured data, with labels corresponding to the state level, divided into training, validation, and test sets in a 7:2:1 ratio; ② Training strategy: using weighted cross-entropy loss. ,in, For the first Zhang's image belongs to the level. , To predict probabilities for the model, Weights ("Severe Fault" samples are few, There are many "normal" samples. ,the remaining Adam optimizer, cosine annealing learning rate scheduling, training for as many epochs as possible to reach validation set accuracy. ③Inference output: Input preprocessed image ( After forward propagation, the model outputs five probability values ​​(e.g.) ), which is directly used as the second membership degree of image data to each level.

[0043] S6. Calculate the comprehensive membership degree of each state category based on the first and second membership degrees corresponding to each state category.

[0044] In a preferred embodiment, calculating the comprehensive membership degree of each state category based on the first membership degree and the second membership degree corresponding to each state category includes: Obtain the subjective weight of structured data, the objective weight of structured data, the subjective weight of unstructured data, and the fusion ratio of the objective weight and subjective weight of unstructured data; The comprehensive weight of structured data is calculated based on the subjective weight of structured data, the objective weight of structured data, and the fusion ratio of subjective weight. The comprehensive weight of unstructured data is calculated based on the subjective weight of unstructured data, the objective weight of unstructured data, and the fusion ratio of subjective weight. Calculate the comprehensive membership degree corresponding to each state category based on the first membership degree, the second membership degree, the comprehensive weight of structured data, and the comprehensive weight of unstructured data. The formula for calculating the comprehensive weight of the structured data is as follows: ; In the formula, This represents the overall weight of structured data; Represents subjective weights in structured data; Represents the objective weights of structured data; Indicates the proportion of subjective weighting; The formula for calculating the comprehensive weight of the unstructured data is as follows: ; In the formula, This represents the overall weight of unstructured data; Represents the subjective weights of unstructured data; Represents the objective weights of unstructured data; The formula for calculating the comprehensive membership degree is: ; In the formula, Indicates the first The comprehensive membership degree corresponding to each state category; Representing structured data for the first The first membership degree of each state category; Representing unstructured data to the first The second membership degree of each state category.

[0045] In a preferred embodiment, based on the optimized weights, the two types of single-source membership degrees are weighted and fused to generate a comprehensive membership degree. .

[0046] In a preferred embodiment, obtaining the subjective weights of structured data and unstructured data includes: Obtain importance scores from experts based on several dimensions for structured data; Based on the importance score corresponding to each dimension, a fuzzy judgment matrix is ​​constructed; Calculate the subjective weights of the structured data based on the fuzzy judgment matrix; Calculate the subjective weight of the unstructured data based on the subjective weight of the structured data; The formula for calculating the subjective weight of the unstructured data is as follows: .

[0047] In this embodiment, experts scored the data based on three dimensions: "fault correlation" (the strength of the physical correlation between the data and the equipment fault), "measurement stability" (the degree of data interference), and "historical consistency" (the matching rate between past assessments and actual faults). A triangular fuzzy number model was used for the scoring. express,( As the lower limit, The most likely value, (This is the upper limit). Example: For the "fault correlation" dimension, if experts believe that structured data is more important than unstructured data, then the score would be... Conversely ; ② Construction of fuzzy judgment matrix: Integrating expert scores into Fuzzy judgment matrix (rows / columns represent structured data and unstructured data, respectively): ; ③ Weight Calculation and Defuzzification: Calculate the geometric mean of the fuzzy numbers in each row: , respectively obtained as well as Then, by fuzzy summation, we obtain... The weights are obtained through fuzzy division. =(A,B,C) and =(L,M,N), and finally obtained by defuzzification. , .

[0048] In a preferred embodiment, obtaining the objective weights of structured data and unstructured data includes: Obtain the marginal probabilities of structured data for each state category and the marginal probabilities of unstructured data for each state category; The first information entropy of the structured data state evaluation result is calculated based on the marginal probabilities of the structured data for each state category. The second information entropy of the unstructured data state evaluation results is calculated based on the marginal probabilities of the unstructured data for each state category. Based on the marginal probabilities of structured data for each state category and unstructured data for each state category, calculate the redundancy between the state evaluation results of structured data and the state evaluation results of unstructured data. Calculate the objective weight of structured data and the objective weight of unstructured data based on the first information entropy, the second information entropy, and the redundancy. The formula for calculating the objective weight of the structured data is as follows: ; In the formula, The first information entropy represents the state assessment result of structured data; The second information entropy represents the state assessment result of unstructured data; This indicates the redundancy between the state assessment results of structured data and the state assessment results of unstructured data. The formula for calculating the objective weight of the unstructured data is as follows: .

[0049] In this embodiment, the "information content" and "correlation" of the two types of data sources are quantified through statistical analysis to generate objective weights. The specific steps are as follows: ① The uncertainty of single-source data is measured by calculating information entropy, using the following formula: ,in Indicates the device's first Levels ( :good, :normal, :Notice, :abnormal, (Critical malfunction) This indicates that in the single-source assessment results, the equipment belongs to the [number]th [source]. The probability of each state level (i.e., membership degree) or ),get as well as ; ② Measure the evaluation results of structured data through mutual information calculation ( Evaluation results of unstructured image data () The redundancy of ) is calculated using the following formula: ,in express The marginal probability, i.e., the evaluation result of structured data, is a level. The probability (equal to the membership degree of the structured data) ), express The edge probability, i.e., the evaluation result of unstructured image data, is a level. The probability (equal to the membership degree of the image data) ), express and The joint probability, i.e., the structured data is evaluated as a rank. And the image data was assessed as graded. The probability of . Joint probability The calculation method is as follows: Select assessment samples (at least 300 groups) from two types of data collected simultaneously over the past year. Each sample group includes the assessment level of structured data. Evaluation level of image data (e.g., "structured data determination") And unstructured data judgment (This is a sample). The statistical "structure level is..." And the image level is The joint probability is obtained by dividing the number of samples in the first place by the total number of samples. ; ③ Through formula Obtain the objective weights of structured and unstructured data respectively. , .

[0050] In a preferred embodiment, obtaining the subjective weight fusion ratio includes: Based on the subjective weight of the structured data, the objective weight of the structured data, the subjective weight of the unstructured data, the objective weight of the unstructured data, the first membership degree corresponding to each state category, and the second membership degree corresponding to each state category, a subjective-objective weight deviation function with the subjective weight adjustment coefficient as the variable is constructed. With the goal of minimizing the total difference between subjective and objective weights, the subjective and objective weight deviation function is solved to obtain the target value of the subjective weight adjustment coefficient, which is used as the subjective weight fusion ratio.

[0051] In this embodiment, the "information content" and "correlation" of the two types of data sources are quantified through statistical analysis to generate objective weights. The specific steps are as follows: ① Define the deviation function with the goal of minimizing the bias between subjective and objective weights. ,in The adjustment coefficient ( ); ② The optimal subjective weight adjustment coefficient t is obtained through iterative processing using the particle swarm optimization algorithm, thus achieving the best result. and Initialize 50 particles (each particle represents a group). ), learning factor Inertial weight for Update according to a linear rule after each iteration. ( (current iteration number), and recalculate After 100 iterations, the optimal subjective weight adjustment coefficient is obtained. Through formula as well as Get the best and .

[0052] S7. The state category with the highest comprehensive membership degree is taken as the state assessment result of the power equipment.

[0053] In this embodiment, the level with the highest value in the comprehensive membership degree is directly selected as the level status of the device.

[0054] In a preferred embodiment, by employing model quantization (floating-point to 8-bit fixed-point conversion) and pruning techniques, the method of this invention is deployed on edge computing nodes (such as NVIDIA Jetson AGX Orin), with a single evaluation time of ≤200ms, meeting the "data acquisition-analysis-early warning" requirements of microgrid devices and reducing latency by 75% compared to cloud deployment. The system supports standardized interfaces such as IEC 61850 and Modbus, enabling rapid adaptation to equipment such as transformers and switchgear, achieving "plug and play". Using the TensorFlow Lite quantization tool, the convolutional layer weights are compressed from 32-bit floating-point to 8-bit fixed-point, with accuracy loss controlled within 3%.

[0055] It should be noted that, compared with the prior art, the beneficial effects of the present invention are: (1) Multi-source data decision layer fusion: directly weighted integration of the membership distribution of structured data and the probability distribution of unstructured data, retaining the parameter essence of structured data and the original visual features of unstructured images. The fusion object is a low-dimensional membership vector (5-dimensional), which reduces computational complexity and is more suitable for the computing power limitations of edge nodes.

[0056] (2) Optimization of subjective and objective weights: Combining AHP and the improved CRITIC method, the subjective and objective weight ratios are dynamically adjusted through the deviation minimization model to avoid the limitations of single weighting.

[0057] like Figure 2 As shown, based on the above-described method embodiments, an embodiment of the present invention provides a power equipment condition assessment device, including: a data acquisition module, a preprocessing module, and a condition assessment module; The data acquisition module is used to acquire structured and unstructured data of power equipment; wherein, the structured data includes the following structured indicators: insulation resistance, oil gas concentration, dielectric loss factor, and DC resistance imbalance coefficient; the unstructured data includes: infrared thermographic images and ultraviolet discharge images; The preprocessing module is used to normalize the structured data and preprocess the unstructured data. The status assessment module is used to generate a first membership degree of the structured data to each status category based on the normalized structured data; generate a second membership degree of the unstructured data to each status category based on the preprocessed unstructured data; calculate the comprehensive membership degree of each status category based on the first and second membership degrees; and take the status category with the highest comprehensive membership degree as the status assessment result of the power equipment; wherein, the status categories include: good, normal, attention, abnormal, and fault.

[0058] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the power equipment condition assessment method provided by any of the above-described method embodiments of the present invention.

[0059] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0060] Based on the above-described method embodiments, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power equipment condition assessment method of any embodiment of the present invention.

[0061] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0062] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0063] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0064] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the power equipment condition assessment method described in any of the above-described method embodiments of the present invention.

[0065] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0066] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for assessing the condition of power equipment, characterized in that, include: Acquire structured and unstructured data of power equipment; wherein the structured data includes the following structured indicators: insulation resistance, oil gas concentration, dielectric loss factor, and DC resistance imbalance coefficient; the unstructured data includes: infrared thermographic images and ultraviolet discharge images; The structured data is normalized. The unstructured data is preprocessed; Based on the normalized structured data, the first membership degree of the structured data to each state category is generated; wherein, the state categories include: good, normal, attention, abnormal, and fault. Based on the preprocessed unstructured data, generate the second membership degree of the unstructured data to each state category; Calculate the comprehensive membership degree of each state category based on the first and second membership degrees corresponding to each state category; The state category with the highest comprehensive membership degree is taken as the state assessment result of the power equipment.

2. The power equipment condition assessment method as described in claim 1, characterized in that, The step of generating the first membership degree of the structured data for each state category based on the normalized structured data includes: For each structured index, calculate the third membership degree of the structured index to each state category based on the structured index. For each state category, the first membership degree of the structured data to the state category is calculated based on all the third membership degrees corresponding to the state category and the weight of each structured indicator. The formula for calculating the third membership degree is as follows: ； In the formula, Indicates the first The structured index for the first The third membership degree of each state category; Indicates the first One structured indicator; Indicates the first The structured indicator belongs to the first... Expectations for each state category; Indicates the first The structured indicator belongs to the first... The entropy of each state category; Indicates the first The hyperentropy of a structured index; N(0,1) is a random number sampled from a standard normal distribution; The formula for calculating the first membership degree is: ； In the formula, Representing structured data for the first The first membership degree of each state category; This represents the total number of structured indicators; Indicates the first The weights of each structured indicator.

3. The power equipment condition assessment method as described in claim 1, characterized in that, The step of generating a second membership degree for each state category of the unstructured data based on the preprocessed unstructured data includes: The preprocessed unstructured data is input into the trained power equipment state prediction model so that the power equipment state prediction model can predict the probability value of the power equipment belonging to each state category based on the preprocessed unstructured data, which is used as the second membership degree. The training process of the power equipment prediction model includes: Obtain several historical unstructured data samples of the power equipment that have been labeled with status level tags; The historical unstructured data samples are preprocessed; Each preprocessed historical unstructured data sample is input into the power equipment state prediction model, so that the power equipment state prediction model predicts the probability of the historical unstructured data sample belonging to each state level based on the preprocessed historical unstructured data sample, and outputs the prediction result corresponding to each state level; a loss value is calculated based on the prediction result and the state level label corresponding to the historical unstructured data sample; and the power equipment state prediction model is adjusted based on the loss value.

4. The power equipment condition assessment method as described in claim 1, characterized in that, The step of calculating the comprehensive membership degree of each state category based on the first and second membership degrees corresponding to each state category includes: Obtain the subjective weight of structured data, the objective weight of structured data, the subjective weight of unstructured data, and the fusion ratio of the objective weight and subjective weight of unstructured data; The comprehensive weight of structured data is calculated based on the subjective weight of structured data, the objective weight of structured data, and the fusion ratio of subjective weight. The comprehensive weight of unstructured data is calculated based on the subjective weight of unstructured data, the objective weight of unstructured data, and the fusion ratio of subjective weight. Calculate the comprehensive membership degree corresponding to each state category based on the first membership degree, the second membership degree, the comprehensive weight of structured data, and the comprehensive weight of unstructured data. The formula for calculating the comprehensive weight of the structured data is as follows: ； In the formula, This represents the overall weight of structured data; Represents subjective weights in structured data; Represents the objective weights of structured data; Indicates the proportion of subjective weighting; The formula for calculating the comprehensive weight of the unstructured data is as follows: ； In the formula, This represents the overall weight of unstructured data; Represents subjective weights in unstructured data; Represents the objective weights of unstructured data; The formula for calculating the comprehensive membership degree is: ； In the formula, Indicates the first The comprehensive membership degree corresponding to each state category; Representing structured data for the first The first membership degree of each state category; Representing unstructured data to the first The second membership degree of each state category.

5. The power equipment condition assessment method as described in claim 4, characterized in that, Obtain subjective weights for structured data and subjective weights for unstructured data, including: Obtain importance scores from experts based on several dimensions for structured data; Based on the importance score corresponding to each dimension, a fuzzy judgment matrix is ​​constructed; Calculate the subjective weights of the structured data based on the fuzzy judgment matrix; Calculate the subjective weight of the unstructured data based on the subjective weight of the structured data; The formula for calculating the subjective weight of the unstructured data is as follows: 。 6. The power equipment condition assessment method as described in claim 4, characterized in that, The acquisition of objective weights for structured data and objective weights for unstructured data includes: Obtain the marginal probabilities of structured data for each state category and the marginal probabilities of unstructured data for each state category; The first information entropy of the structured data state evaluation result is calculated based on the marginal probabilities of the structured data for each state category. The second information entropy of the unstructured data state evaluation results is calculated based on the marginal probabilities of the unstructured data for each state category. Based on the marginal probabilities of structured data for each state category and unstructured data for each state category, calculate the redundancy between the state evaluation results of structured data and the state evaluation results of unstructured data. Calculate the objective weight of structured data and the objective weight of unstructured data based on the first information entropy, the second information entropy, and the redundancy. The formula for calculating the objective weight of the structured data is as follows: ； In the formula, The first information entropy represents the state assessment result of structured data; The second information entropy represents the state assessment result of unstructured data; This indicates the redundancy between the state assessment results of structured data and the state assessment results of unstructured data. The formula for calculating the objective weight of the unstructured data is as follows: 。 7. The power equipment condition assessment method as described in claim 4, characterized in that, Obtain the subjective weight fusion ratio, including: Based on the subjective weight of the structured data, the objective weight of the structured data, the subjective weight of the unstructured data, the objective weight of the unstructured data, the first membership degree corresponding to each state category, and the second membership degree corresponding to each state category, a subjective-objective weight deviation function with the subjective weight adjustment coefficient as the variable is constructed. With the goal of minimizing the total difference between subjective and objective weights, the subjective and objective weight deviation function is solved to obtain the target value of the subjective weight adjustment coefficient, which is used as the subjective weight fusion ratio.

8. A power equipment condition assessment device, characterized in that, include: Data acquisition module, preprocessing module, and status assessment module; The data acquisition module is used to acquire structured and unstructured data of power equipment; wherein, the structured data includes the following structured indicators: insulation resistance, oil gas concentration, dielectric loss factor, and DC resistance imbalance coefficient; the unstructured data includes: infrared thermographic images and ultraviolet discharge images; The preprocessing module is used to normalize the structured data and preprocess the unstructured data. The status assessment module is used to generate a first membership degree of the structured data to each status category based on the normalized structured data; generate a second membership degree of the unstructured data to each status category based on the preprocessed unstructured data; calculate the comprehensive membership degree of each status category based on the first and second membership degrees; and take the status category with the highest comprehensive membership degree as the status assessment result of the power equipment; wherein, the status categories include: good, normal, attention, abnormal, and fault.

9. A terminal device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the power equipment condition assessment method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the power equipment condition assessment method as described in any one of claims 1-7.

Images