Energy consumption analysis-driven electric energy meter preventive maintenance scheduling method and platform

Abstract

The invention relates to the technical field of intelligent electric energy meters, and discloses an energy consumption analysis-driven electric energy meter preventive maintenance scheduling method and platform. According to the method, a deep learning method based on LSTM is utilized, a fault prediction model is constructed through historical energy consumption data, energy consumption data of the electric energy meter are collected in real time and preprocessed, then the fault prediction model is utilized to predict the fault trend of the electric energy meter, and the remaining service life of the electric energy meter is evaluated. And formulating and implementing a preventive maintenance scheduling plan according to the prediction result and the residual service life evaluation. In addition, a corresponding platform is also constructed and comprises a data acquisition module, a data processing module, a fault prediction module, a scheduling planning module, a maintenance execution module and a performance evaluation module. The reliability and the service life of the electric energy meter are improved, the maintenance resource configuration is optimized, the fault rate and the power failure time are reduced, and a comprehensive, efficient and reliable solution is provided for preventive maintenance of the electric energy meter.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent electric energy meters and discloses an electric energy meter preventive maintenance scheduling method and platform driven by energy consumption analysis. Background Art

[0002] In the power grid system, the energy meter is an important metering device, and its stability and accuracy are directly related to the reliability of power supply and the accuracy of measurement. However, with the continuous expansion of the scale of the power grid and the increase in the use time of the energy meter, the failure of the energy meter is frequent, which poses a serious threat to the safe operation of the power grid. Traditional energy meter maintenance methods mostly adopt regular inspections and post-failure repairs, which is not only inefficient, but also difficult to effectively prevent the occurrence of failures. Therefore, how to achieve preventive maintenance of energy meters and discover and solve potential failures in advance has become an important issue in the current power grid operation and maintenance.

[0003] In recent years, with the rapid development of big data and artificial intelligence technologies, data-driven preventive maintenance methods have gradually attracted attention. Among them, the use of deep learning methods to analyze the energy consumption data of electric energy meters and build fault prediction models provides new ideas for preventive maintenance of electric energy meters. By collecting the energy consumption data of electric energy meters in real time, and performing preprocessing and fault prediction, it is possible to accurately judge the fault trend of electric energy meters, and then formulate a scientific and reasonable preventive maintenance scheduling plan to improve the reliability and service life of electric energy meters. Summary of the invention

[0004] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide an energy consumption analysis-driven preventive maintenance scheduling method and platform for electric energy meters, so as to solve the problems of inaccurate fault prediction, untimely maintenance scheduling and low efficiency of electric energy meters in the prior art.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions: First, the energy consumption analysis-driven preventive maintenance scheduling method for electric energy meters includes: Step S100: using a LSTM-based deep learning method and using the historical energy consumption data of the electric energy meter to build a fault prediction model; Step S200: collecting energy consumption data of electric energy meters from the power grid system in real time, and performing data preprocessing on the collected energy consumption data; Step S300: predicting the fault trend of the electric energy meter using a fault prediction model based on the preprocessed real-time energy consumption data, and evaluating its remaining service life; Step S400: Preset a preventive maintenance scheduling plan for the electric energy meter, and implement maintenance scheduling of the electric energy meter according to the fault prediction result and the remaining service life assessment; Step S500: After the maintenance scheduling of the electric energy meter, the performance of the electric energy meter is evaluated.

[0006] Preferably, in a possible implementation manner of the first aspect, the historical energy consumption data includes historical current data, historical voltage data, historical power data, historical fault data and a timestamp; the energy consumption data includes current data, voltage data and power data.

[0007] Preferably, in a possible implementation manner of the first aspect, the fault prediction model structure includes: Input layer, used to receive preprocessed energy consumption data; LSTM layer, used to extract time series features; The fully connected layer receives the features transmitted by the LSTM layer and performs weighted summation and activation processing on these features to perform fault prediction; The output layer is used to output the fault prediction results.

[0008] Preferably, in a possible implementation manner of the first aspect, the data preprocessing includes data cleaning, missing value filling and outlier processing.

[0009] Preferably, in a possible implementation manner of the first aspect, step S300 specifically includes: The preprocessed real-time energy consumption data is input into the fault prediction model, and the fault prediction results of the electric energy meter are output. The results include the fault type, fault probability and the expected time point of the fault. At the same time, based on the fault prediction results and the preset mapping relationship between fault and remaining service life, the remaining service life of the electric energy meter is evaluated.

[0010] Preferably, in a possible implementation manner of the first aspect, the preventive maintenance scheduling plan includes: Determine the maintenance priority order based on the failure prediction results and remaining service life assessment of the energy meter; Establish maintenance time windows for different electric energy meters, and arrange maintenance personnel and required spare parts; Set the specific content and execution standards of maintenance tasks, and perform preventive maintenance before the electricity meter fails or reaches its critical service life.

[0011] Preferably, in a possible implementation manner of the first aspect, step S500 specifically includes: After the energy meter is maintained, the energy consumption data of the energy meter is collected again, and compared and analyzed with the energy consumption data before maintenance to evaluate the maintenance effect; Adjust the parameters of the fault prediction model based on the evaluation results to optimize the model performance.

[0012] Preferably, in a possible implementation manner of the first aspect, the method further includes recording all energy consumption data, fault prediction results, maintenance scheduling results and performance evaluation results in a database.

[0013] Second, the energy consumption analysis-driven electricity meter preventive maintenance scheduling platform includes: A data acquisition module is used to collect energy consumption data of electric energy meters from the power grid system in real time; A data processing module is used to pre-process historical energy consumption data and real-time energy consumption data; The fault prediction module uses an LSTM-based fault prediction model to predict the fault trend of the electric energy meter and evaluate the remaining service life; Scheduling plan module, which formulates preventive maintenance scheduling plans based on fault prediction results and remaining service life; Maintenance execution module, responsible for implementing the maintenance scheduling plan and recording the maintenance process; The performance evaluation module is used to evaluate the performance of the electric energy meter after maintenance and to optimize the fault prediction model based on the evaluation results.

[0014] The beneficial effects of the present invention are: by analyzing historical and real-time energy consumption data through deep learning methods, a high-precision fault prediction model is constructed, and real-time monitoring and early warning of the status of the electric energy meter are realized, thereby improving the reliability and service life of the electric energy meter. The method formulates a preventive maintenance scheduling plan based on the fault prediction results and the remaining service life assessment, optimizes the maintenance resource allocation, reduces the failure rate, and reduces the power outage time and economic losses. At the same time, the performance of the electric energy meter after maintenance is evaluated through the performance evaluation module, and the fault prediction model is optimized based on the evaluation result feedback, ensuring that the accuracy of the model prediction and the scientific nature of the maintenance scheduling are continuously improved, providing a comprehensive, efficient and reliable solution for the preventive maintenance of the electric energy meter. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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 description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flow chart of a method for preventive maintenance scheduling of electric energy meters driven by energy consumption analysis is provided for this application.

[0017] Figure 2 A structural diagram of an energy meter preventive maintenance scheduling platform driven by energy consumption analysis is provided for this application.

[0018] Explanation of the accompanying drawings: 1-data acquisition module, 2-data processing module, 3-fault prediction module, 4-scheduling planning module, 5-maintenance execution module, 6-performance evaluation module. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0020] Embodiment 1: Figure 1 As shown, the present invention provides an energy consumption analysis driven electric energy meter preventive maintenance scheduling method, comprising: Step S100: Adopting the LSTM-based deep learning method and using the historical energy consumption data of the electric energy meter to build a fault prediction model.

[0021] Specifically, the historical energy consumption data includes historical current data, historical voltage data, historical power data, historical fault data and timestamp information, all of which are obtained from the power grid system.

[0022] Historical current data records the current changes of the equipment during its past operation, reflecting the equipment's power consumption, load characteristics, and potential abnormal or fault conditions; historical voltage data records the fluctuations of the equipment's input voltage. By analyzing these fluctuations, we can evaluate the stability of the power system, the adaptability of the equipment, and possible voltage quality problems; historical power data includes active power and reactive power, etc. By analyzing these data, we can evaluate the equipment's energy efficiency, operating status, and whether there is abnormal energy consumption behavior; historical fault data records the fault events that occurred in the equipment's past operation and related information.

[0023] In this embodiment, a deep learning method based on LSTM is used to construct a fault prediction model, and the fault prediction model structure includes an input layer, an LSTM layer, a fully connected layer, and an output layer.

[0024] The input layer is responsible for receiving preprocessed historical energy consumption data, including historical current, voltage, power and other time series data and timestamp information. These data are organized into a three-dimensional tensor with a shape of (batch_size, sequence_length, input_size), where batch_size represents the number of samples used in each training iteration, sequence_length represents the number of time steps in each sample, and input_size represents the number of features in each time step.

[0025] The LSTM layer is used to capture long-term dependencies in time series data. Through the control of the forget gate, input gate, and output gate, the LSTM layer can effectively remember important information and forget unimportant information. The parameters of the LSTM layer include input_size (the number of input features), hidden_size (the size of the hidden state of the LSTM unit), num_layers (the number of stacks of LSTM layers), and batch_first (a Boolean value used to indicate whether the first dimension of the input tensor is the batch size). The output of the LSTM layer includes hidden states and cell states, where the hidden state of the last layer is used as the input of the fully connected layer.

[0026] The fully connected layer is responsible for mapping the features extracted by the LSTM layer to the final prediction target. This layer contains a weight matrix and a bias vector, which are used to calculate the weighted sum of the input features. The parameters of the fully connected layer include in_features (the number of input features, i.e. the size of the hidden state of the LSTM layer) and out_features (the number of output features, i.e. the number of targets that the model needs to predict). After the fully connected layer, the sigmoid activation function is applied to introduce nonlinear relationships.

[0027] The output of the fully connected layer is the model's estimate of the prediction target. For fault prediction tasks, the output includes fault type (classification problem), fault probability (regression problem), and the time point at which the fault is expected to occur (regression problem). Among them, the prediction of the fault type is the responsibility of a subset of nodes in the fully connected layer. The number of outputs of these nodes is equal to the number of fault categories, and the softmax activation function is applied to convert the output into a probability distribution. The prediction of the fault probability is the responsibility of another subset of nodes, and its output is a real number between 0 and 1, indicating the possibility of a fault. The prediction of the time point at which the fault is expected to occur is also the responsibility of another subset of nodes, and its output is a real number representing the interval from the current time point to the time point at which the fault is expected to occur.

[0028] As the final layer of the model, the output layer is responsible for converting the output of the fully connected layer into the model's final estimate of the prediction target.

[0029] Step S200: collecting energy consumption data of electric energy meters from the power grid system in real time, and performing data preprocessing on the collected energy consumption data.

[0030] Specifically, energy consumption data mainly includes current data, voltage data and power data, all of which are obtained in real time from the power grid system.

[0031] The data preprocessing stage includes multiple steps such as data cleaning, missing value filling and outlier processing. In this embodiment, the read_csv function of the Pandas library is used to read the real-time energy consumption data.

[0032] The read data is stored in a DataFrame object named "data". To ensure the uniqueness of the data, the drop_duplicates method of DataFrame is used to remove duplicate rows in "data". For missing values ​​in the three columns of current (labeled as "current"), voltage (labeled as "voltage"), and power (labeled as "power"), first try to fill them with the previous non-null value (i.e., forward filling). If there are still missing values ​​after this step, they are further filled with the average value of the current column.

[0033] In order to effectively identify and process outliers in the current, voltage, and power columns, the Z scores (i.e., standard scores) of these columns are first calculated and their absolute values ​​are taken. Subsequently, a threshold is set (set to 3 in this embodiment), and rows with Z scores greater than the threshold are considered to contain outliers. For these rows containing outliers, the median of the current column is used to replace the outliers.

[0034] To verify the effect of data preprocessing, first print the first five rows of the processed DataFrame for a quick review. Then, use the isnull method to check whether there are still missing values ​​in each column of the DataFrame, and use the sum method to calculate the number of missing values ​​in each column to ensure the quality of data preprocessing.

[0035] Finally, use the to_csv method of DataFrame to save the processed data into a new CSV file named "preprocessed_energy_consumption_data.csv". In the saving process, choose not to save the row index to keep the data neat.

[0036] Step S300: predicting the fault trend of the electric energy meter using a fault prediction model based on the preprocessed real-time energy consumption data, and evaluating its remaining service life.

[0037] Specifically, the preprocessed real-time energy consumption data is input into the fault prediction model to output the fault prediction results of the electric energy meter, which include the fault type, fault probability and the expected time point of the fault. At the same time, based on the fault prediction results and the preset mapping relationship between fault and remaining service life, the remaining service life of the electric energy meter is evaluated.

[0038] In this embodiment, the real-time energy consumption data after data cleaning, missing value filling and outlier processing is organized according to the input format of the fault prediction model and input into the trained fault prediction model. These energy consumption data cover key parameters such as current, voltage, and power, and are organized into a three-dimensional tensor to meet the input requirements of the model.

[0039] The fault prediction model uses its internal neural network architecture to deeply mine and analyze the input energy consumption data to extract features related to the fault. These features are used to predict the type of fault, that is, to accurately determine the specific type of fault that the electric energy meter may encounter. At the same time, the model will also output a probability value of the fault occurrence, which is between 0 and 1, intuitively reflecting the possibility of the fault. In addition, the model also has the ability to predict the expected time point of the fault, that is, it can estimate the time interval from the current time point to the expected fault occurrence.

[0040] After obtaining the fault prediction results, the remaining service life of the energy meter is evaluated based on the preset fault and remaining service life mapping relationship. This mapping relationship can convert key information such as fault type, fault probability and expected fault time point into an estimated value of the remaining service life of the energy meter.

[0041] In this embodiment, the mapping relationship defines the corresponding relationship between the fault type and the impact coefficient, the fault probability and the life reduction, and the predicted fault occurrence time point and the urgency level.

[0042] Fault type and impact coefficient: Different fault types have significantly different impacts on the remaining service life of the energy meter. To this end, an impact coefficient matrix is ​​defined to accurately convert the fault type into the specific impact on the remaining service life. Fault types include fault type A, fault type B, and fault type C.

[0043] Fault type A is a serious fault with a large impact. The impact coefficient is set to 0.8. This type of fault directly affects the metering accuracy, safety performance or basic functions of the energy meter, and may cause the energy meter to fail to work properly or need to be replaced immediately. Specifically, it includes metering module failure: the energy meter cannot accurately measure the electric energy, such as damage to the metering chip, open circuit or short circuit of the metering circuit; power module failure: the energy meter cannot be powered or the power supply is unstable, such as damage to the power circuit and battery failure; communication module failure: the energy meter cannot communicate with the remote system, such as damage to the communication chip and communication interface failure.

[0044] Fault type B is a medium fault with medium impact. The impact coefficient is set to 0.5. Although this type of fault will not immediately cause the energy meter to stop working, it will affect some of its functions or performance and needs to be repaired as soon as possible to avoid further deterioration. Specifically, it includes display screen failure, which causes the energy meter display screen to be unable to display information normally, such as display screen damage and display drive circuit failure; key failure, which causes the energy meter key to be unable to operate normally, such as key failure and key circuit failure; clock module failure, which causes the energy meter time to display inaccurately or fail to record timestamps, such as clock chip damage and clock circuit failure.

[0045] Fault type C is a minor fault with little impact. The impact coefficient is set to 0.2. This type of fault has little impact on the metering accuracy, safety performance or basic functions of the energy meter, but may affect the user experience or require regular maintenance. Specifically, it includes shell damage: the appearance of the energy meter is damaged, but it does not affect its internal functions, such as scratches and dents on the shell; indicator light failure: the indicator light of the energy meter cannot indicate the status normally, such as indicator light damage, indicator light drive circuit failure; software minor errors: cause the energy meter to have minor errors or abnormalities in certain specific situations, but do not affect its normal metering and communication functions, such as software logic errors and improper parameter settings.

[0046] Failure probability and life reduction: The higher the failure probability, the shorter the remaining service life. The failure probability ranges from 0 to 1, and the reduction factor is proportional to the failure probability.

[0047] Expected time of failure and urgency: The closer the expected time of failure is, the more urgent the maintenance is, and the remaining service life should be shortened accordingly. The urgency adjustment value is determined based on the expected time of failure. Specifically, a time window T is set (1 hour in this embodiment). If the expected time of failure falls within this time window, the urgency is high and the urgency adjustment value is set to 1.5; if it is outside this time window, the urgency is low and the urgency adjustment value is set to 1.

[0048] In the calculation process of the remaining service life, the initial remaining service life of the electric energy meter is first determined based on factors such as the design life of the electric energy meter and historical maintenance records. Then, based on the fault prediction results (including fault type, fault probability, and expected time of fault occurrence), the initial remaining service life is dynamically adjusted using the influence coefficient, reduction factor, and urgency adjustment value in the mapping relationship. The calculation formula for the final remaining service life is: Final remaining service life = Initial remaining service life - (influence coefficient × reduction factor × urgency adjustment value).

[0049] Step S400: Preset a preventive maintenance scheduling plan for the electric energy meter, and implement maintenance scheduling of the electric energy meter according to the fault prediction result and the remaining service life evaluation.

[0050] Specifically, the preventive maintenance scheduling plan includes: determining the maintenance priority order based on the fault prediction results and remaining service life assessment of the electricity meter; formulating maintenance time windows for different electricity meters, arranging maintenance personnel and required spare parts; setting the specific content and execution standards of the maintenance tasks, and performing preventive maintenance before the electricity meter fails or reaches its critical service life.

[0051] In this embodiment, the maintenance priority order of each electric energy meter is first determined based on the fault prediction results and remaining service life evaluation of the electric energy meter. For electric energy meters with higher failure probability, closer predicted failure time point or shorter remaining service life, higher priority is given to ensure that these devices can receive timely maintenance attention.

[0052] Secondly, a specific maintenance time window is set for each power meter. In the process of setting, the overall operation requirements of the power grid system, the actual availability of maintenance personnel, and the real-time status of spare parts inventory are taken into consideration. By comprehensively considering these factors, we strive to ensure that the maintenance work proceeds smoothly while minimizing the impact on the daily operation of the power grid system.

[0053] Next, when formulating the maintenance plan, the specific content and execution standards of the maintenance tasks were clarified, including the specific division of responsibilities of maintenance personnel, the precise preparation of required spare parts and tools, various safety precautions during maintenance, and strict quality inspection after maintenance.

[0054] Finally, according to the maintenance plan, maintenance personnel and required spare parts are arranged, and preventive maintenance measures are taken before the electric energy meter fails or reaches its critical service life.

[0055] Step S500: After the maintenance scheduling of the electric energy meter, the performance of the electric energy meter is evaluated.

[0056] Specifically, after the electric energy meter completes maintenance, the operating data of the electric energy meter is recollected and compared with the data before maintenance to evaluate the maintenance effect; the parameters of the fault prediction model are adjusted according to the evaluation results to optimize the model performance.

[0057] In this embodiment, the operation data of the electric energy meter that has completed the maintenance operation is firstly collected again, and these data cover performance indicators such as current, voltage, power, etc. These data are compared and analyzed with the historical data before maintenance, so as to evaluate the actual effect of the maintenance operation.

[0058] In the comparative analysis phase, the changes in data before and after maintenance are observed to directly reflect the degree of improvement in the performance of the energy meter. For example, key indicators such as the reduction in current and voltage fluctuations and the improvement in power factor. The changes in these indicators provide a reference for evaluating the maintenance effect.

[0059] In addition, the parameters of the fault prediction model are adjusted based on the results of the performance evaluation. If the electric energy meter after maintenance shows more stable performance, this verifies the accuracy of the current fault prediction model to a certain extent; on the contrary, if the electric energy meter after maintenance still frequently fails or has large performance fluctuations, the model will be analyzed in depth and a series of optimization measures will be considered. These optimization measures may include adjusting key parameters such as the number of layers and the number of neurons of the LSTM network, increasing the size of the training data set to improve the generalization ability of the model, and introducing other advanced machine learning algorithms for ensemble learning.

[0060] In summary, by evaluating the performance of the electric energy meter after maintenance, we can not only accurately judge the effect of the maintenance work, but also continuously optimize the fault prediction model based on the evaluation results, thereby continuously improving the prediction accuracy and reliability of the model.

[0061] In one embodiment, the method further includes recording all energy consumption data, fault prediction results, maintenance scheduling results, and performance evaluation results into a database.

[0062] Embodiment 2: Figure 2 As shown, the present invention provides an energy consumption analysis-driven electric energy meter preventive maintenance scheduling platform, including: The data acquisition module 1 is used to collect energy consumption data of electric energy meters from the power grid system in real time; Data processing module 2, used for preprocessing historical energy consumption data and real-time energy consumption data; Fault prediction module 3 uses an LSTM-based fault prediction model to predict the fault trend of the electric energy meter and evaluate the remaining service life; Scheduling plan module 4, formulates preventive maintenance scheduling plan based on fault prediction results and remaining service life; Maintenance execution module 5, responsible for implementing the maintenance scheduling plan and recording the maintenance process; The performance evaluation module 6 is used to evaluate the performance of the electric energy meter after maintenance and to optimize the fault prediction model based on the evaluation result feedback.

[0063] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An energy consumption analysis-driven preventive maintenance scheduling method for electric energy meters, characterized in that: include: Step S100: using a LSTM-based deep learning method and using the historical energy consumption data of the electric energy meter to build a fault prediction model; Step S200: collecting energy consumption data of electric energy meters from the power grid system in real time, and performing data preprocessing on the collected energy consumption data; Step S300: predicting the fault trend of the electric energy meter using a fault prediction model based on the preprocessed real-time energy consumption data, and evaluating its remaining service life; Step S400: Preset a preventive maintenance scheduling plan for the electric energy meter, and implement maintenance scheduling of the electric energy meter according to the fault prediction result and the remaining service life assessment; Step S500: After the maintenance scheduling of the electric energy meter, the performance of the electric energy meter is evaluated.

2. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The historical energy consumption data includes historical current data, historical voltage data, historical power data, historical fault data and timestamp; the energy consumption data includes current data, voltage data and power data.

3. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The fault prediction model structure includes: Input layer, used to receive preprocessed energy consumption data; LSTM layer, used to extract time series features; The fully connected layer receives the features transmitted by the LSTM layer and performs weighted summation and activation processing on these features to perform fault prediction; The output layer is used to output the fault prediction results.

4. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The data preprocessing includes data cleaning, missing value filling and outlier processing.

5. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The step S300 specifically includes: The preprocessed real-time energy consumption data is input into the fault prediction model, and the fault prediction results of the electric energy meter are output. The results include the fault type, fault probability and the expected time point of the fault. At the same time, based on the fault prediction results and the preset mapping relationship between fault and remaining service life, the remaining service life of the electric energy meter is evaluated.

6. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The preventive maintenance scheduling plan includes: Determine the maintenance priority order based on the failure prediction results and remaining service life assessment of the energy meter; Establish maintenance time windows for different electric energy meters, and arrange maintenance personnel and required spare parts; Set the specific content and execution standards of maintenance tasks, and perform preventive maintenance before the electricity meter fails or reaches its critical service life.

7. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The step S500 specifically includes: After the energy meter is maintained, the energy consumption data of the energy meter is collected again, and compared and analyzed with the energy consumption data before maintenance to evaluate the maintenance effect; Adjust the parameters of the fault prediction model based on the evaluation results to optimize the model performance.

8. The method for scheduling preventive maintenance of electric energy meters according to claim 1, characterized in that: The method also includes recording all energy consumption data, fault prediction results, maintenance scheduling results and performance evaluation results in a database.

9. Energy consumption analysis driven electric energy meter preventive maintenance scheduling platform, characterized by: include: A data acquisition module is used to collect energy consumption data of electric energy meters from the power grid system in real time; A data processing module is used to pre-process historical energy consumption data and real-time energy consumption data; The fault prediction module uses an LSTM-based fault prediction model to predict the fault trend of the electric energy meter and evaluate the remaining service life; Scheduling plan module, which formulates preventive maintenance scheduling plans based on fault prediction results and remaining service life; Maintenance execution module, responsible for implementing the maintenance scheduling plan and recording the maintenance process; The performance evaluation module is used to evaluate the performance of the electric energy meter after maintenance and to optimize the fault prediction model based on the evaluation results.

Images