Electric energy meter verification fault intelligent judgment system

Abstract

The invention discloses an electric energy meter verification fault intelligent judgment system, and relates to the technical field of electric energy meter fault judgment, and the system comprises the following components: a high-precision monitoring and collection unit which collects various parameters of an electric energy meter in real time through a high-precision metering chip and a sensor, so as to more comprehensively understand the operation environment of the electric energy meter, and improve the detection accuracy of the electric energy meter; according to the method, the fault mode recognition model is constructed through the logistic regression model, the relation between the features and the fault modes can be accurately mapped, the cross entropy loss function and the optimization algorithm are adopted, the accuracy of model prediction is effectively improved, and the accuracy of model prediction is improved. According to the method, model hyper-parameters are adjusted and optimized through the random search technology, the performance of the model is further improved, the stability and reliability of the model in practical application are ensured through inspection of a verification set and a test set, powerful support is provided for rapid and accurate judgment of electric energy meter faults, and the efficiency and accuracy of electric energy meter verification are improved.

Description

Technical Field

[0001] The invention relates to the technical field of electric energy meter fault judgment, and in particular to an electric energy meter fault verification intelligent judgment system. Background Art

[0002] Traditional fault detection methods mainly rely on manual detection and instrument analysis. This method is not only inefficient, but also easily affected by human factors, resulting in low accuracy and reliability of fault detection. In addition, with the rapid development of smart grids, electricity meters, as an important part of smart grids, have increasing complexity and diversity in their operating data. Traditional methods can no longer meet the needs of efficient and accurate identification of electricity meter faults.

[0003] Traditional data processing methods are often unable to effectively extract key features related to fault modes, which seriously affects the accuracy of fault detection. Due to the lack of effective feature selection and optimization methods, it is difficult to select feature subsets that contribute more to classification when building fault mode recognition models, further reducing the efficiency and accuracy of fault detection.

[0004] Therefore, in view of the deficiencies and shortcomings of the prior art, the present invention proposes an intelligent fault judgment system for electric energy meter calibration to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and provide an intelligent fault judgment system for electric energy meter calibration. It can accurately judge the operating status of the electric energy meter through the steps of data acquisition, preprocessing, intelligent feature extraction and fault pattern recognition. It uses high-precision metering chips and sensors to collect parameters in real time, adopts machine learning algorithms to extract key features, and establishes a mapping relationship between fault modes and features. Through model training and hyperparameter tuning, the model performance is further optimized, which will strongly support the calibration and maintenance of electric energy meters and improve the stability and reliability of the power system.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: an intelligent fault judgment system for electric energy meter verification, the system comprising the following components:

[0007] High-precision monitoring and acquisition unit: Through high-precision metering chips and sensors, various parameters of the energy meter are collected in real time to gain a more comprehensive understanding of the operating environment of the energy meter, and the frequency and accuracy of data collection are adjusted to capture subtle changes in the operation of the energy meter;

[0008] Data Optimization and Cleaning Center: Develop more sophisticated data cleaning strategies based on data types and characteristics, use a variety of denoising methods to improve data quality, and select appropriate data standardization methods for subsequent analysis;

[0009] Key feature learning engine: It uses advanced machine learning algorithms to extract features from pre-processed data, and automatically learns the inherent laws and patterns in the data to extract key features that reflect the operating status of the electricity meter;

[0010] Fault mode analysis and modeling unit: Based on the extracted key features, a fault mode recognition model is constructed through a logistic regression model, and the model is trained and learned using historical data to establish a mapping relationship between fault modes and features;

[0011] Fault prediction and alarm module: Input new operating data into the trained fault pattern recognition model to determine whether the electricity meter is in a normal state or has potential faults. For abnormal states, the alarm mechanism will be automatically triggered to send alarm information to relevant personnel.

[0012] Furthermore, the high-precision monitoring and acquisition unit detects the current, voltage, and power-related parameters of the electric energy meter in real time through sensors, adjusts the frequency of data acquisition according to the application scenarios and needs of the electric energy meter, realizes accurate measurement of the parameters of the electric energy meter, converts the relevant data into electrical signals, amplifies, filters, and isolates the signals output by the sensors through the signal conditioning circuit, uses the high-precision metering chip to receive the conditioned signals, and converts them into digital signals, compresses the collected data to reduce the data transmission requirements and storage space occupancy, and transmits the collected data to the data optimization and cleaning center through a specific communication protocol.

[0013] Furthermore, the data optimization and cleaning center checks whether there are missing current, voltage, and power data in the data, fills the missing values ​​by using linear interpolation, identifies duplicate records in the data, merges the duplicate records, and standardizes the data using Z-score standardization. The mathematical formula of Z-score standardization is: Where: x is an observation value in the current data, μ is the mean of the current data, and σ is the standard deviation of the current data, which eliminates the impact of different dimensions or value ranges between different features.

[0014] Furthermore, the key feature learning engine SVM can be used to extract key features from the operation data of the electric energy meter, and RBF is selected as the kernel function. The mathematical formula of RBF is: K(x,x′)=exp-γ||xx′|| 2), where: γ is the parameter of the kernel function, ||xx′|| is the Euclidean distance between two vectors, and the data of the normal or abnormal state of the electric energy meter is used to train the SVM model. During the training process, SVM will find an optimal decision hyperplane to maximize the classification interval. The mathematical formula of the hyperplane is: w·x+b=0, where: w is the normal vector of the hyperplane, b is the intercept, and the features that contribute more to the classification are selected according to the decision boundary of SVM as the features reflecting the operating status of the electric energy meter.

[0015] Furthermore, the features are further screened and optimized by filtering feature selection to select the feature subset that has the greatest influence on fault mode identification. For each feature, the value of mutual information is calculated to measure the correlation between the feature and the normal or abnormal state of the electric energy meter. Among them: X is a certain feature, Y is the normal or abnormal state of the energy meter, x i and i is the ith observation of X and Y, and are the means of X and Y respectively, n is the number of observations, and the features are sorted according to the calculated feature statistics to evaluate the correlation between each feature and the normal or abnormal state of the electric energy meter. The larger the statistic value, the higher the correlation between the feature and the target variable, and the greater the contribution to the classification. Based on the importance sorting of the features, a threshold is set, and the features whose statistic values ​​exceed the threshold are selected as the final feature set, and the feature set is randomly divided into training set, validation set and test set.

[0016] Furthermore, the fault mode analysis and modeling unit uses a logistic regression model to establish a mapping relationship between the fault mode and the feature. The mathematical expression of the logistic regression model is: θ (x) = σ(θ T x), where: h θ (x) represents the probability that the model predicts a failure mode given the input x and parameter θ. It is a sigmoid function, which is used to convert the output of the linear regression model into a probability value, randomly assign initial values ​​to the parameters θ of the logistic regression model, and use the cross entropy loss function to measure the difference between the model prediction and the actual cross entropy loss function. The mathematical expression of the cross entropy loss function is: Where: m is the number of samples, y is (i) is the actual label of the i-th sample, h θ (x (i)) is the predicted probability of the model for the i-th sample. An optimization algorithm is selected to minimize the loss function J(θ). In each iteration, the parameter θ is updated according to the gradient of the loss function with respect to the parameter. The hyperparameters of the model are tuned by random search techniques. Finally, the performance of the model is evaluated by using the validation set.

[0017] Furthermore, the specific steps of selecting the gradient descent method to minimize the loss function are:

[0018] (1) Initialize the model parameters θ and select an appropriate learning rate α. The learning rate controls the step size of each parameter update.

[0019] (2) For the current parameter θ, the loss function J(θ) is calculated using the training set. The mathematical expression is:

[0020]

[0021] (3) Use the calculated gradient to update the model parameters. The update formula is:

[0022] (4) Repeat steps (2) and (3) until the preset number of iterations is reached;

[0023] (5) Use the validation set to evaluate the performance of the trained model.

[0024] Furthermore, the specific steps of tuning the hyperparameters of the model by random search technology are:

[0025] (1) Determine the hyperparameters that need to be tuned and set a reasonable search range for each hyperparameter;

[0026] (2) Randomly select a set of hyperparameter combinations from the defined hyperparameter space;

[0027] (3) Use randomly selected hyperparameter combinations to train the fault pattern recognition model;

[0028] (4) Evaluate the performance of the model on the validation set and record the performance indicators;

[0029] (5) Repeat steps (2) to (4), randomly select hyperparameter combinations multiple times and evaluate model performance;

[0030] (6) Compare the model performance corresponding to all hyperparameter combinations obtained by random search and select the hyperparameters with the best performance;

[0031] (7) Retrain the model on the full training set using the selected best hyperparameter combination.

[0032] Furthermore, the performance of the model is evaluated using the validation set by calculating the accuracy: the ratio of correctly classified samples to the total number of samples. Recall rate: the proportion of all true positive examples that are correctly predicted as positive examples by the model. Precision: The proportion of true positive samples among all samples predicted by the model as positive samples. Among them: TP represents true positive examples, FP represents false positive examples, FN represents false negative examples, and TN represents true negative examples. According to the evaluation results on the validation set, the hyperparameters of the model are adjusted. On the model tuned on the validation set, the final performance evaluation is performed on the test set. According to the evaluation results on the test set, the model structure is adjusted, and different algorithms and enhancement strategies are tried.

[0033] Furthermore, the fault prediction and alarm module collects new operating data from the electric energy meter or other related equipment, performs necessary preprocessing on the data, extracts relevant feature data from the preprocessed data, and inputs the extracted feature data into the fault pattern recognition model. The model will output whether the electric energy meter is in a normal state or has a potential fault based on the input feature data. If the electric energy meter has a potential fault, the alarm mechanism is triggered. After the alarm is triggered, the system will send an alarm message to relevant personnel on the current state of the electric energy meter, the type of fault, and the time when the fault occurred, so that the relevant personnel can take timely measures.

[0034] Compared with the prior art, this electric energy meter fault detection intelligent judgment system has the following beneficial effects:

[0035] 1. The present invention combines the SVM model with the RBF kernel function to efficiently extract key features from the operation data of the electric energy meter, effectively identify the normal or abnormal state of the electric energy meter, and optimize the features using a filtering feature selection method, thereby further improving the accuracy and efficiency of fault mode recognition. The correlation between the features and the state of the electric energy meter is evaluated by calculating the mutual information value, and the threshold is set based on the importance ranking to select features, ensuring that the selected feature set contributes greatly to the classification, thereby improving the accuracy and reliability of the electric energy meter fault judgment and providing strong support for the calibration of the electric energy meter.

[0036] 2. The present invention constructs a fault pattern recognition model through a logistic regression model, which can accurately map the relationship between features and fault patterns. It adopts a cross-entropy loss function and an optimization algorithm to effectively improve the accuracy of model prediction. It tunes the model hyperparameters through random search technology to further improve the performance of the model. After being tested on the validation set and the test set, the stability and reliability of the model in practical applications are ensured, which provides strong support for the rapid and accurate judgment of electric energy meter faults and improves the efficiency and accuracy of electric energy meter calibration.

[0037] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 The present invention is a flow chart of an intelligent fault judgment system for electric energy meter calibration. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Embodiment 1:

[0042] The present embodiment provides an intelligent fault judgment system for electric energy meter calibration, which realizes real-time monitoring and fault prediction of the operating status of the electric energy meter through data collection, preprocessing, intelligent feature extraction and fault pattern recognition. It uses SVM and logistic regression to extract key features and build a fault pattern recognition model, which effectively improves the accuracy and efficiency of fault judgment. At the same time, it uses a random search strategy to optimize the model hyperparameters, further improving the model performance. Through the fault prediction and alarm module, it can detect and handle faults in a timely manner, providing a strong guarantee for the safe and stable operation of the electric energy meter and improving work efficiency and reliability.

[0043] The high-precision monitoring and acquisition unit uses high-precision metering chips and sensors to detect the current, voltage, and power of the electric energy meter in real time and convert them into electrical signals. The signal conditioning circuit amplifies, filters, and isolates the signal output by the sensor to improve the stability and reliability of the signal. The high-precision metering chip is used to receive the conditioned signal and convert it into a digital signal. The collected data is transmitted to the data optimization and cleaning center through a specific communication protocol.

[0044] The data optimization and cleaning center pre-processes the collected electric energy meter operation data, including data cleaning, denoising, and standardization. First, check whether there is missing current, voltage, and power data in the data, and use linear interpolation to fill the missing values. Then, identify duplicate records in the data and merge them. Finally, use the Z-score standardization method to standardize the data. The mathematical formula of z-score standardization is: Eliminate the impact of different dimensions or value ranges between different features.

[0045] The key feature learning engine uses advanced machine learning algorithms to extract features from preprocessed data, uses support vector machine (SVM) as the feature extraction algorithm, and selects radial basis function (RBF) as the kernel function. The mathematical formula of RBF is: K(x,x′)=exp(-γ|||xx′|| 2 ), by training the SVM model, an optimal decision hyperplane is found. The mathematical formula of the hyperplane is: w·x+b=0, which maximizes the classification interval. According to the decision boundary of SVM, the features that contribute more to the classification are selected as the key features reflecting the operating status of the electric energy meter. In order to further optimize the feature set, the filtering feature selection method is used to screen and optimize the features. By calculating the mutual information value between the feature and the normal or abnormal state of the electric energy meter, the mutual information value calculation formula is: Evaluate the relevance of each feature to the target variable, rank the features based on their importance, set a threshold, and select the features whose statistical values ​​exceed the threshold as the final feature set.

[0046] The fault mode analysis and modeling unit identifies the fault mode of the electric energy meter by constructing a logistic regression model, randomly initializes the parameters of the logistic regression model, and uses the cross entropy loss function to measure the difference between the model prediction and the actual fault mode. The gradient descent method combined with the learning rate is selected to control the step size of the parameter update, and the model parameters are updated by iteratively calculating the gradient of the loss function. The random search technique is used to randomly select combinations from the predefined hyperparameter space to train the model, and its performance is evaluated on the validation set to select the best hyperparameter combination. Finally, the system uses the selected best hyperparameter combination to retrain the model on the complete training set, and finally improves the model by calculating the accuracy, recall and precision of the model on the test set, thereby realizing fast and accurate identification of the fault mode of the electric energy meter and improving the efficiency and accuracy of the electric energy meter calibration.

[0047] The fault prediction and alarm module inputs the new operating data into the trained fault pattern recognition model to determine whether the electricity meter is in a normal state or has potential faults. For abnormal states, the system will automatically trigger the alarm mechanism and send alarm information to relevant personnel so that corresponding measures can be taken in time.

[0048] The electric energy meter calibration fault intelligent judgment system described in this embodiment realizes real-time monitoring of the electric energy meter operating status and intelligent fault judgment through high-precision data acquisition, preprocessing, intelligent feature extraction and fault pattern recognition, improves the efficiency and accuracy of electric energy meter fault detection, and provides a strong guarantee for the stable operation of the smart grid.

[0049] Embodiment 2:

[0050] In this embodiment, an intelligent fault determination system for electric energy meters is provided, with particular emphasis on the fault mode analysis and modeling unit.

[0051] The logistic regression model is used to establish the mapping relationship between the fault mode and the feature. The mathematical expression of the logistic regression model is: θ (x) = σ(θ T x), randomly assign initial values ​​to the parameters θ of the logistic regression model, and use the cross entropy loss function to measure the difference between the model prediction and the actual situation. The mathematical expression of the cross entropy loss function is:

[0052] Select the gradient descent algorithm to minimize the loss function, initialize the model parameters θ, select an appropriate learning rate α, and for the current parameters θ, use the training set to calculate the loss function J(θ). The mathematical expression is: Use the calculated gradient to update the model parameters. The update formula is: This process is repeated until the preset number of iterations is reached.

[0053] Use random search technology to tune the hyperparameters of the model. First, determine the hyperparameters that need to be tuned (such as learning rate and number of iterations) and set a reasonable search range for each hyperparameter. Then, randomly select a set of hyperparameter combinations from the defined hyperparameter space and use this set of hyperparameters to train the fault mode recognition model. Evaluate the performance of the model on the validation set and record the performance indicators. Repeat this process multiple times, compare the model performance corresponding to all hyperparameter combinations obtained by random search, select the hyperparameters with the best performance, and finally, use the selected best hyperparameter combination to retrain the model on the complete training set.

[0054] Use the validation set to evaluate the performance of the model and calculate the accuracy Recall and precision The accuracy rate indicates the ratio of correctly classified samples to the total number of samples. The recall rate indicates the ratio of all true positive examples that are correctly predicted as positive by the model. The precision rate indicates the ratio of true positive examples among all samples predicted as positive by the model. According to the evaluation results on the validation set, the hyperparameters of the model are adjusted. The final performance evaluation is performed on the model after being tuned on the validation set using the test set. According to the evaluation results on the test set, the model structure can be adjusted and different algorithms or enhancement strategies can be tried.

[0055] Embodiment three:

[0056] Based on the first and second embodiments, the specific process steps of the system are as follows:

[0057] (1) Capture various parameter data of the electric energy meter in real time through high-precision metering chips and sensors, and convert the data into digital signals;

[0058] (2) Preprocess the collected raw data to reduce noise and interference and improve data quality;

[0059] (3) Use machine learning algorithms (such as SVM) to extract features from the preprocessed data, and use a filtering feature selection method to select the feature subset that is most relevant to the fault pattern recognition of the electric energy meter;

[0060] (4) Use the logistic regression model to establish the failure mode, use the cross entropy loss function to measure the difference between the model prediction and the actual situation, select the gradient descent algorithm to minimize the loss function, use the random search technique to tune the model's hyperparameters, and use the validation set to evaluate the model's performance;

[0061] (5) The new data is input into the fault pattern recognition model to determine the current state of the electricity meter. If the state is abnormal, the alarm mechanism is triggered and an alarm message is sent to relevant personnel.

[0062] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An intelligent fault judgment system for electric energy meter verification, characterized in that: The system consists of the following components: High-precision monitoring and acquisition unit: Through high-precision metering chips and sensors, various parameters of the energy meter are collected in real time to gain a more comprehensive understanding of the operating environment of the energy meter, and the frequency and accuracy of data collection are adjusted to capture subtle changes in the operation of the energy meter; Data Optimization and Cleaning Center: Develop more sophisticated data cleaning strategies based on data types and characteristics, use a variety of denoising methods to improve data quality, and select appropriate data standardization methods for subsequent analysis; Key feature learning engine: It uses advanced machine learning algorithms to extract features from pre-processed data, and automatically learns the inherent laws and patterns in the data to extract key features that reflect the operating status of the electricity meter; Fault mode analysis and modeling unit: Based on the extracted key features, a fault mode recognition model is constructed through a logistic regression model, and the model is trained and learned using historical data to establish a mapping relationship between fault modes and features; Fault prediction and alarm module: Input new operating data into the trained fault pattern recognition model to determine whether the electricity meter is in a normal state or has potential faults. For abnormal states, the alarm mechanism will be automatically triggered to send alarm information to relevant personnel.

2. The intelligent fault judgment system for electric energy meter verification according to claim 1 is characterized in that: The high-precision monitoring and acquisition unit detects the current, voltage, and power-related parameters of the electric energy meter in real time through sensors, adjusts the frequency of data acquisition according to the application scenarios and needs of the electric energy meter, realizes accurate measurement of the parameters of the electric energy meter, converts the relevant data into electrical signals, amplifies, filters, and isolates the signals output by the sensors through the signal conditioning circuit, receives the conditioned signals using the high-precision metering chip, and converts them into digital signals, compresses the collected data to reduce the data transmission requirements and storage space occupancy, and transmits the collected data to the data optimization and cleaning center through a specific communication protocol.

3. The intelligent fault judgment system for electric energy meter verification according to claim 1 is characterized in that: The data optimization and cleaning center checks whether there are missing current, voltage, and power data in the data, fills the missing values ​​by using linear interpolation, identifies duplicate records in the data, merges the duplicate records, and standardizes the data using Z-score standardization. The mathematical formula for Z-score standardization is: Where: x is an observation value in the current data, μ is the mean of the current data, and σ is the standard deviation of the current data, which eliminates the impact of different dimensions or value ranges between different features.

4. The intelligent fault determination system for electric energy meter according to claim 1 is characterized in that: The key feature learning engine SVM can be used to extract key features from the operation data of the electric energy meter, and RBF is selected as the kernel function. The mathematical formula of RBF is: K(x,x′)=exp(-γ||xx′|| 2 ), where: γ is the parameter of the kernel function, ||xx′|| is the Euclidean distance between two vectors, and the data of the normal or abnormal state of the electric energy meter is used to train the SVM model. During the training process, SVM will find an optimal decision hyperplane to maximize the classification interval. The mathematical formula of the hyperplane is: w·x+b=0, where: w is the normal vector of the hyperplane, b is the intercept, and the features that contribute more to the classification are selected according to the decision boundary of SVM as the features reflecting the operating status of the electric energy meter.

5. The intelligent fault determination system for electric energy meter according to claim 4 is characterized in that: The filtering feature selection further screens and optimizes the features to select the feature subset that has the greatest influence on fault mode identification. For each feature, the value of mutual information is calculated to measure the correlation between the feature and the normal or abnormal state of the electric energy meter. Among them: X is a certain feature, Y is the normal or abnormal state of the energy meter, x i and i is the ith observation of X and Y, and y are the means of X and Y respectively, n is the number of observations. According to the calculated feature statistics, the features are sorted, and the correlation of each feature with the normal or abnormal state of the electric energy meter is evaluated. The larger the statistic value, the higher the correlation between the feature and the target variable, and the greater the contribution to the classification. Based on the importance sorting of the features, a threshold is set, and the features with statistical values ​​exceeding the threshold are selected as the final feature set, and the feature set is randomly divided into training set, validation set and test set.

6. The intelligent fault determination system for electric energy meter according to claim 1 is characterized in that: The fault mode analysis and modeling unit uses a logistic regression model to establish a mapping relationship between fault modes and features. The mathematical expression of the logistic regression model is: θ (x) = σ(θ T x), where: h θ (x) represents the probability that the model predicts a failure mode given the input x and parameter θ. It is a sigmoid function, which is used to convert the output of the linear regression model into a probability value, randomly assign initial values ​​to the parameters θ of the logistic regression model, and use the cross entropy loss function to measure the difference between the model prediction and the actual cross entropy loss function. The mathematical expression of the cross entropy loss function is: Where: m is the number of samples, y is (i) is the actual label of the i-th sample, h θ (x (i) ) is the predicted probability of the model for the i-th sample. An optimization algorithm is selected to minimize the loss function J(θ). In each iteration, the parameter θ is updated according to the gradient of the loss function with respect to the parameter. The hyperparameters of the model are tuned by random search techniques. Finally, the performance of the model is evaluated by using the validation set.

7. The intelligent fault determination system for electric energy meter according to claim 6 is characterized in that: The specific steps of selecting the gradient descent method to minimize the loss function are: (1) Initialize the model parameters θ and select an appropriate learning rate α. The learning rate controls the step size of each parameter update. (2) For the current parameter θ, the loss function J(θ) is calculated using the training set. The mathematical expression is: (3) Use the calculated gradient to update the model parameters. The update formula is: (4) Repeat steps (2) and (3) until the preset number of iterations is reached; (5) Use the validation set to evaluate the performance of the trained model.

8. The intelligent fault determination system for electric energy meter according to claim 6 is characterized in that: The specific steps of tuning the hyperparameters of the model by random search technology are: (1) Determine the hyperparameters that need to be tuned and set a reasonable search range for each hyperparameter; (2) Randomly select a set of hyperparameter combinations from the defined hyperparameter space; (3) Use randomly selected hyperparameter combinations to train the fault pattern recognition model; (4) Evaluate the performance of the model on the validation set and record the performance indicators; (5) Repeat steps (2) to (4), randomly select hyperparameter combinations multiple times and evaluate model performance; (6) Compare the model performance corresponding to all hyperparameter combinations obtained by random search and select the hyperparameters with the best performance; (7) Retrain the model on the full training set using the selected best hyperparameter combination.

9. The intelligent fault determination system for electric energy meter according to claim 6 is characterized in that: The performance of the model is evaluated using the validation set by calculating the accuracy: the ratio of correctly classified samples to the total number of samples. Recall rate: the proportion of all true positive examples that are correctly predicted as positive examples by the model. Precision: The proportion of true positive samples among all samples predicted by the model as positive samples. Among them: TP represents true positive examples, FP represents false positive examples, FN represents false negative examples, and TN represents true negative examples. According to the evaluation results on the validation set, the hyperparameters of the model are adjusted. On the model tuned on the validation set, the final performance evaluation is performed on the test set. According to the evaluation results on the test set, the model structure is adjusted, and different algorithms and enhancement strategies are tried.

10. The intelligent fault determination system for electric energy meter according to claim 1, characterized in that: The fault prediction and alarm module collects new operating data from the electric energy meter or other related equipment, performs necessary preprocessing on the data, extracts relevant feature data from the preprocessed data, and inputs the extracted feature data into the fault pattern recognition model. The model will output the result of whether the electric energy meter is in a normal state or has a potential fault based on the input feature data. If the electric energy meter has a potential fault, the alarm mechanism is triggered. After the alarm is triggered, the system will send an alarm message to relevant personnel on the current state of the electric energy meter, the type of fault, and the time when the fault occurred, so that the relevant personnel can take timely measures.

Images