Nondestructive detection method and device for faulty electric energy meter based on broadband impedance spectrum and computer equipment

Abstract

The embodiment of the application provides a kind of based on wide frequency impedance spectrum's fault electric energy meter nondestructive testing method, device, computer equipment, storage medium and computer program product, it is related to electric power detection technical field.The method comprises: obtaining the target impedance data of the electric energy meter to be measured;According to target impedance data, obtain the impedance gray scale chart of the electric energy meter to be measured;Using the fault identification model pre-trained, based on impedance gray scale chart, obtain the fault probability of the electric energy meter to be measured;According to fault probability, obtain the fault detection result of the electric energy meter to be measured.The method improves the intelligentization of electric energy meter fault detection, reduces the operation cost of electric energy meter fault detection, and then, the efficiency, simplicity and accuracy of electric energy meter fault detection can be improved.

Description

Technical Field

[0001] This application relates to the field of power testing technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for non-destructive testing of faulty energy meters based on broadband impedance spectrum. Background Technology

[0002] With the rapid development of smart grids, electricity meters, as core devices for electricity metering and information collection, have been widely used in power systems. Electricity meters not only perform the basic function of accurately measuring electricity consumption, but also transmit electricity consumption data in real time through communication networks, providing data support for power management, user demand response, and energy optimization. This functionality has greatly improved the automation level of the power system and effectively enhanced the utilization efficiency of power resources. However, as the service life of electricity meters increases and their working environment becomes more complex, the types of faults they experience are gradually increasing and their manifestations are becoming more diverse.

[0003] Especially in the context of smart grids, electricity meters are widely installed in various environments, operating under harsh conditions such as high temperature and humidity, low temperature and extreme cold, and strong electromagnetic interference. This significantly increases the risk of meter failure. Furthermore, due to differences in equipment components and manufacturing processes from different suppliers, the failure manifestations of installed electricity meters vary. Some meters may only develop potential problems after prolonged operation, while others may show obvious signs of failure within a short period. This complexity undoubtedly increases the difficulty and complexity of fault detection for retired electricity meters.

[0004] In existing research on electricity meters, various methods have been proposed to address the issues of fault detection and quality control. These include using correlation models to analyze the relationship between meter fault phenomena and component failures, and employing image processing techniques for automated detection of meter display panel quality. However, most studies remain limited to open-cover inspection or visual inspection, failing to comprehensively and effectively identify faults in internal components. While open-cover inspection allows for direct observation of internal components, it is complex and involves high labor costs and safety risks. Visual inspection methods can only identify surface damage or defects, failing to detect internal circuit faults. Penetrating inspection methods such as X-ray imaging can identify damage to internal circuitry, but their high equipment cost and complex operation make them unsuitable for large-scale testing, resulting in low efficiency in electricity meter fault detection. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, computer equipment, storage medium, and computer program product for non-destructive testing of faulty energy meters based on broadband impedance spectrum to address the above-mentioned technical problems.

[0006] Firstly, this application provides a non-destructive testing method for faulty energy meters based on broadband impedance spectrum. The method includes:

[0007] Obtain the target impedance data of the energy meter under test;

[0008] Based on the target impedance data, obtain the impedance grayscale image of the energy meter under test;

[0009] Using a pre-trained fault identification model, the fault probability of the energy meter under test is obtained based on the impedance grayscale image.

[0010] Based on the fault probability, the fault detection result of the energy meter under test is obtained.

[0011] In one embodiment, obtaining the target impedance data of the energy meter under test includes: obtaining the initial impedance data of the energy meter under test; and performing data preprocessing on the initial impedance data to obtain the target impedance data.

[0012] In one embodiment, the step of preprocessing the initial impedance data to obtain the target impedance data includes: cleaning the initial impedance data to obtain cleaned initial impedance data; and normalizing the cleaned initial impedance data to obtain the target impedance data.

[0013] In one embodiment, obtaining the impedance grayscale image of the energy meter under test based on the target impedance data includes: obtaining the grayscale value of each data point among multiple data points of the target impedance data; and drawing the multiple data points on a two-dimensional plane according to a preset arrangement rule based on the grayscale value to obtain the impedance grayscale image of the energy meter under test.

[0014] In one embodiment, the step of using a pre-trained fault identification model to obtain the fault probability of the energy meter under test based on the impedance grayscale image includes: inputting the impedance grayscale image into the pre-trained fault identification model, determining and outputting the fault probability of the energy meter under test.

[0015] In one embodiment, the step of inputting the impedance grayscale image into the pre-trained fault identification model to determine and output the fault probability of the energy meter under test includes: inputting the impedance grayscale image into the pre-trained fault identification model, whereby the feature extraction module of the fault identification model determines and outputs the impedance characteristics of the energy meter under test based on the impedance grayscale image; and the classification module of the fault identification model determines and outputs the fault probability of the energy meter under test based on the impedance characteristics output by the feature extraction module.

[0016] In one embodiment, before acquiring the target impedance data of the energy meter under test, the method further includes:

[0017] Obtain sample impedance data for each of the various energy meter fault types; based on the sample impedance data, obtain a sample impedance grayscale image corresponding to each energy meter fault type; divide the sample impedance grayscale image into a training sample impedance grayscale image and a test sample impedance grayscale image; use the training sample impedance grayscale image to obtain an initial fault identification model; optimize the initial fault identification model based on the test sample impedance grayscale image to obtain the pre-trained fault identification model.

[0018] Secondly, this application provides a non-destructive testing device for faulty energy meters based on a broadband impedance spectrum. The device includes:

[0019] The data measurement module is used to acquire the target impedance data of the energy meter under test;

[0020] The drawing module is used to obtain the impedance grayscale image of the energy meter under test based on the target impedance data.

[0021] The calculation module is used to obtain the fault probability of the energy meter under test based on the impedance grayscale image using a pre-trained fault identification model.

[0022] The result acquisition module is used to acquire the fault detection result of the energy meter under test based on the fault probability.

[0023] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0024] Obtain the target impedance data of the energy meter under test;

[0025] Based on the target impedance data, obtain the impedance grayscale image of the energy meter under test;

[0026] Using a pre-trained fault identification model, the fault probability of the energy meter under test is obtained based on the impedance grayscale image.

[0027] Based on the fault probability, the fault detection result of the energy meter under test is obtained.

[0028] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0029] Obtain the target impedance data of the energy meter under test;

[0030] Based on the target impedance data, obtain the impedance grayscale image of the energy meter under test;

[0031] Using a pre-trained fault identification model, the fault probability of the energy meter under test is obtained based on the impedance grayscale image.

[0032] Based on the fault probability, the fault detection result of the energy meter under test is obtained.

[0033] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0034] Obtain the target impedance data of the energy meter under test;

[0035] Based on the target impedance data, obtain the impedance grayscale image of the energy meter under test;

[0036] Using a pre-trained fault identification model, the fault probability of the energy meter under test is obtained based on the impedance grayscale image.

[0037] Based on the fault probability, the fault detection result of the energy meter under test is obtained.

[0038] In the aforementioned non-destructive testing method, apparatus, computer equipment, storage medium, and computer program product for faulty energy meters based on broadband impedance spectrum, firstly, the target impedance data of the energy meter under test can be obtained; nextly, an impedance grayscale image of the energy meter under test can be obtained based on the target impedance data; furthermore, a pre-trained fault identification model can be used to obtain the fault probability of the energy meter under test based on the impedance grayscale image; finally, the fault detection result of the energy meter under test can be obtained based on the fault probability. The method provided in this application embodiment can combine a pre-trained fault identification model to obtain the fault detection result of the energy meter under test based on its impedance data, avoiding the need to open the meter for inspection, improving the intelligence of energy meter fault detection, reducing the operational cost of energy meter fault detection, and thus improving the efficiency, simplicity, and accuracy of energy meter fault detection. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1A flowchart illustrating a non-destructive testing method for faulty energy meters based on broadband impedance spectrum, provided in an embodiment of this application;

[0041] Figure 2 A schematic diagram of a convolutional neural network model provided in an embodiment of this application;

[0042] Figure 3 A schematic diagram illustrating a process for acquiring target impedance data, provided in an embodiment of this application;

[0043] Figure 4 This application provides a schematic diagram of a data preprocessing process.

[0044] Figure 5 A schematic diagram of a process for obtaining an impedance grayscale image provided in an embodiment of this application;

[0045] Figure 6 A schematic diagram of a process for obtaining a fault identification model provided in an embodiment of this application;

[0046] Figure 7 A schematic diagram of a grayscale image provided in an embodiment of this application;

[0047] Figure 8 A CNN model provided in this application provides accuracy curves on the training and test sets.

[0048] Figure 9 A loss function curve of a CNN model on the training set and test set is provided for an embodiment of this application;

[0049] Figure 10 A structural block diagram of a non-destructive testing device for faulty energy meters based on broadband impedance spectrum provided in this application embodiment;

[0050] Figure 11 This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0052] In one exemplary embodiment, such as Figure 1As shown, a non-destructive testing method for faulty energy meters based on broadband impedance spectrum is provided. This embodiment illustrates the method by applying it to a server. It is understood that this method can also be applied to terminals, and furthermore, to systems including both terminals and servers, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0053] Step 102: Obtain the target impedance data of the energy meter under test.

[0054] The energy meter under test can be any energy meter requiring fault detection. In this step, firstly, the initial impedance data of the energy meter under test can be obtained; then, the initial impedance data can be preprocessed to obtain the target impedance data corresponding to the initial impedance data. Obtaining the initial impedance data of the energy meter under test can include: measuring the impedance of the energy meter under test; specifically, applying AC voltage signals of different frequencies to the port of the energy meter using an impedance testing instrument and measuring the port current to obtain the port impedance, which includes the impedance amplitude and phase angle. The impedance amplitude reflects the impedance capability of the internal circuit of the energy meter to current, while the phase angle reveals the phase response of the internal components of the energy meter to current and voltage. The data preprocessing can include data cleaning and data normalization. Data cleaning removes invalid or noisy data to ensure data quality. The cleaned data contains only the true impedance measurements. Normalization addresses the potential for significant differences in the dimensions and range of the cleaned initial impedance data, requiring normalization to the range of 0 to 1. In this step, the min-max scaling method is used to normalize each data point among multiple data points of the initial impedance data after cleaning, so as to obtain the target impedance data of the energy meter under test, which can be found in formula (1):

[0055]

[0056] in, This represents the normalized value of a single data point. It is the minimum value in the column containing the data. It is the maximum value in the column containing the data. This refers to the actual value of a single data point.

[0057] Step 104: Obtain the impedance grayscale image of the energy meter under test based on the target impedance data.

[0058] The target impedance data can be converted into a grayscale image, which is an image that represents numerical data using grayscale values, typically ranging from 0 to 255. Specifically, the first step is to map each data point of the target impedance data to a grayscale value between 0 and 255. The second step is to plot each target impedance data point on a two-dimensional plane according to a certain arrangement rule to generate the impedance grayscale image of the energy meter under test. Each pixel in the impedance grayscale image corresponds to an impedance measurement value. Black (grayscale value of 0) represents low impedance, and white (grayscale value of 255) represents high impedance, as shown in equation (2).

[0059]

[0060] in, for The normalized impedance measurement, i.e., the target impedance data, This is the grayscale value converted from the target impedance data.

[0061] Step 106: Using a pre-trained fault identification model, the fault probability of the energy meter under test is obtained based on the impedance grayscale image.

[0062] Among them, such as Figure 2 As shown, the pre-trained fault identification model can be a Convolutional Neural Network (CNN). This pre-trained fault identification model can include an input layer, at least one convolutional layer, at least one pooling layer, a fully connected layer, and an output layer. The input layer can receive a grayscale image of the input electricity meter; considering the model size, the input layer size is set to 300×400×1. The convolutional layers can use multiple kernels to extract features from the input image, with depths of 32 and 64 respectively. The kernel size is 5×5, and the stride is 1. Convolutional operations can capture local features in grayscale images, such as impedance abrupt changes or anomalous responses at specific frequencies. Pooling layers can downsample the feature maps output by convolutional layers, reducing data dimensionality and model complexity while preserving key features. The pooling layer has a 2×2 filter size and a stride of 2, and the model uses max pooling. In one possible implementation, the pre-trained fault recognition model may include a flattening layer, which flattens the pooled feature maps into one-dimensional vectors for processing by subsequent fully connected layers. The fully connected layer is responsible for further nonlinear transformations of the flattened features to form the final classification result. The fully connected layer has 64 nodes, and a Dropout layer is added during training to prevent overfitting. The output layer can use a nonlinear activation function to convert the CNN output into a probability distribution of faults and non-faults.

[0063] The convolutional layer extracts features from the input data; the pooling layer samples the input data. Both the convolutional and pooling layers include activation functions. The impedance grayscale image can include multiple impedance data points.

[0064] Specifically, the impedance grayscale image of the energy meter under test can be input into the pre-trained fault identification model. The convolutional layer can be used to extract initial features from multiple impedance data. First, multiple impedance data are vectorized to obtain multiple impedance data vectors, which can be combined into an impedance data vector matrix. Second, this impedance data vector matrix is ​​input into the convolutional layer, and a convolution operation is performed between the convolution kernel and the impedance data vector matrix, i.e., the inner product operation is performed between the impedance data vector matrix and the convolution kernel to obtain the convolution result corresponding to the impedance data vector matrix. Next, a nonlinear transformation is performed on the convolution result based on the activation function, and a bias vector is added to obtain the initial feature vector. Third, the initial feature vector is input into the pooling layer, which can perform feature sampling on the initial feature vector. Then, a nonlinear transformation is performed on the feature sampling result based on the activation function, and a bias vector is added to obtain the impedance characteristics of the energy meter under test.

[0065] This fully connected layer can classify impedance characteristics and obtain fault tags for the energy meter under test. Additionally, the fully connected layer can include an activation function, which comprises a weight matrix and a bias constant.

[0066] Specifically, impedance characteristics can be input into a fully connected layer. Based on the weight matrix and bias vector of the activation function, a nonlinear transformation is performed on the impedance characteristics, followed by normalization to obtain the fault label of the energy meter under test. This fault label indicates whether the energy meter under test has a fault.

[0067] Step 108: Obtain the fault detection results of the energy meter under test based on the fault probability.

[0068] The presence of a fault in the energy meter under test can be characterized by fault probability. This fault probability includes the probability of a fault existing and the probability of no fault existing. The presence of a fault can be determined based on these probabilities. If the probability of a fault existing is greater than the probability of no fault existing, the energy meter under test is considered to be faulty.

[0069] In this embodiment, the method first acquires the target impedance data of the energy meter under test; then, based on the target impedance data, it acquires the impedance grayscale image of the energy meter under test; furthermore, using a pre-trained fault identification model, it acquires the fault probability of the energy meter under test based on the impedance grayscale image; finally, it acquires the fault detection result of the energy meter under test based on the fault probability. This method, provided in this application embodiment, combines a pre-trained fault identification model with the impedance data of the energy meter under test to acquire the fault detection result, avoiding the need to open the meter for inspection. This improves the intelligence of energy meter fault detection, reduces the operational cost of energy meter fault detection, and thus improves the efficiency, simplicity, and accuracy of energy meter fault detection.

[0070] In one exemplary embodiment, such as Figure 3 As shown, step 102 may include steps 302 to 304. Wherein:

[0071] Step 302: Obtain the initial impedance data of the energy meter under test.

[0072] Step 304: Perform data preprocessing on the initial impedance data to obtain the target impedance data.

[0073] This data preprocessing may include data cleaning and data normalization.

[0074] In one exemplary embodiment, such as Figure 4 As shown, step 304 may include steps 402 to 404. Wherein:

[0075] Step 402: Perform data cleaning on the initial impedance data to obtain the cleaned initial impedance data.

[0076] Data cleaning involves removing invalid or noisy data to ensure data quality. The cleaned data contains only the true impedance measurements.

[0077] Step 404: Normalize the initial impedance data after cleaning to obtain the target impedance data.

[0078] The dimensions and range of the initial impedance data after cleaning may vary significantly, requiring normalization to the range of 0 to 1. In this step, the min-max scaling method is used to normalize each data point among multiple data points of the initial impedance data after cleaning, to obtain the target impedance data of the energy meter under test, as shown in equation (1).

[0079] In the method of this embodiment, after obtaining the initial impedance data of the energy meter under test, the initial impedance data can be cleaned and normalized to obtain the normalized impedance measurement value of the energy meter under test, which improves the accuracy of data processing and thus improves the accuracy of fault detection.

[0080] In one exemplary embodiment, such as Figure 5 As shown, step 104 may include steps 502 to 504. Wherein:

[0081] Step 502: Obtain the grayscale value of each data point among multiple data points of the target impedance data.

[0082] Step 504: Based on the grayscale values, multiple data points are plotted on a two-dimensional plane according to a preset arrangement rule to obtain the impedance grayscale map of the energy meter under test.

[0083] The target impedance data can be converted into a grayscale image, which is an image that represents numerical data using grayscale values, typically ranging from 0 to 255. Specifically, the first step is to map each data point of the target impedance data to a grayscale value between 0 and 255. The second step is to plot each target impedance data point on a two-dimensional plane according to a certain arrangement rule, generating an impedance grayscale image of the energy meter under test. Each pixel in the impedance grayscale image corresponds to an impedance measurement value; black (grayscale value of 0) represents low impedance, and white (grayscale value of 255) represents high impedance, as shown in equation (2).

[0084] In the method of this embodiment, the target impedance data of the energy meter under test can be converted into a grayscale image, and then graph convolution can be performed on the grayscale image, which avoids the complex feature extraction process, improves the computational efficiency, and thus improves the detection efficiency of fault detection.

[0085] In one exemplary embodiment, step 106 may include:

[0086] The impedance grayscale image is input into a pre-trained fault identification model to determine and output the fault probability of the energy meter under test.

[0087] In an exemplary embodiment, the step of inputting the impedance grayscale image into a pre-trained fault identification model to determine and output the fault probability of the energy meter under test may include:

[0088] The impedance grayscale image is input into a pre-trained fault identification model. The feature extraction module of the fault identification model determines and outputs the impedance characteristics of the energy meter under test based on the impedance grayscale image. The classification module of the fault identification model determines and outputs the fault probability of the energy meter under test based on the impedance characteristics output by the feature extraction module.

[0089] The pre-trained fault identification model can be a Convolutional Neural Network (CNN). This model may include an input layer, at least one convolutional layer, at least one pooling layer, a fully connected layer, and an output layer. The feature extraction module of the fault identification model may include at least one convolutional layer and at least one pooling layer; the classification module may include at least one fully connected layer. The input layer can receive a grayscale image of the input electricity meter; considering the model size, the input layer size is set to 300×400×1. The convolutional layers can use multiple kernels to extract features from the input image, with depths of 32 and 64 respectively. The kernel size is 5×5, and the stride is 1. Convolutional operations can capture local features in grayscale images, such as impedance abrupt changes or anomalous responses at specific frequencies. Pooling layers can downsample the feature maps output by convolutional layers, reducing data dimensionality and model complexity while preserving key features. The pooling layer has a 2×2 filter size and a stride of 2, and the model uses max pooling. In one possible implementation, the pre-trained fault recognition model may include a flattening layer, which flattens the pooled feature maps into one-dimensional vectors for processing by subsequent fully connected layers. The fully connected layer is responsible for further nonlinear transformations of the flattened features to form the final classification result. The fully connected layer has 64 nodes, and a Dropout layer is added during training to prevent overfitting. The output layer can use a nonlinear activation function to convert the CNN output into a probability distribution of faults and non-faults.

[0090] The convolutional layer extracts features from the input data; the pooling layer samples the input data. Both the convolutional and pooling layers include activation functions. The impedance grayscale image can include multiple impedance data points.

[0091] Specifically, the impedance grayscale image of the energy meter under test can be input into the pre-trained fault identification model. The convolutional layer can be used to extract initial features from multiple impedance data. First, multiple impedance data are vectorized to obtain multiple impedance data vectors, which can be combined into an impedance data vector matrix. Second, this impedance data vector matrix is ​​input into the convolutional layer, and a convolution operation is performed between the convolution kernel and the impedance data vector matrix, i.e., the inner product operation is performed between the impedance data vector matrix and the convolution kernel to obtain the convolution result corresponding to the impedance data vector matrix. Next, a nonlinear transformation is performed on the convolution result based on the activation function, and a bias vector is added to obtain the initial feature vector. Third, the initial feature vector is input into the pooling layer, which can perform feature sampling on the initial feature vector. Then, a nonlinear transformation is performed on the feature sampling result based on the activation function, and a bias vector is added to obtain the impedance characteristics of the energy meter under test.

[0092] This fully connected layer can classify impedance characteristics and obtain fault tags for the energy meter under test. Additionally, the fully connected layer can include an activation function, which comprises a weight matrix and a bias constant.

[0093] Specifically, impedance characteristics can be input into a fully connected layer. Based on the weight matrix and bias vector of the activation function, a nonlinear transformation is performed on the impedance characteristics, followed by normalization to obtain the fault label of the energy meter under test. This fault label indicates whether the energy meter under test has a fault.

[0094] In one exemplary embodiment, such as Figure 6 As shown, the embodiments of this application further include the following steps to provide an implementation method for obtaining a fault identification model, specifically including steps 602 to 610. Wherein:

[0095] Step 602: Obtain sample impedance data for each fault type of the sample energy meter in the various energy meter fault types.

[0096] This process identifies multiple fault types in the sample energy meter, including but not limited to those related to capacitors (power modules), LCD screens, optocouplers (metering circuits), main chips, Bluetooth modules, thermistors, metering chips, transformers, and voltage regulator chips (main circuits). Next, multiple sample impedance data points for each fault type are acquired. Additionally, multiple sample impedance data points for the energy meter under fault-free conditions are acquired as reference samples. In one possible implementation, as shown in Table 1 (fault type classification table), the multiple fault types are numbered to obtain the corresponding fault number, and the number of sample impedance data points for each fault type is determined. For example, capacitors (power modules) correspond to 3 sample impedance data points, and LCD screens correspond to 4 sample impedance data points. (See Table 1 and...) Figure 7 As shown, there are 10 types of faults in the sample energy meters (including normal meters). The specific names and corresponding sample numbers for each fault type are shown in Table 1, and the 10 fault types are numbered 0-9. Among the normal energy meters, two are from the same manufacturer, and the other is from a different manufacturer. For the faulty energy meters, except for those of types 6-8 (which are from the same manufacturer), the energy meters with faults in types 1-5 and 9 are from different manufacturers. This distribution indicates that although some faulty meters are from the same manufacturer, the distribution of fault types is quite broad, involving multiple different manufacturers. This step can also be understood as obtaining the sample impedance data of the sample energy meters and the fault type label for that sample impedance data. This fault type label is used to mark the fault type of the corresponding sample impedance data.

[0097]

[0098] Table 1 Fault Type Classification Table

[0099] like Figure 7 As shown in the figure, (a) and (b) are grayscale images of two normal energy meters from different manufacturers, while (c) and (d) are grayscale images of energy meters from the same manufacturer that are faulty due to different faulty components (where (c) and (d) are faults caused by the thermistor module and the metering chip module, respectively). (e) and (f) show grayscale images of energy meters from different manufacturers that are faulty due to different faulty components (where (e) and (f) are faults caused by the main chip module and the Bluetooth module, respectively). These images show that regardless of whether the energy meter is from the same manufacturer or which specific faulty component is causing the fault, the grayscale image of a faulty energy meter is significantly different from that of a normal energy meter. This difference allows the convolutional neural network to accurately identify faulty energy meters, regardless of their origin or the specific faulty component. The underlying mechanism is that although energy meters produced by different manufacturers may differ in design and manufacturing, the internal equivalent circuit of a normal energy meter is basically consistent; similarly, the internal circuit of a faulty energy meter will exhibit characteristics that are significantly different from those of a normal energy meter due to the presence of the fault. Therefore, convolutional neural networks can effectively distinguish normal electricity meters from faulty electricity meters by capturing these feature differences.

[0100] Step 604: Based on the sample impedance data, obtain the sample impedance grayscale image corresponding to each energy meter fault type.

[0101] In this process, at least one sample impedance data for each type of electricity meter fault can be preprocessed to obtain processed sample impedance data for each type of electricity meter fault. This data preprocessing can include data cleaning and data normalization. Data cleaning removes invalid or noisy data to ensure data quality; the cleaned data contains only the true impedance measurements. Normalization addresses the potential for significant differences in the dimensions and ranges of the cleaned sample impedance data, requiring normalization to the range of 0 to 1. This step employs the Min-Max Scaling method to normalize each data point among multiple data points of the cleaned sample impedance data, yielding processed sample impedance data for each type of electricity meter fault, as shown in equation (1). Furthermore, based on at least one processed sample impedance data for each type of electricity meter fault, a sample impedance grayscale image corresponding to each fault type can be obtained. Specifically, in the first step, each data point among multiple data points of at least one processed sample impedance data is mapped to a grayscale value between 0 and 255. The second step is to plot the grayscale values ​​corresponding to at least one processed sample impedance data on a two-dimensional plane according to a certain arrangement rule, thereby generating a sample impedance grayscale map corresponding to each type of electricity meter fault. Each pixel in the sample impedance grayscale map corresponds to a sample impedance measurement value. Black (grayscale value of 0) represents low impedance, and white (grayscale value of 255) represents high impedance. See equation (3) for details.

[0102]

[0103] in, for The normalized impedance measurement, i.e., the processed sample impedance data, The grayscale values ​​represent the processed sample impedance data.

[0104] Step 606: Divide the sample impedance grayscale image into a training sample impedance grayscale image and a test sample impedance grayscale image.

[0105] Among them, multiple sample impedance grayscale images corresponding to various types of electricity meter faults can be divided into training sample impedance grayscale images and test sample impedance grayscale images. The training sample impedance grayscale images can be used to train the fault identification model to be trained, and the trained fault identification model can be obtained. Then, the trained fault identification model can be optimized using the test sample impedance grayscale images to obtain the optimized fault identification model, which is the pre-trained fault identification model.

[0106] Step 608: Use the impedance grayscale image of the training samples to obtain the initial fault identification model.

[0107] The weights of the convolutional kernels in the fault identification model to be trained can be adjusted using the training sample impedance grayscale image, enabling the model to accurately identify the fault state of the electricity meter. Specifically, the training sample impedance grayscale image and its actual fault label can be input into the fault identification model to be trained. The feature extraction module in the model can obtain the sample impedance features corresponding to the training sample impedance grayscale image. The classification module can obtain the predicted fault label of the training sample impedance grayscale image based on the sample impedance features output by the feature extraction module. Then, the model can be trained based on the difference between the actual fault label and the predicted fault label of the training sample impedance grayscale image, resulting in a trained fault identification model. The feature extraction module of the fault identification model may include at least one convolutional layer and at least one pooling layer; the classification module may include at least one fully connected layer.

[0108] Step 610: Optimize the initial fault identification model based on the impedance grayscale image of the test sample to obtain a pre-trained fault identification model.

[0109] In the testing phase, the grayscale image of the test sample impedance can be input into the trained fault identification model. The trained fault identification model determines and outputs the predicted fault label of the grayscale image of the test sample impedance. Then, the classification performance of the trained fault identification model can be judged based on the difference between the actual fault label and the predicted fault label of the grayscale image of the test sample impedance. For example, the classification accuracy and error performance index can be calculated, and the classification performance of the trained fault identification model can be evaluated based on the classification accuracy and error performance index to obtain the evaluation result of the trained fault identification model. Then, the trained fault identification model can be optimized based on the evaluation result to obtain the optimized fault identification model, which is the pre-trained fault identification model of the final version mentioned above. In one possible implementation, the accuracy of the trained fault identification model, which can be called the initial fault identification model, can be obtained, and the prediction accuracy of the initial fault identification model can be evaluated based on the accuracy. See formula (4):

[0110]

[0111] Depend on Figure 8The accuracy curves of the CNN model on the training and test sets show that it performs exceptionally well in fault meter identification, achieving high accuracy on both sets with minimal difference between them. This indicates that the model did not exhibit significant overfitting during training and can effectively generalize to unseen data. Particularly in distinguishing between normal and faulty meters, the model demonstrates high accuracy and robustness regardless of the component causing the fault. These results suggest that the CNN model is well-suited for fault identification of decommissioned energy meters, exhibiting strong predictive power and stability.

[0112] The difference between the actual fault label and the predicted fault label in the impedance grayscale image of the training sample, and the difference between the actual fault label and the predicted fault label in the impedance grayscale image of the test sample, can be measured by the loss function. The loss function is an important indicator for measuring the model prediction error. It represents the difference between the model prediction value, i.e., the predicted fault label, and the actual value, i.e., the actual fault label. The smaller the value, the more accurate the model prediction. In one possible implementation, the binary cross-entropy loss function can be used to measure the difference between the actual fault label and the predicted fault label. See equation (5):

[0113]

[0114] in, It is the actual fault label (0 or 1). It predicts the fault label (model output), and the predicted fault label can be the predicted probability.

[0115] In addition, by Figure 9 The loss function curves of the CNN model on the training and test sets show that, as the number of training iterations increases, both the training and test losses decrease rapidly in the initial iterations, indicating that the model quickly learns and adjusts the weights to reduce the value of the loss function at the beginning of training. As training progresses, the rate of decrease in the loss function gradually slows down and eventually plateaus, indicating that the model is gradually approaching its optimal state and the loss no longer decreases significantly. Overall, the CNN neural network performs robustly in this task, and the steady decrease in the loss function shows that the model can effectively learn and reduce errors during continuous optimization. This performance demonstrates that the model has good training effectiveness and maintains low error when applied to test data.

[0116] The success rate of classifying different types of electricity meter faults is shown in Table 2.

[0117]

[0118] Table 2 Summary of Model Classification Success Rate

[0119] Each fault type of electricity meter sample contains 45 impedance grayscale images, comprehensively representing its internal circuit structure. Under different fault types, the CNN model achieved a high classification success rate for faulty electricity meters, especially in categories such as "LCD screen," "optical coupler (metering circuit)," and "main chip," where the recognition success rate reached 100%. Although the recognition success rate was slightly lower for certain specific fault types (such as "capacitor (power module)" and "transformer"), overall, the model can accurately distinguish between normal and faulty electricity meters. This indicates that the CNN model can not only identify faulty electricity meters but also effectively ignore the specific component type of the fault, making accurate binary classification judgments.

[0120] In this embodiment, the method can be combined with a pre-trained fault identification model to obtain the fault detection result of the energy meter under test based on the impedance data of the energy meter under test. This avoids opening the cover of the energy meter for testing, improves the intelligence of energy meter fault detection, reduces the operation cost of energy meter fault detection, and thus improves the efficiency, simplicity and accuracy of energy meter fault detection.

[0121] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0122] Based on the same inventive concept, this application also provides a broadband impedance spectrum-based non-destructive testing device for faulty energy meters, used to implement the aforementioned broadband impedance spectrum-based non-destructive testing method for faulty energy meters. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the broadband impedance spectrum-based non-destructive testing device for faulty energy meters provided below can be found in the limitations of the broadband impedance spectrum-based non-destructive testing method for faulty energy meters described above, and will not be repeated here.

[0123] In one embodiment, such as Figure 10As shown, a non-destructive testing device for faulty energy meters based on broadband impedance spectrum is provided, including: a data measurement module 1002, a plotting module 1004, a calculation module 1006, and a result acquisition module 1008, wherein:

[0124] Data measurement module 1002 is used to acquire the target impedance data of the energy meter under test;

[0125] The drawing module 1004 is used to obtain the impedance grayscale image of the energy meter under test based on the target impedance data.

[0126] The calculation module 1006 is used to obtain the fault probability of the energy meter under test based on the impedance grayscale image using a pre-trained fault identification model.

[0127] The result acquisition module 1008 is used to acquire the fault detection result of the energy meter under test based on the fault probability.

[0128] In one embodiment, the data measurement module 1002 is further configured to: acquire the initial impedance data of the energy meter under test; and perform data preprocessing on the initial impedance data to obtain the target impedance data.

[0129] In one embodiment, the data measurement module 1002 is further configured to: perform data cleaning on the initial impedance data to obtain cleaned initial impedance data; and perform normalization processing on the cleaned initial impedance data to obtain the target impedance data.

[0130] In one embodiment, the calculation module 1006 is further configured to: obtain the grayscale value of each of the multiple data points of the target impedance data; and draw the multiple data points on a two-dimensional plane according to a preset arrangement rule based on the grayscale value to obtain the impedance grayscale image of the energy meter under test.

[0131] In one embodiment, the calculation module 1006 is further configured to: input the impedance grayscale image into the pre-trained fault identification model, determine and output the fault probability of the energy meter under test.

[0132] In one embodiment, the calculation module 1006 is further configured to: input the impedance grayscale image into the pre-trained fault identification model, and have the feature extraction module of the fault identification model determine and output the impedance characteristics of the energy meter under test based on the impedance grayscale image; and have the classification module of the fault identification model determine and output the fault probability of the energy meter under test based on the impedance characteristics output by the feature extraction module.

[0133] In one embodiment, the calculation module 1006 is further configured to: acquire sample impedance data of the sample energy meter under each of the various energy meter fault types; acquire sample impedance grayscale images corresponding to each of the energy meter fault types based on the sample impedance data; divide the sample impedance grayscale images into training sample impedance grayscale images and test sample impedance grayscale images; obtain an initial fault identification model using the training sample impedance grayscale images; and optimize the initial fault identification model based on the test sample impedance grayscale images to obtain the pre-trained fault identification model.

[0134] Each module in the aforementioned non-destructive testing device for faulty energy meters based on broadband impedance spectrum can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0135] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores data related to non-destructive testing of faulty energy meters based on broadband impedance spectroscopy. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a non-destructive testing method for faulty energy meters based on broadband impedance spectroscopy.

[0136] Those skilled in the art will understand that Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0137] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0138] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0139] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0140] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0141] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0142] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0143] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A non-destructive testing method for faulty electric energy meters based on broadband impedance spectroscopy, characterized in that, The method comprises: obtaining initial impedance data of a to-be-tested electric energy meter; performing data cleaning on the initial impedance data to obtain cleaned initial impedance data; and performing normalization processing on the cleaned initial impedance data to obtain target impedance data; mapping each of a plurality of data points of the target impedance data to a gray value between 0 and 255; and drawing a plurality of the data points on a two-dimensional plane according to a preset arrangement rule based on the gray value to obtain an impedance gray scale diagram of the to-be-tested electric energy meter; obtaining a fault probability of the to-be-tested electric energy meter based on the impedance gray scale diagram by using a pre-trained fault recognition model; obtaining a fault detection result of the to-be-tested electric energy meter according to the fault probability.

2. The method of claim 1, wherein, The method comprises: inputting the impedance gray scale diagram into the pre-trained fault recognition model to determine and output the fault probability of the to-be-tested electric energy meter.

3. The method of claim 2, wherein, The method comprises: inputting the impedance gray scale diagram into the pre-trained fault recognition model to determine and output the fault probability of the to-be-tested electric energy meter. The method further comprises, before the obtaining of the target impedance data of the to-be-tested electric energy meter:

4. The method according to any one of claims 1 to 3, characterized in that, obtaining sample impedance data of a sample electric energy meter under each electric energy meter fault type in a plurality of electric energy meter fault types; obtaining a sample impedance gray scale diagram corresponding to each electric energy meter fault type according to the sample impedance data; dividing the sample impedance gray scale diagram into a training sample impedance gray scale diagram and a test sample impedance gray scale diagram; obtaining an initial fault recognition model by using the training sample impedance gray scale diagram; optimizing the initial fault recognition model according to the test sample impedance gray scale diagram to obtain the pre-trained fault recognition model. The device comprises:

5. A non-destructive testing device for faulty electric energy meter based on wideband impedance spectroscopy, characterized in that, a data measurement module configured to obtain initial impedance data of a to-be-tested electric energy meter; perform data cleaning on the initial impedance data to obtain cleaned initial impedance data; and perform normalization processing on the cleaned initial impedance data to obtain target impedance data; a drawing module configured to map each of a plurality of data points of the target impedance data to a gray value between 0 and 255; and draw a plurality of the data points on a two-dimensional plane according to a preset arrangement rule based on the gray value to obtain an impedance gray scale diagram of the to-be-tested electric energy meter; a calculation module configured to obtain a fault probability of the to-be-tested electric energy meter based on the impedance gray scale diagram by using a pre-trained fault recognition model; a result obtaining module configured to obtain a fault detection result of the to-be-tested electric energy meter according to the fault probability. The processor implements the steps of the method of any one of claims 1-4 when executing the computer program. 6.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-5 when the computer program is executed by the processor. The processor implements the steps of the method of any one of claims 1-4 when executing the computer program.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, which is executed by a processor, implements the steps of the method according to any one of claims 1 to 4.

Images