A smart meter performance evaluation method, system, device and storage medium

Abstract

The present application discloses a method, system, device and storage medium for evaluating the performance of smart meters. The method includes: collecting time series of electric energy data from the total meter and sub-meter in different substations, and pre-processing the collected data to obtain a set of original electric energy data time series; performing cluster analysis on the electric energy data time series in the set of original electric energy data time series; training an evaluation model based on each category of electric energy data time series set, obtaining an electric energy prediction model for the corresponding category and determining a corresponding threshold index; and evaluating the performance of smart meters based on the electric energy prediction model of each category and its corresponding threshold index. The present application is based on a cluster analysis algorithm and combined with a long short-term memory network model to obtain a predicted value for each category of electric energy data curve, and judges the performance of smart meters based on the difference between the predicted value and the actual value, providing technical support for the operation, maintenance and precise replacement of smart meters in substations.

Description

Technical Field

[0001] The present application relates to the technical field of smart meter status evaluation, and in particular to a smart meter performance evaluation method, system, device, and storage medium. Background Art

[0002] In recent years, with the rapid development of smart grids and the widespread deployment of electricity consumption information collection systems, smart meters have been widely adopted. By the first half of 2023, hundreds of millions of smart meters had been installed nationwide, gradually replacing traditional mechanical meters and moving toward digital energy measurement. However, as the number of smart meter installations continues to grow, efficient state estimation, fault detection, and maintenance remain a major challenge.

[0003] To ensure the accuracy of electricity market settlements, State Grid currently replaces smart meters through periodic calibration and rotation. State Grid Corporation of China began a comprehensive rotation of electricity meters in 2009, and since 2013, has used batch failure rates as the basis for this rotation. However, this rotation system presents several challenges. Firstly, some smart meters, despite reaching their service life, may continue to operate without malfunctions, resulting in a waste of resources. Secondly, some meters that have not reached their rotation period may experience sudden failures due to environmental or human factors, significantly increasing both verification and repair costs and social costs.

[0004] Therefore, it is necessary to propose a more accurate method for online monitoring of the operating status of low-voltage smart meters, conduct quantitative analysis of the operating status of meters, and formulate corresponding meter replacement strategies based on evaluation indicators. Summary of the Invention

[0005] In order to solve the problems that the existing periodic rotation method of low-voltage smart meters is not scientific and accurate enough, and the number of metering anomalies is huge, and its management consumes a lot of manpower and material resources, this application proposes a smart meter performance evaluation method, system, equipment and storage medium. This application is based on a cluster analysis algorithm and combines a long-short-term memory network model to obtain the predicted value of each type of electric energy data curve, and judges the performance of the smart meter based on the difference between the predicted value and the actual value, providing technical support for the operation, maintenance and precise replacement of smart meters in the substation.

[0006] In the first aspect, the present application is implemented through the following technical solutions:

[0007] A method for evaluating the performance of a smart meter, the method comprising:

[0008] Collect the time series of electric energy data of the total meter and sub-meter in different substations respectively, and pre-process the collected data to obtain the original electric energy data time series set;

[0009] performing cluster analysis on the electric energy data time series in the original electric energy data time series set;

[0010] Based on the time series set of electric energy data for each category, the evaluation model is trained separately to obtain the electric energy prediction model for the corresponding category and determine the corresponding threshold index;

[0011] Based on the electric energy prediction models of various categories and their corresponding threshold indicators, the performance of smart meters is evaluated.

[0012] In some embodiments, performing cluster analysis on the electric energy data time series in the original electric energy data time series set specifically includes:

[0013] Normalize the time series of electric energy data;

[0014] Construct the minority class cluster:

[0015] Identify minority and majority class samples;

[0016] Randomly selecting a sample from the minority class samples as a cluster center, calculating the Euclidean distance between the cluster center and each sample in the majority class samples, taking the minimum value of the Euclidean distance between the cluster center and each sample in the majority class samples as a threshold, calculating the Euclidean distance between the cluster center and other samples in the minority class samples and obtaining the maximum value of the Euclidean distance within the threshold range;

[0017] Find out whether there are any samples in the minority class whose Euclidean distance to the cluster center is within the threshold range. If so, these samples and the cluster center form a cluster with the cluster center as the center and the average of the threshold and the maximum value as the radius; if not, the cluster center forms a cluster alone, and the cluster radius is equal to half of the threshold;

[0018] Re-determine the minority class samples and the majority class samples, and perform the above steps until the clustering of all power data time series is completed.

[0019] In some embodiments, the evaluation model is trained based on each category of the electric energy data time series set to obtain the electric energy prediction model for the corresponding category and determine the corresponding threshold index, specifically including:

[0020] First, within each category, the cluster center is selected and a related evaluation model is established, wherein the evaluation model adopts a long short-term memory neural network model; then, the evaluation model is trained using training samples of each category. During each training process, a section of data in the time series is used as input to predict one or more future values, and the training error is obtained by subtracting the predicted value from the actual measured value;

[0021] After each training is completed, the root mean square error is used to verify whether the trained model meets the expectations. If not, the training error is fed back to optimize the model parameters and continue with the next training until the expectations are met.

[0022] The trained model is used to predict the power of the test samples, and the error analysis between the predicted values ​​and the actual measured values ​​is performed to determine the threshold indicators for evaluating the power performance.

[0023] In some embodiments, the residual between the predicted power value and the measured power value of the test sample is used as a threshold indicator for evaluating power performance.

[0024] In some embodiments, the evaluation of the performance of the smart meter based on the electric energy prediction models of various categories and their corresponding threshold indicators specifically includes:

[0025] Collect the time series of electric energy data from a smart meter and perform preprocessing;

[0026] Perform cluster analysis on the pre-processed electric energy data time series to determine its category;

[0027] Input the preprocessed electric energy data time series into the electric energy prediction model of the corresponding category to obtain the electric energy prediction value;

[0028] The predicted electric energy value is compared with the measured electric energy value. If the difference between them exceeds the threshold index, it is determined that the operating state of the smart meter is poor; otherwise, it is determined that the operating state of the smart meter is good.

[0029] In a second aspect, the present application proposes a smart meter performance evaluation system, the evaluation system comprising:

[0030] A data preprocessing module is used to preprocess the collected time series of total meter and sub-meter electric energy data of different substations to obtain a time series set of original electric energy data;

[0031] A cluster analysis module, configured to perform cluster analysis on the original electric energy data time series;

[0032] A model training module, which trains an evaluation model based on a time series set of electric energy data for each category, obtains an electric energy prediction model for the corresponding category, and determines a corresponding threshold indicator;

[0033] and a performance evaluation module, which evaluates the performance of the smart meter based on various types of electric energy prediction models and their corresponding threshold indicators.

[0034] In some embodiments, the model training module further includes:

[0035] A construction unit, wherein the construction unit first selects a time series of electric energy data corresponding to a cluster center in each category and establishes a related evaluation model, wherein the evaluation model adopts a long short-term memory neural network model;

[0036] A parameter optimization unit, wherein the parameter optimization unit uses the training samples in each category to train the evaluation model. After each training is completed, the root mean square error is used to verify whether the trained model meets the expectations. If not, the training error of the training is fed back to optimize the model parameters and continue the next training until the expectations are met.

[0037] And, an error analysis unit, which uses the trained model to predict the power of test samples in its corresponding category, performs error analysis on the power prediction value and the power measured value, and determines the threshold index for evaluating the power performance.

[0038] In some embodiments, the performance evaluation module further comprises:

[0039] A prediction unit, wherein the prediction unit inputs the electric energy data time series, which has been preprocessed by the preprocessing module and whose category is determined by the cluster analysis module, into an electric energy prediction model of the corresponding category to obtain an electric energy prediction value;

[0040] and a comparison unit, which compares the electric energy prediction value output by the prediction unit with the electric energy actual measurement value. If the difference exceeds the threshold index, it is determined that the operating state of the smart meter is poor; otherwise, it is determined that the operating state of the smart meter is good.

[0041] In a third aspect, the present application proposes an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0042] In a fourth aspect, the present application proposes a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0043] This application proposes a smart meter performance evaluation method, system, device, and storage medium. This method uses a Euclidean distance-based clustering analysis algorithm to perform cluster analysis on meter data from various sites, obtaining time series of smart meter energy data of different categories (i.e., significantly different). The method then uses the energy data time series of each category to train a long-short-term memory network model to obtain an energy prediction model corresponding to that category. A threshold indicator is set based on the difference between the predicted energy value and the measured energy value, serving as a criterion for determining the performance of the smart meter. This method provides technical support for the operation, maintenance, and precise replacement of smart meters in substations.

[0044] The smart meter performance evaluation method, system, device and storage medium proposed in this application have a wider scope of application and stronger versatility compared to existing status evaluations that are only applicable to one type of meter. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the embodiments of the present application, constitute a part of the present application, and do not constitute a limitation of the embodiments of the present application. In the drawings:

[0046] Figure 1 This is a flow chart of the evaluation method according to an embodiment of the present application;

[0047] Figure 2 This is a functional block diagram of the evaluation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the objectives, technical solutions and advantages of this application more clear, the present application is further described in detail below in conjunction with examples and drawings. The schematic implementation methods of this application and their descriptions are only used to explain this application and are not intended to limit this application.

[0049] Example:

[0050] The existing periodic rotation method of low-voltage smart meters easily leads to resource waste. In addition, the number of metering anomalies is huge, and managing them consumes a lot of manpower and material resources. To address this, this embodiment proposes a method for evaluating the performance of smart meters.

[0051] like Figure 1 As shown, the method proposed in this embodiment specifically includes the following steps:

[0052] Step 100: collecting time series of electric energy data of the master meter and sub-meter in different substations respectively, and pre-processing the data to obtain a set of original electric energy data time series.

[0053] The step 100 specifically includes the following sub-steps:

[0054] Step 101: Collect time series of electric energy data of the master meter and sub-meter in different substations.

[0055] Step 102 , filtering the collected raw electric energy data time series to remove interference and noise such as incompleteness and inconsistency that may exist in the metering data, thereby providing a more accurate and reliable data basis for subsequent clustering of electric energy curves and model training of electric energy meters.

[0056] Step 200 : performing cluster analysis on the electric energy data time series in the original electric energy data time series set.

[0057] The step 200 specifically includes the following sub-steps:

[0058] Step 201: Perform cluster analysis on the power data time series in the original power data time series set to generate a sample data set. The specific process is as follows:

[0059] First, in order to eliminate the dimensional relationship between data variables, the data needs to be normalized and preprocessed to [-1, 1].

[0060] Then, we construct the minority class cluster. Let the data set be X, the data feature dimension be n, and the minority class samples be represented as E={e1,e2,...,e c}, the majority class F={f1,f2,...,f m c is the number of samples in the minority class, and m is the number of samples in the majority class; the abnormal data is the minority class, and the normal data is the majority class. Randomly select a sample from the minority class E as the cluster center e a , and then calculate the following related parameters for discrimination.

[0061] The Euclidean distance between the cluster center and the majority class sample is shown in formula (1).

[0062]

[0063] The threshold χ is the distance between the cluster center and the majority class sample closest to it, as shown in formula (2).

[0064] χ=min{D(e a ,f i )},i∈{1,2,…m} (2)

[0065] The distance between the cluster center and other minority class samples is calculated as shown in formula (3), and then the maximum value of the distance in the interval (0,χ] is obtained by formula (4), which is recorded as d.

[0066]

[0067] The cluster radius R is shown in the following formula (5):

[0068]

[0069] According to formula (1)-formula (5), we can get a a As the center, R is the radius of the minority class cluster. Then, find out whether there are any samples in the remaining minority class samples whose Euclidean distance to the cluster center is less than or equal to the threshold χ. If so, the value of the maximum distance between these minority class samples and the cluster center is recorded as d, thus forming a cluster with the cluster center as the center and the radius R as the average value of χ and d; if not, the cluster center forms a cluster alone, and the radius R is equal to After that, the above steps are repeated until all clusters are formed. To further reduce the training time of the prediction model, this embodiment uses a construction covering algorithm of Euclidean distance to perform cluster analysis on the historical power time series of smart meters based on the similarity of power data curves.

[0070] Step 202: Divide the sample data set generated by clustering into a training sample set and a validation sample set according to a preset ratio. For example, 70% of the sample data is used as the training sample set and 30% is used as the validation sample set.

[0071] Step 300: Based on the time series of each category of power data, an evaluation model is trained to obtain a power forecast model for the corresponding category and determine the corresponding threshold index. This embodiment can use, but is not limited to, a long short-term memory neural network model as the evaluation model. The model is continuously trained based on the time series of power data in the substation to obtain the corresponding power forecast model. The specific process is as follows:

[0072] In step 301, within each category, the power data time series corresponding to the cluster center is first selected and a related evaluation model is established. This embodiment can employ, but is not limited to, a long short-term memory neural network model as the evaluation model. The evaluation model is trained using training sample data from each category. During model training, a segment of time series data is used as input, and one or more values ​​are predicted ahead of the window. Each predicted value is subtracted from the actual power measurement value to obtain the training error, completing a complete training cycle.

[0073] Step 302: After each training is completed, the root mean square error (RMSE) shown in formula (6) is used to verify whether the trained model meets the expectations. If not, the training error is fed back to optimize the model parameters and continue with the next training until the training meets the expectations.

[0074]

[0075] In the formula, num represents the number of training samples; yi is the actual value of the i-th data; The model prediction value for the i-th data.

[0076] Step 303: Use the trained model to predict the power consumption of the test sample set, perform error analysis between the predicted value and the actual measured value, and determine the threshold index for evaluating power performance. This embodiment assumes that the power meter is operating well during the model establishment process, so the residual corresponding to the test sample set is within the normal range, which is used as the threshold of the evaluation model. The calculation formula for the prediction residual is as follows:

[0077] error=W p -W r (7)

[0078] Where: error is the residual of electric energy data; W p is the electric energy data predicted by the model; W r The actual electric energy data of the smart meter.

[0079] Step 400 : Evaluate the performance of the smart meter based on the various types of electric energy prediction models and their corresponding threshold indicators.

[0080] The step 400 specifically includes the following steps:

[0081] Step 401: collecting the time series of electric energy data of a smart meter and preprocessing it;

[0082] Step 402: Perform cluster analysis on the pre-processed electric energy data time series to determine the category to which it belongs.

[0083] Step 403: input the pre-processed electric energy data time series into the electric energy prediction model of the corresponding category to obtain the electric energy prediction value.

[0084] Step 404 : Compare the predicted power value with the measured power value. If the difference exceeds a threshold, it is determined that the operating state of the smart meter is poor; otherwise, it is determined that the operating state of the smart meter is good.

[0085] The evaluation method proposed in this embodiment is based on the Euclidean distance cluster analysis algorithm. By performing cluster analysis on the data measured by the smart meters at each site, the electric energy data time series of several types of smart meters with obvious differences are obtained. Then, the long short-term memory network model is used to train the electric energy data time series of each type respectively, so as to obtain the predicted value of each type of electric energy data time series. Finally, based on the data obtained in the actual substation, the threshold is set by using the difference between the predicted electric energy value and the actual value to determine the evaluation index of the smart meter performance. This provides technical support for the operation, maintenance and precise replacement of smart meters in the substation, and improves the accuracy of fault analysis and positioning.

[0086] Based on the same technical concept as above, this embodiment also proposes a smart meter performance evaluation system, such as Figure 2 As shown, the evaluation system proposed in this embodiment includes:

[0087] The data preprocessing module is used to preprocess the collected time series of energy data from the master and sub-meters of different substations to obtain a set of raw energy data time series. This data preprocessing module can also be used to generate the energy data time series of smart meters for later performance evaluation. Specifically, a filtering algorithm can be used to process the collected energy data time series to remove interference and noise in the metered data and improve data reliability. This is described in detail in steps 101-102 above and will not be further elaborated here.

[0088] The cluster analysis module is used to perform cluster analysis on the original power data time series. The specific clustering process is described in steps 201-202 above and will not be further elaborated here. This cluster analysis can also perform adaptive clustering on newly collected power data time series, that is, identify the categories of the power data time series.

[0089] The model training module trains the evaluation model based on the time series set of electric energy data of each category, obtains the electric energy prediction model of the corresponding category and determines the corresponding threshold index.

[0090] and a performance evaluation module, which evaluates the performance of smart meters based on various types of electric energy prediction models and their corresponding threshold indicators.

[0091] Furthermore, the model training module also includes:

[0092] The construction unit first selects the time series of electric energy data corresponding to the cluster center in each category and establishes a related evaluation model. The long short-term memory neural network model is used as the evaluation model.

[0093] The parameter optimization unit uses the training samples in each category to train the evaluation model. After each training is completed, the root mean square error is used to verify whether the trained model meets the expectations. If so, the training is completed. Otherwise, the training error is fed back to optimize the model parameters and continue with the next training.

[0094] And, an error analysis unit, which uses the trained model to predict the power of the test sample, performs error analysis between the predicted value and the actual measured value, and determines the threshold index for evaluating the power performance.

[0095] Furthermore, the performance evaluation module also includes:

[0096] The prediction unit inputs the electric energy data time series processed by the preprocessing module and classified by the clustering analysis module into the electric energy prediction model of the corresponding category to obtain the electric energy prediction value.

[0097] and a comparison unit, which compares the electric energy prediction value output by the prediction unit with the electric energy actual measurement value. If the difference exceeds a threshold indicator, it is determined that the operating state of the smart meter is poor; otherwise, it is determined that the operating state of the smart meter is good.

[0098] This embodiment uses actual area data to verify the above evaluation method proposed in this embodiment, as follows:

[0099] To address the problem of estimating smart grid operation errors in a typical distribution network in a specific region, the evaluation method proposed in this embodiment was used to evaluate two types of smart energy meters: AFPM and PZ. The distribution network was equipped with one master smart energy meter and 125 sub-smart energy meters. The training data used historical smart energy meter operation data from April to June 2022, with a sampling frequency of 15 minutes. The sample test data used smart grid operation input data from May 1 to 15, 2022. The error estimation results are shown in Table 1.

[0100] Table 1. Estimation results of operation errors of smart energy meters in distribution network substations

[0101]

[0102] As can be seen from Table 1, the evaluation method proposed in this embodiment is effective and provides technical support for operation and maintenance detection and fault diagnosis.

[0103] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0104] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0105] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0107] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of this application. It should be understood that the above description is only the specific implementation methods of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application should be included in the scope of protection of this application.

Claims

1. A method for evaluating the performance of a smart meter, characterized in that: The evaluation method includes: Collect the time series of electric energy data of the total meter and sub-meter in different substations respectively, and pre-process the collected data to obtain the original electric energy data time series set; performing cluster analysis on the electric energy data time series in the original electric energy data time series set; Based on the time series set of electric energy data for each category, the evaluation model is trained separately to obtain the electric energy prediction model for the corresponding category and determine the corresponding threshold index; Based on the electric energy prediction models of various categories and their corresponding threshold indicators, the performance of the smart meter is evaluated; the cluster analysis of the electric energy data time series in the original electric energy data time series set specifically includes: Normalize the time series of electric energy data; Construct the minority class cluster: Identify minority and majority class samples; Randomly selecting a sample from the minority class samples as a cluster center, calculating the Euclidean distance between the cluster center and each sample in the majority class samples, taking the minimum value of the Euclidean distance between the cluster center and each sample in the majority class samples as a threshold, calculating the Euclidean distance between the cluster center and other samples in the minority class samples and obtaining the maximum value of the Euclidean distance within the threshold range; Find out whether there are any samples in the minority class whose Euclidean distance to the cluster center is within the threshold range. If so, these samples and the cluster center form a cluster with the cluster center as the center and the average of the threshold and the maximum value as the radius; if not, the cluster center forms a cluster alone, and the cluster radius is equal to half of the threshold; Re-determine the minority class samples and the majority class samples, and perform the above steps until the clustering of all power data time series is completed.

2. A method for evaluating the performance of a smart meter according to claim 1, characterized in that: The aforementioned evaluation model is trained based on each category of electric energy data time series set, to obtain the electric energy prediction model of the corresponding category and determine the corresponding threshold index, specifically including: First, within each category, the cluster center is selected and a related evaluation model is established, wherein the evaluation model adopts a long short-term memory neural network model; then, the evaluation model is trained using training samples of each category. During each training process, a section of data in the time series is used as input to predict one or more future values, and the training error is obtained by subtracting the predicted value from the actual measured value; After each training is completed, the root mean square error is used to verify whether the trained model meets the expectations. If not, the training error is fed back to optimize the model parameters and continue with the next training until the expectations are met. The trained model is used to predict the power of the test samples, and the error analysis between the predicted values ​​and the actual measured values ​​is performed to determine the threshold indicators for evaluating the power performance.

3. A method for evaluating the performance of a smart meter according to claim 2, characterized in that: The residual between the predicted power value and the measured power value of the test sample is used as the threshold indicator for evaluating power performance.

4. A method for evaluating the performance of a smart meter according to any one of claims 1 to 3, characterized in that: The aforementioned evaluation of the performance of smart meters based on the electric energy prediction models of various categories and their corresponding threshold indicators specifically includes: Collect the time series of electric energy data from a smart meter and perform preprocessing; Perform cluster analysis on the pre-processed electric energy data time series to determine its category; Input the preprocessed electric energy data time series into the electric energy prediction model of the corresponding category to obtain the electric energy prediction value; The predicted electric energy value is compared with the measured electric energy value. If the difference between them exceeds the threshold index, it is determined that the operating state of the smart meter is poor; otherwise, it is determined that the operating state of the smart meter is good.

5. A smart meter performance evaluation system, characterized in that: The evaluation system comprises: A data preprocessing module is used to preprocess the collected time series of total meter and sub-meter electric energy data of different substations to obtain a time series set of original electric energy data; A cluster analysis module, configured to perform cluster analysis on the original electric energy data time series; A model training module, which trains an evaluation model based on a time series set of electric energy data for each category, obtains an electric energy prediction model for the corresponding category, and determines a corresponding threshold indicator; and a performance evaluation module, which evaluates the performance of the smart meter based on various types of power prediction models and their corresponding threshold indicators; The cluster analysis of the original electric energy data time series specifically includes: Normalize the time series of electric energy data; Construct the minority class cluster: Identify minority and majority class samples; Randomly selecting a sample from the minority class samples as a cluster center, calculating the Euclidean distance between the cluster center and each sample in the majority class samples, taking the minimum value of the Euclidean distance between the cluster center and each sample in the majority class samples as a threshold, calculating the Euclidean distance between the cluster center and other samples in the minority class samples and obtaining the maximum value of the Euclidean distance within the threshold range; Find out whether there are samples in the minority class whose Euclidean distance to the cluster center is within the threshold range. If so, these samples and the cluster center are combined to form a cluster with the cluster center as the center and the radius as the average of the threshold and the maximum value; if not, the cluster center forms a cluster alone, and the cluster radius is equal to half of the threshold; Re-determine the minority class samples and the majority class samples, and perform the above steps until the clustering of all power data time series is completed.

6. A smart meter performance evaluation system according to claim 5, characterized in that: The model training module also includes: A construction unit, wherein the construction unit first selects a time series of electric energy data corresponding to a cluster center in each category and establishes a related evaluation model, wherein the evaluation model adopts a long short-term memory neural network model; A parameter optimization unit, wherein the parameter optimization unit uses the training samples in each category to train the evaluation model. After each training is completed, the root mean square error is used to verify whether the trained model meets the expectations. If not, the training error of the training is fed back to optimize the model parameters and continue the next training until the expectations are met. And, an error analysis unit, which uses the trained model to predict the power of test samples in its corresponding category, performs error analysis on the power prediction value and the power measured value, and determines the threshold index for evaluating the power performance.

7. The smart meter performance evaluation system according to claim 5, characterized in that: The performance evaluation module also includes: A prediction unit, wherein the prediction unit inputs the electric energy data time series, which has been preprocessed by the preprocessing module and whose category is determined by the cluster analysis module, into an electric energy prediction model of the corresponding category to obtain an electric energy prediction value; and a comparison unit, which compares the electric energy prediction value output by the prediction unit with the electric energy actual measurement value. If the difference exceeds the threshold index, it is determined that the operating state of the smart meter is poor; otherwise, it is determined that the operating state of the smart meter is good.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images