Metering Verification Method and System for Grouped Power Users

Abstract

The present invention discloses a metering verification method and system for grouped power users, which relates to the technical field of power metering. The method includes: obtaining the power energy types of each power user in the power user group; grouping the power user group to output multiple power user groups; constructing corresponding power metering models according to the multiple power user groups to output multiple power metering models; performing ensemble learning on the multiple power metering models to obtain an ensemble power metering model, optimizing the model parameters of the multiple power metering models, and outputting the optimized multiple power metering models; collecting multiple real-time power operation data sets corresponding to the multiple power user groups; mapping and inputting the multiple real-time power operation data sets into the optimized multiple power metering models for verification, and outputting a metering verification result. The technical problems of low efficiency and insufficient accuracy of power user metering verification in the prior art are solved, and the technical effects of improving metering verification efficiency and accuracy are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power metering, and particularly to a metering verification method and system for grouped power users. Background Art

[0002] With the transformation of the global energy structure and the rapid development of smart grid technology, power metering verification, as a key link to ensure the stable operation of the power system, protect the rights and interests of power users, and promote the efficient use of energy, has become increasingly important. Traditional power metering verification methods mostly rely on manual on-site meter reading and periodic calibration, which are not only inefficient but also difficult to reflect the actual power consumption of power users in real time. Especially in the context of the increasing access of distributed energy, diversified power transactions, and the surging demand for user-side energy management, traditional methods have been unable to meet the requirements of modern power system management. Summary of the Invention

[0003] This application provides a metering verification method and system for grouped power users, which solves the technical problems of low efficiency and insufficient accuracy in power user metering verification in the prior art.

[0004] In view of the above problems, this application provides a metering verification method and system for grouped power users.

[0005] In the first aspect of this application, a metering verification method for grouped power users is provided. The method includes:

[0006] Obtain the power energy types of each power user in the power user group; group the power user group according to the power energy types, and output multiple power user groups, where each user group has the same energy type; construct corresponding power metering models according to the multiple power user groups, and output multiple power metering models; perform ensemble learning on the multiple power metering models to obtain an ensemble power metering model, and optimize the model parameters of the multiple power metering models according to the model parameters of the ensemble power metering model, and output the optimized multiple power metering models; collect multiple real-time power operation data sets corresponding to the multiple power user groups; map and input the multiple real-time power operation data sets into the optimized multiple power metering models for verification, and output the metering verification results.

[0007] In the second aspect of this application, a metering verification system for grouped power users is provided. The system includes:

[0008] Energy type acquisition module, which is used to acquire the power energy types of each power user in the power user group; clustering module, which is used to cluster the power user group according to the power energy type and output multiple power user groups, where each user group has the same energy type; model construction module, which is used to construct corresponding power metering models according to the multiple power user groups and output multiple power metering models; optimization module, which is used to perform ensemble learning on the multiple power metering models, obtain an integrated power metering model, optimize the model parameters of the multiple power metering models according to the model parameters of the integrated power metering model, and output the optimized multiple power metering models; data acquisition module, which is used to acquire multiple real-time power operation data sets corresponding to the multiple power user groups; verification module, which is used to map and input the multiple real-time power operation data sets into the optimized multiple power metering models for verification and output a metering verification result.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] First, acquire the power energy types of each power user in the power user group. Next, cluster the power user group according to the power energy type and output multiple power user groups, where each user group has the same energy type. Further, construct corresponding power metering models according to the multiple power user groups and output multiple power metering models. Then, perform ensemble learning on the multiple power metering models, obtain an integrated power metering model, optimize the model parameters of the multiple power metering models according to the model parameters of the integrated power metering model, and output the optimized multiple power metering models. Finally, acquire multiple real-time power operation data sets corresponding to the multiple power user groups; map and input the multiple real-time power operation data sets into the optimized multiple power metering models for verification and output a metering verification result. This solves the technical problems of low efficiency and insufficient accuracy in power user metering verification in the prior art and achieves the technical effect of improving the efficiency and accuracy of metering verification. Description of the Drawings

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 It is a schematic flowchart of a metering verification method for clustering power users provided in an embodiment of this application.

[0013] Figure 2 This is a schematic structural diagram of the metering verification system for grouped power users provided by an embodiment of the present application.

[0014] Explanation of reference numerals: Energy type acquisition module 11, grouping module 12, model construction module 13, optimization module 14, data acquisition module 15, verification module 16. Specific implementation manners

[0015] By providing a metering verification method and system for grouped power users, the present application solves the technical problems of low efficiency and insufficient accuracy in metering verification of power users in the prior art.

[0016] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0017] It should be noted that the terms "include" and "have" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0018] Embodiment 1, as Figure 1 shown, the present application provides a metering verification method for grouped power users, wherein the method includes:

[0019] Obtain the power energy types of each power user in the power user group.

[0020] In a power user group, the power energy types of each power user may vary due to various factors such as geographical location, energy resources, and technical conditions. By querying the data of the power company, obtain the power energy types of each power user in the power user group, and the power energy types include thermal power, wind energy, solar energy, etc.

[0021] Group the power user group according to the power energy type, and output multiple power user groups, wherein each user group has users with the same energy type.

[0022] Classify users according to the pre-identified power energy types (such as thermal power, wind power, solar energy, etc.), establish an independent user group list for each energy type, and initially divide multiple power user groups to ensure that users within each group share the same energy type.

[0023] Furthermore, clustering the power user groups according to the power energy type and outputting multiple power user groups, the method includes:

[0024] Performing a first clustering of the power user groups according to the power energy type, and outputting a first type of power user group and a second type of power user group, wherein the first type of power user group is a power user group with a single energy label, and the second type of power user group is a power user group with a mixed energy label; clustering the first type of power user group according to the power energy type, and outputting multiple single power user groups; obtaining the number of mixed items, and clustering the second type of power user group in ascending order according to the number of mixed items, and outputting multiple mixed power user groups; using the multiple single power user groups and the multiple mixed power user groups to output multiple power user groups.

[0025] Specifically, according to the energy labels of power users, perform a preliminary clustering of the power user groups, and divide the user groups into a power user group with a single energy label (the first type of power user group) and a power user group with a mixed energy label (the second type of power user group). The first type of power user group means that each user only uses one energy, such as thermal power or wind energy. The second type of power user group means that the user uses a combination of multiple energies, such as a mixture of thermal power and solar energy. For the first type of power user group, perform a more refined clustering according to the specific single energy type of each user (such as thermal power, wind energy, solar energy), and output multiple power user groups of a single energy type (i.e., multiple single power user groups). For example, classify thermal power users into the thermal power user group, and wind energy users into the wind energy user group, etc. For the second type of power user group, first determine the number of energy types included in the mixed energy label of each user, that is, the number of mixed items; according to the number of mixed items, perform clustering in ascending order to form multiple mixed power user groups with different numbers of mixed items. For example, users with 2 mixed items (such as a mixture of thermal power and wind energy) are classified into one group, and users with 3 mixed items (such as a mixture of thermal power, wind energy, and solar energy) are classified into another group. Combine and output the multiple single power user groups and multiple mixed power user groups obtained before to obtain the complete clustering result. These user groups respectively reflect the characteristics of power users with single and mixed energy labels, providing a data basis for the construction of subsequent power measurement models.

[0026] Construct corresponding power measurement models according to the multiple power user groups, and output multiple power measurement models.

[0027] Select a suitable metering model architecture according to the energy characteristics of each power user group. For user groups with a single energy type, a relatively simple metering model can be selected, such as a univariate model based on the energy consumption curve; while for user groups with a mixed energy type, a more complex multivariate model is required to meet the metering needs of multiple energy sources. Use the historical power data of each user group to train the corresponding metering model; for the model of a single energy type user group, the typical power consumption data of this user group can be used for univariate training to quickly establish the model; for the model of a mixed energy type user group, the relevant data of each energy type are used for joint training to ensure that the model can accurately measure and predict the power consumption of the multi-energy combination. Export the trained metering models of each power user group as independent model instances; each single power user group corresponds to a single energy metering model for monitoring the energy consumption of single energy users; each mixed power user group corresponds to a mixed energy metering model for accurately measuring and predicting the energy consumption of multi-energy combination users.

[0028] Furthermore, construct corresponding power metering models according to the multiple power user groups, and output multiple power metering models. The multiple power metering models include multiple single power metering models. Among them, training a single power metering model includes:

[0029] Collect a single historical power operation dataset sample of the corresponding user group; extract features from the single historical power operation dataset sample and output a single fused power operation feature vector; obtain an initial time series model according to the power energy type corresponding to each single power metering model; train the initial time series model through the single fused power operation feature vector and output multiple single power metering models.

[0030] Preferably, historical power operation data is collected from corresponding single power user groups. These data samples include the power consumption situations of these user groups at different time periods. A single historical power operation data set may cover information such as energy consumption peaks, power consumption time distributions, power factors, etc., comprehensively reflecting the power consumption characteristics of users of a single energy type. Feature extraction is performed on the collected single historical power operation data set samples, including extracting key factors such as time, load, power, etc., to generate a single fused power operation feature vector. According to the power energy type corresponding to each single power metering model, a suitable time series model is selected and initialized, such as ARIMA, LSTM, RNN, etc., to fully capture the time dependence relationship and fluctuation law of a single energy type. The single fused power operation feature vector is input into the initialized time series model for training, and the model learns the relationships between various features and the time evolution trend. Through the training of the feature vector, the time series model can accurately predict the energy consumption pattern of users of this single energy type, generating a trained single power metering model. After the training is completed, multiple single power metering models are output, and each model is for a user group of a single energy type.

[0031] Furthermore, corresponding power metering models are constructed according to the multiple power user groups, and multiple power metering models are output. The multiple power metering models further include multiple hybrid power metering models. Among them, training the hybrid power metering model includes:

[0032] Collecting hybrid-historical power operation data set samples and hybrid energy ratios of the corresponding user groups; performing data fusion on the hybrid-historical power operation data set samples to output hybrid-fused power operation data set samples; performing hybrid feature extraction according to the hybrid-fused power operation data set samples to output hybrid-fused power operation feature vectors; initializing a multi-layer time series model, and training the hybrid-fused power operation feature vectors according to the multi-layer time series model to output multiple hybrid power metering models.

[0033] Preferably, for each hybrid energy user group, historical power operation data samples and the hybrid energy ratio of the user group (for example, thermal power accounts for 30%, wind power accounts for 50%, and solar energy accounts for 20%) are collected; the collected hybrid-historical power operation data set samples are subjected to data fusion and weighted and combined according to the proportions of each energy type to construct a complete hybrid-fused power operation data set sample. The hybrid-fused power operation data set sample integrates the power consumption characteristics of different energy sources and forms a hybrid data set with multi-energy characteristics, providing a multi-level data basis for feature extraction. Hybrid features (such as load response characteristics, time variation trends, energy switching frequencies, etc.) are extracted from the hybrid-fused power operation data set to generate a hybrid-fused power operation feature vector. According to the multi-energy characteristics of hybrid power users, a multi-layer time series model, such as a multi-layer LSTM or GRU structure, is selected and initialized to capture the temporal correlation and hierarchical relationship between different energy types; the hybrid-fused power operation feature vector is input into the multi-layer time series model for training so that the model can learn the interaction between different energy characteristics and the overall energy consumption trend; after training is completed, multiple hybrid power metering models are output, and each model is optimized for a specific hybrid energy user group and can accurately reflect the multi-energy characteristics of the user group, providing data support for subsequent real-time power metering and verification.

[0034] Furthermore, according to the multiple power user groups, corresponding power metering models are constructed, and multiple power metering models are output. The method further includes:

[0035] Identifying the user group quantity index in the multiple power user groups; when the user group quantity index is less than a preset user group quantity index, it is marked, and the marked power user group is output; activating a generative adversarial network, and obtaining a generated power operation data set sample corresponding to the marked power user group according to the generative adversarial network; and performing a power metering model according to the historical power operation data set sample corresponding to the marked power user group and the generated power operation data set sample.

[0036] Preferably, count the quantity index of the power user groups, that is, the number of users included in each power user group; when the quantity index of a certain power user group is lower than the preset user group quantity threshold, identify this power user group as an identified power user group. The sample data of these user groups is less, which may lead to insufficient model training. Therefore, data needs to be supplemented; use a generative adversarial network (GAN) to generate additional data samples for the identified power user group; GAN consists of a generator and a discriminator. The generator generates realistic power operation data samples, and the discriminator continuously adjusts the output of the generator, making the generated samples gradually approach the characteristics of real samples. The samples generated by GAN can simulate the electricity consumption characteristics of the identified power user group, providing sufficient data support for subsequent model training; the power operation data samples generated by GAN supplement the historical data set of the identified power user group. The data set corresponding to the identified power user group includes its original historical power operation data set samples and the power operation data set samples generated by GAN; use the expanded data set samples to train the power metering model of the identified power user group; the multiple trained power metering models include models optimized for the identified power user groups with insufficient sample data, ensuring that the metering accuracy of all user groups meets the standard requirements. Through the generative adversarial network (GAN), the user groups with insufficient data are effectively supplemented, enabling the power metering model to still achieve high-precision power metering verification in the case of insufficient samples.

[0037] Perform ensemble learning on the multiple power metering models to obtain an ensemble power metering model, and optimize the model parameters of the multiple power metering models according to the model parameters of the ensemble power metering model, and output the optimized multiple power metering models.

[0038] Optimize multiple power metering models through ensemble learning to improve the accuracy and generalization ability of the overall model. Specifically, construct an ensemble learning framework and incorporate multiple power metering models (including single power metering models and hybrid power metering models) into the ensemble learning. Commonly used ensemble learning methods such as Bagging, Boosting, or Stacking can be selected to effectively integrate the advantages of each model; integrate each power metering model, and measure the performance of each model on different user groups through weight assignment. Among them, the weight assignment is based on indicators such as the metering error, prediction stability, and applicability of each model to ensure that the models with better performance have a greater weight in the integration; generate an integrated power metering model according to the ensemble learning strategy, enabling it to comprehensively combine the prediction results of multiple models, so as to achieve higher-precision metering verification for different types and characteristics of power user groups; generate an integrated power metering model according to the ensemble learning strategy, enabling it to comprehensively combine the prediction results of multiple models, so as to achieve higher-precision metering verification for different types and characteristics of power user groups; use the parameters of the integrated power metering model as feedback information to optimize and adjust the parameters of each power metering model. By comparing the error differences between the integrated model and the single model, optimize the key parameters in each model to make it more conform to the real electricity consumption characteristics of the user group; after completing the parameter optimization, output the optimized power metering models, including single power metering models and hybrid power metering models.

[0039] Furthermore, optimize the model parameters of the multiple power metering models according to the model parameters of the integrated power metering model. Among them, the model parameter update formula for each power metering model is:

[0040] ; where The model parameters of the i-th power metering model at the (t + 1)-th iteration are the model parameters of the i-th power metering model at the t-th iteration, is the learning rate, which is used to control the speed of model parameter update, is the gradient of the loss function L with respect to the model parameter , reflecting the sensitivity of each model parameter to the error; through repeated iteration until convergence, output the optimized multiple power metering models.

[0041] Preferably, set the loss function L, usually the prediction error of the model (such as mean square error, cross entropy, etc.), to evaluate the current prediction effect of the model; calculate the gradient of the loss function with respect to each model parameter , that is, the rate of change of the loss function with respect to each parameter. The gradient value reflects the impact of parameter changes on the error and guides the direction of parameter update. Using the parameter update formula, the parameters of each power metering model are iteratively updated, where, The model parameters of the i-th power metering model at the (t + 1)-th iteration, are the model parameters of the i-th power metering model at the t-th iteration, is the learning rate, which is used to control the speed of model parameter update, is the gradient of the loss function L with respect to the model parameters , reflecting the sensitivity of each model parameter to the error. Through continuous iteration until the loss function L converges, that is, the parameter update amplitude is small enough to be negligible, indicating that the model has reached the optimal state. At this time, the sensitivity of the model parameters to the error approaches zero and no longer significantly affects the prediction results. When the loss function reaches the minimum value and converges, the optimized multiple power metering models are output. These model parameters have been effectively optimized by the feedback of the integrated power metering model, and the accuracy and robustness have been significantly improved.

[0042] Collect multiple real-time power operation data sets corresponding to the multiple power user groups.

[0043] Establish a real-time data acquisition interface for each power user group. The interface can be connected to smart meters, sensor networks or energy management systems to collect the power operation data of each user group in real time, so as to ensure that the latest power operation data is used in the verification process.

[0044] Map and input the multiple real-time power operation data sets into the optimized multiple power metering models for verification, and output the metering verification results.

[0045] Map the real-time power operation data set of each power user group into the corresponding optimized power metering model. The mapping is matched according to the energy type (single energy or mixed energy) and model type (single power metering model or mixed power metering model) of the user group to ensure that the data is accurately input into the corresponding model. Input the mapped real-time data into the corresponding power metering model. The model analyzes the data through calculation, including measuring the characteristics such as the real-time power consumption, load fluctuation, and power factor of the user, and finally outputs the metering verification results of each power user group, including the actual power consumption, metering error, and verification status of each user group.

[0046] Furthermore, mapping and inputting the multiple real-time power operation data sets into the optimized multiple power metering models for verification and outputting the metering verification results, the method includes:

[0047] Map the input of the multiple real-time power operation data sets into multiple optimized power metering models. Each real-time power operation data set includes the power operation data of each power user in the same user group. Perform predictions according to the multiple power metering models and output the predicted power indicators of each power user. Connect to the terminal of the power metering log to obtain the recorded power indicators of each power user. Compare the predicted power indicators with the recorded power indicators, identify users with a difference greater than the preset indicator difference as abnormal users, and output the metering verification result.

[0048] Specifically, map the real-time power operation data set of each power user group into the corresponding optimized power metering model. Use the optimized power metering model to process the real-time data to generate the predicted power indicators of each power user. The predicted power indicators include the predicted real-time power consumption, load conditions, etc., reflecting the model's expectations for each user's power consumption behavior. Connect to the log terminal of the power metering system to obtain the recorded power indicators of each power user. The recorded power indicators are the actually measured power consumption data, usually updated in real time by smart meters or sensors, serving as the actual basis for metering. Compare the predicted power indicators of each user with their recorded power indicators item by item, calculate the difference between the two, and judge whether there is a significant deviation by setting a preset indicator difference threshold. If the difference is less than the preset threshold, the metering result is considered normal. If the difference is greater than the preset threshold, it indicates metering abnormality. Users with a difference exceeding the preset threshold in the comparison result will be marked as abnormal users for further verification and tracking. Summarize the comparison results of all users to generate the metering verification result. The metering verification result includes the predicted power indicators, recorded power indicators, deviation analysis of each user, and the identification information of abnormal users.

[0049] In summary, the embodiments of the present application have at least the following technical effects:

[0050] First, obtain the power energy types of each power user in the power user group. Then, divide the power user group according to the power energy types to output multiple power user groups. Each user group has users with the same energy type. Further, construct corresponding power metering models according to the multiple power user groups to output multiple power metering models. Then, perform ensemble learning on the multiple power metering models to obtain an ensemble power metering model, and optimize the model parameters of the multiple power metering models according to the model parameters of the ensemble power metering model to output multiple optimized power metering models. Finally, collect multiple real-time power operation data sets corresponding to the multiple power user groups; map the input of the multiple real-time power operation data sets into the multiple optimized power metering models for verification and output the metering verification result. This solves the technical problems of low efficiency and insufficient accuracy in power user metering verification in the prior art, and achieves the technical effects of improving metering verification efficiency and accuracy.

[0051] Embodiment 2. Based on the same inventive concept as the metering verification method for grouped power users in the foregoing embodiment, as Figure 2 shown, the present application provides a metering verification system for grouped power users, wherein the system includes:

[0052] An energy type acquisition module 11, which is configured to acquire the power energy types of each power user in the power user group; a grouping module 12, which is configured to group the power user group according to the power energy types and output a plurality of power user groups, wherein each user group has users with the same energy type; a model construction module 13, which is configured to construct corresponding power metering models according to the plurality of power user groups and output a plurality of power metering models; an optimization module 14, which is configured to perform ensemble learning on the plurality of power metering models, obtain an integrated power metering model, optimize the model parameters of the plurality of power metering models according to the model parameters of the integrated power metering model, and output the optimized plurality of power metering models; a data acquisition module 15, which is configured to acquire a plurality of real-time power operation data sets corresponding to the plurality of power user groups; a verification module 16, which is configured to map and input the plurality of real-time power operation data sets into the optimized plurality of power metering models for verification and output a metering verification result.

[0053] Further, the verification module 16 is configured to execute the following method:

[0054] Map and input the plurality of real-time power operation data sets into the optimized plurality of power metering models, wherein each real-time power operation data set includes the power operation data of each power user in the same user group; perform prediction according to the plurality of power metering models and output the predicted power indexes of each power user; connect to the terminal of the power metering log to obtain the recorded power indexes of each power user; compare the predicted power indexes with the recorded power indexes, identify the users with a difference greater than the preset index difference as abnormal users, and output the metering verification result.

[0055] Further, the grouping module 12 is configured to execute the following method:

[0056] Cluster the power user groups once according to the power energy type, and output the first type of power user group and the second type of power user group. Among them, the first type of power user group is a power user group with a single energy label, and the second type of power user group is a power user group with a mixed energy label; cluster the first type of power user group according to the power energy type, and output multiple single power user groups; obtain the number of mixed items, and cluster the second type of power user group in ascending order according to the number of mixed items, and output multiple mixed power user groups; use the multiple single power user groups and the multiple mixed power user groups to output multiple power user groups.

[0057] Further, the model construction module 13 is used to execute the following method:

[0058] Collect the single-historical power operation dataset samples of the corresponding user group; perform feature extraction on the single-historical power operation dataset samples, and output single-fused power operation feature vectors; obtain the initialized time series model according to the power energy type corresponding to each single power measurement model; train the initialized time series model through the single-fused power operation feature vectors, and output multiple single power measurement models.

[0059] Further, the model construction module 13 is used to execute the following method:

[0060] Collect the mixed-historical power operation dataset samples and the mixed energy ratio of the corresponding user group; perform data fusion on the mixed-historical power operation dataset samples, and output mixed-fused power operation dataset samples; perform mixed feature extraction according to the mixed-fused power operation dataset samples, and output mixed-fused power operation feature vectors; initialize the multi-layer time series model, and train the mixed-fused power operation feature vectors according to the multi-layer time series model, and output multiple mixed power measurement models.

[0061] Further, the optimization module 14 is used to execute the following method:

[0062] The model parameter update formula for each power measurement model is: ; where The model parameters of the i-th power measurement model at the (t + 1)-th iteration, are the model parameters of the i-th power measurement model at the t-th iteration, is the learning rate, which is used to control the speed of model parameter update, is the gradient of the loss function L with respect to the model parameter , reflecting the sensitivity of each model parameter to the error; through repeated iteration until convergence, output the optimized multiple power measurement models.

[0063] Further, the model construction module 13 is used to execute the following method:

[0064] Identify the user group quantity index in the multiple power user groups; when the user group quantity index is less than the preset user group quantity index, perform identification and output the identified power user group; activate the generative adversarial network, and obtain the generative power operation data set sample corresponding to the identified power user group according to the generative adversarial network; perform a power metering model based on the historical power operation data set sample corresponding to the identified power user group and the generative power operation data set sample.

[0065] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the advantages or disadvantages of the embodiments. And the above description of specific embodiments of this specification has been made. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0066] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

[0067] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A metering verification method for grouped power users, characterized in that The method includes: Obtaining the power energy types of each power user in the power user group; Grouping the power user group according to the power energy type, and outputting multiple power user groups, where each user group has the same energy type; Constructing corresponding power metering models according to the multiple power user groups, and outputting multiple power metering models; Performing ensemble learning on the multiple power metering models to obtain an ensemble power metering model, and optimizing the model parameters of the multiple power metering models according to the model parameters of the ensemble power metering model, and outputting the optimized multiple power metering models; Collecting multiple real-time power operation data sets corresponding to the multiple power user groups; Mapping and inputting the multiple real-time power operation data sets into the optimized multiple power metering models for verification, and outputting a metering verification result; Grouping the power user group according to the power energy type, and outputting multiple power user groups, the method includes: Performing a first grouping on the power user group according to the power energy type, and outputting a first type of power user group and a second type of power user group, where the first type of power user group is a power user group with a single energy label, and the second type of power user group is a power user group with a mixed energy label; Grouping the first type of power user group according to the power energy type, and outputting multiple single power user groups; Obtaining the number of mixed items, and grouping the second type of power user group in ascending order of the number of mixed items, and outputting multiple mixed power user groups; Outputting multiple power user groups with the multiple single power user groups and the multiple mixed power user groups; Constructing corresponding power metering models according to the multiple power user groups, and outputting multiple power metering models, the multiple power metering models include multiple single power metering models, where training a single power metering model includes: Collecting a single-historical power operation data set sample of the corresponding user group; Performing feature extraction on the single-historical power operation data set sample, and outputting a single-fused power operation feature vector; Obtaining an initial time series model according to the power energy type corresponding to each single power metering model; Training the initial time series model through the single-fused power operation feature vector, and outputting multiple single power metering models.

2. The metering verification method for grouped power users according to claim 1, wherein Mapping and inputting the multiple real-time power operation data sets into the optimized multiple power metering models for verification, and outputting a metering verification result, the method includes: Mapping and inputting the multiple real-time power operation data sets into the optimized multiple power metering models, where each real-time power operation data set includes the power operation data of each power user in the same user group; Performing prediction according to the multiple power metering models, and outputting the predicted power indicators of each power user; Connecting to the terminal of the power metering log, and obtaining the recorded power indicators of each power user; Comparing the predicted power indicators with the recorded power indicators, and identifying the users with a difference greater than the preset indicator difference as abnormal users, and outputting a metering verification result.

3. The metering verification method for grouped power users according to claim 1, characterized in that, Construct corresponding power metering models according to the multiple power user groups, and output multiple power metering models. The multiple power metering models further include multiple hybrid power metering models. Among them, training the hybrid power metering models includes: Collect the hybrid-historical power operation dataset samples and the hybrid energy ratio of the corresponding user group; Perform data fusion on the hybrid-historical power operation dataset samples to output hybrid-fused power operation dataset samples; Extract hybrid features according to the hybrid-fused power operation dataset samples to output hybrid-fused power operation feature vectors; Initialize a multi-layer time series model, and train the hybrid-fused power operation feature vectors according to the multi-layer time series model to output multiple hybrid power metering models.

4. The metering verification method for grouped power users according to claim 1, characterized in that, Optimize the model parameters of the multiple power metering models according to the model parameters of the integrated power metering model. Among them, the model parameter update formula for each power metering model is: ； Among them, The model parameter of the i-th power metering model at the (t + 1)-th iteration, is the model parameter of the i-th power metering model at the t-th iteration, is the learning rate, which is used to control the speed of model parameter update, is the gradient of the loss function L with respect to the model parameter , reflecting the sensitivity of each model parameter to the error; Through repeated iteration until convergence, output the optimized multiple power metering models.

5. The metering verification method for grouped power users according to claim 1, wherein Construct corresponding power metering models according to the multiple power user groups, and output multiple power metering models. The method further includes: Identify the user group quantity index in the multiple power user groups; When the user group quantity index is less than the preset user group quantity index, make a mark and output the marked power user group; Activate the generative adversarial network, and obtain the generative power operation dataset samples corresponding to the marked power user group according to the generative adversarial network; Perform a power metering model according to the historical power operation dataset samples corresponding to the marked power user group and the generative power operation dataset samples.

6. The metering verification system for grouped power users is characterized in that, For implementing the metering verification method for the grouped power users according to any one of claims 1-5, the system includes: An energy type acquisition module, which is used to acquire the power energy types of each power user in the power user group; A grouping module, which is used to group the power user group according to the power energy type and output multiple power user groups. Among them, the users in each user group have the same energy type; A model construction module, which is used to construct corresponding power metering models according to the multiple power user groups and output multiple power metering models; An optimization module, which is used to perform ensemble learning on the multiple power metering models, obtain an integrated power metering model, optimize the model parameters of the multiple power metering models according to the model parameters of the integrated power metering model, and output the optimized multiple power metering models; A data acquisition module, which is used to acquire multiple real-time power operation datasets corresponding to the multiple power user groups; A verification module, which is used to map and input the multiple real-time power operation datasets into the optimized multiple power metering models for verification and output the metering verification results.

Images