A kind of gateway electric energy meter measurement calibration method and system

Abstract

The present application relates to the technical field of gateway electric energy meter calibration, and discloses a gateway electric energy meter calibration method and system. The method comprises: monitoring the real-time operation of the gateway electric energy meter, the measurement value and the current environmental condition, and synchronously querying the associated past calibration record; classifying and arranging the environmental condition data in the past record to form multiple environmental category groups, extracting historical deviation data and calibration time set from each group, and calculating to obtain a historical evaluation index set; compressing each evaluation index set to obtain an evaluation index sequence and organizing the evaluation index sequence into a time series data set according to time, analyzing and deducing the error growth and reduction trend; matching the current environmental condition with each environmental category group to determine the best matching group, adjusting the error trend according to the environmental difference and outputting; selecting a calibration strategy according to the adjusted trend, generating a calibration guidance scheme and executing a prompt action.

Description

A method and system for metering and verifying energy meters at the gateway Technical Field

[0001] This invention relates to the field of energy meter calibration technology, specifically to a method and system for energy meter calibration. Background Technology

[0002] In the operation of power systems, gateway energy meters are crucial devices for measuring the exchange of electrical energy between different regions and entities within the power grid. Their measurement results directly impact key tasks such as power transaction settlement, power grid operation status assessment, and power resource dispatch. With the rapid development of the power industry, the power grid coverage is continuously expanding, and electricity load is constantly increasing. The operating environment of gateway energy meters is also becoming increasingly complex. Significant differences exist in environmental conditions such as temperature, humidity, voltage fluctuations, and electromagnetic interference across different regions, all of which affect the measurement accuracy of gateway energy meters.

[0003] Currently, the industry primarily uses periodic calibration for metering at key points (such as annually or biennially). This involves disassembling the meters from the operating site and sending them to a professional calibration laboratory for testing and adjustment of their metering accuracy under standard environmental conditions. While this periodic calibration method can ensure the metering accuracy of key points to a certain extent, it has significant limitations. Periodic calibration requires disassembling and reassembling the meters, which not only consumes considerable manpower, resources, and time but may also damage the meters or related electrical equipment during the process. Furthermore, disassembly and reassembly can interrupt electricity metering at key points, affecting the normal settlement of electricity transactions. Periodic calibration cannot reflect the metering status of key points in the actual operating environment in real time. Because the calibration is conducted under standard conditions, and the actual operating environment of the meters differs from the standard environment, the calibration results cannot accurately reflect the metering deviation in real-world scenarios. This may result in some meters with metering errors going undetected, leading to metering disputes or economic losses.

[0004] In addition, some calibration methods attempt to incorporate environmental factors, but most only provide simple compensation for a single environmental factor (such as temperature), failing to comprehensively consider the combined effects of multiple environmental conditions. These methods typically do not systematically classify and organize environmental condition data from past calibration records, nor do they extract deviation information and calibration time information from historical data to construct evaluation indicators, and they cannot deduce error change trends through time series analysis. In practical applications, these methods cannot formulate appropriate calibration strategies based on the correlation between current and historical environmental conditions, often resulting in over- or under-calibration. For example, when an electricity meter is in an environment similar to one with historically large measurement deviations, failure to adjust the calibration strategy in a timely manner may delay calibration, leading to a continuous accumulation of measurement errors; conversely, if calibration is still performed at a fixed frequency when environmental conditions are stable and measurement deviations are small, it will result in wasted resources. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for metering and verifying energy meters at checkpoints, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a method for metering and verifying a gated energy meter, the method comprising:

[0007] Monitor the meter readings and current environmental conditions of the energy meter at the monitoring point in real time, and query the past verification records associated with the meter readings.

[0008] The environmental condition data in the past verification records are classified and organized into multiple environmental category groups. From each environmental category group, the historical deviation data set and the historical verification time set are extracted, and multiple historical evaluation index sets are calculated.

[0009] Information compression processing is performed on each set of historical evaluation indicators to obtain a compressed evaluation indicator sequence, which is then organized into multiple time series datasets in chronological order. The development patterns of these multiple time series datasets are analyzed to deduce the error growth trend and the error reduction trend.

[0010] The current environmental conditions are matched with the multiple environmental category groups to determine the best matching category group. Based on the degree of difference between the current environmental conditions and the reference environmental conditions of the best matching category group, the error growth trend and error reduction trend are adjusted, and the adjusted growth trend and reduction trend are output.

[0011] Based on the adjusted growth and decline trends, a verification strategy is selected, a verification guidance plan is generated, and prompts are executed.

[0012] Preferably, the monitoring involves checking the meter readings and current environmental conditions of the energy meter at the monitoring point during real-time operation, and querying past verification records associated with the meter readings, specifically including:

[0013] The system reads the instantaneous power consumption value of the energy meter at the checkpoint in real time, and simultaneously collects ambient temperature, humidity, and power grid frequency parameters as the current environmental conditions; using the instantaneous power consumption value as a keyword, it searches the historical database for previous verification files of the same energy meter to obtain past verification records.

[0014] Preferably, the environmental condition data in the past verification records are classified and organized into multiple environmental category groups. From each environmental category group, a set of historical deviation data and a set of historical verification times are extracted, and multiple sets of historical evaluation indicators are calculated, specifically including:

[0015] Cluster analysis is performed on various environmental condition parameters stored in past verification records to divide them into multiple environmental category groups. For each environmental category group, the standard error range recorded in the historical verification archives is retrieved to generate a historical deviation data set. Multiple sets of actual error values ​​and multiple sets of verification time points submitted by operators under the corresponding environmental category group are extracted. The deviation between the multiple sets of actual error values ​​and the historical deviation data set is calculated, and multiple sets of basic evaluation indicators are obtained by grouping them. Based on the time span ratio between the preset time window and the multiple sets of verification time points, the multiple sets of basic evaluation indicators are weighted and fused to obtain multiple sets of historical evaluation indicators.

[0016] Preferably, each historical evaluation indicator set undergoes information compression processing to obtain a compressed evaluation indicator sequence, which is then organized into multiple time series datasets in chronological order, specifically including:

[0017] Select one set from multiple historical evaluation indicator sets as the processing object, and designate a central evaluation indicator within this set; measure the dispersion of other evaluation indicators in this set relative to the central evaluation indicator, and assign sampling probabilities based on the dispersion to form an initial probability distribution, where the dispersion is inversely proportional to the sampling probability; randomly sample the set according to the initial probability distribution, retain the key evaluation indicators, and form a compressed dataset; sort the evaluation indicators in the compressed dataset according to their time labels to generate a time series dataset; iteratively perform compression and sorting operations on the remaining historical evaluation indicator sets to obtain multiple time series datasets.

[0018] Preferably, the set is randomly sampled according to the initial probability distribution, key evaluation indicators are retained, and a compressed dataset is formed, specifically including:

[0019] From the set, a subset of evaluation indicators is randomly selected according to a preset sampling rate as a candidate compression set; the dispersion of evaluation indicators within the candidate compression set and the central evaluation indicator is calculated, the sampling probability is updated, and a candidate probability distribution is obtained; the overlap between the candidate probability distribution and the initial probability distribution is compared, and this is used as the compression quality score; a subset of evaluation indicators is randomly selected again, and a new compression quality score is calculated; the sampling process is repeated until the quality score stabilizes, and the subset of evaluation indicators with the highest quality score is selected as the compressed dataset.

[0020] Preferably, the development patterns of the multiple time series datasets are analyzed to deduce the error growth trend and error reduction trend, specifically including:

[0021] Collect time-series data from multiple sets of electricity meters to construct a sample database. Identify error change patterns from each sample time series and extract sample growth and decrease trends. Use the sample time series data as training input and the sample growth and decrease trends as training outputs to construct a trend prediction model. Apply the trend prediction model to perform pattern recognition on multiple time series datasets, outputting multiple category growth trends and multiple category decrease trends. Calculate the correlation strength between the current environmental conditions and multiple environmental category groups. Based on the correlation strength, linearly combine the multiple category growth trends and multiple category decrease trends to obtain the error growth trend and error decrease trend.

[0022] Preferably, the current environmental conditions are matched with the multiple environmental category groups to determine the best matching category group. Based on the degree of difference between the current environmental conditions and the reference environmental conditions of the best matching category group, the error growth trend and error reduction trend are adjusted, and the adjusted growth trend and reduction trend are output. Specifically, this includes:

[0023] Select the environmental category group with the strongest correlation as the best matching category group, and obtain the reference environmental conditions of the best matching category group; measure the Euclidean distance between the current environmental conditions and the reference environmental conditions, and set an adjustment factor based on the distance; use the adjustment factor to perform scaling operations on the error growth trend and error reduction trend to generate the adjusted growth trend and reduction trend.

[0024] Preferably, the verification strategy is selected based on the adjusted growth and decline trends, a verification guidance plan is generated, and prompts are executed, specifically including:

[0025] The system aggregates adjusted growth and decline trend samples from historical operations. Based on the numerical relationship of each sample, it defines sample guidance schemes to form a sample scheme library. The sample guidance schemes include prompt levels, and the sample numerical size is negatively correlated with the prompt level. The system trains a decision engine using adjusted growth and decline trend samples as input features and sample guidance schemes as output labels. The decision engine is then used to classify the current adjusted growth and decline trends to generate a validation guidance scheme.

[0026] Preferably, the method further includes: after generating the verification guidance scheme, checking the consistency between the scheme and the real-time measurement values; if the consistency is lower than a set threshold, restarting the information compression processing and time series dataset generation process.

[0027] Preferably, the present invention also includes a gate energy meter metering verification system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the gate energy meter metering verification method described above.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] During the verification process, the meter readings and current environmental conditions of the electricity meters at the checkpoints are monitored in real time, and related past verification records are queried, thus combining real-time and historical data. By integrating these two types of data, the one-sidedness of verification caused by relying solely on real-time data can be avoided, as well as the problem of being unable to trace the root cause of errors due to neglecting historical operating conditions, allowing the verification work to be carried out based on more comprehensive information.

[0030] In processing past verification records, this method categorizes environmental condition data into multiple environmental category groups. Then, it extracts historical deviation data sets and historical verification time sets from each group and calculates a set of historical evaluation indicators. This categorization method systematically organizes complex and diverse historical environmental data, creating clear groupings of historical verification information under different environmental conditions. This facilitates subsequent analysis of the metering characteristics of electricity meters for different environmental scenarios. Furthermore, the construction of the historical evaluation indicator set transforms scattered historical deviation data and verification time data into valuable indicators, providing effective data support for subsequent error trend analysis.

[0031] By compressing historical evaluation indicators into a time-series dataset and then analyzing its development patterns to deduce error growth and reduction trends, this process can uncover the inherent patterns of error changes from a large amount of historical data. Information compression removes redundant data, reduces data processing complexity, and retains key information. Organizing the dataset chronologically provides a clear view of how error indicators change over time. Analysis of this dataset clearly reveals the direction and speed of error changes in electricity meters under different time periods and environmental conditions, providing a clear basis for formulating subsequent verification strategies.

[0032] By performing similarity matching between the current environmental conditions and multiple environmental category groups to determine the best matching category group, and adjusting the error trend according to the degree of environmental differences, the error trend can be made to better reflect the current operating environment of the electricity meter. Different environmental conditions have varying degrees of impact on the metering error. Finding the historical environmental group that is closest to the current environment through similarity matching ensures that the referenced historical error data has a high degree of relevance. Furthermore, adjusting the trend according to environmental differences can compensate for the deviation between historical data and the current environment, making the adjusted error trend more accurately reflect the potential metering deviations of the electricity meter and improving the reliability of the error trend.

[0033] Selecting a verification strategy and generating a verification guidance plan based on the adjusted error trend makes the verification work more targeted and reasonable. Based on accurate error trends, it's possible to determine whether the electricity meter needs verification, when to perform verification, and which verification method to use, avoiding the blindness inherent in traditional periodic verification. For situations where errors are increasing rapidly, a verification plan can be generated promptly and verification prompts can be triggered to prevent error accumulation; for situations where errors are stable or decreasing slowly, the verification cycle can be appropriately extended to reduce unnecessary disassembly and installation operations and resource consumption, while also avoiding the impact of verification interruptions on electricity metering and settlement. Attached Figure Description

[0034] Figure 1 is a schematic diagram of the working principle of the metering and verification method for a gate energy meter according to the present invention.

[0035] Figure 2 is a flowchart of the generation of environmental condition classification and historical assessment index set;

[0036] Figure 3 is a compressed analysis chart of measurement assessment indicators under environmental category groups;

[0037] Figure 4 is a flowchart illustrating the derivation of the error growth trend and the error reduction trend;

[0038] Figure 5 shows the relationship between environmental similarity matching indicators. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Referring to Figure 1, this invention provides a method for metering and verifying a boundary energy meter. The method includes: monitoring the metering value and current environmental conditions of the boundary energy meter in real-time operation, and querying past verification records associated with the metering value. The environmental condition data in the past verification records are categorized and organized into multiple environmental category groups. Historical deviation data sets and historical verification time sets are extracted from each environmental category group, and multiple historical evaluation index sets are calculated. Information compression processing is performed on each historical evaluation index set to obtain a compressed evaluation index sequence, which is then organized into multiple time-series datasets in chronological order. The development patterns of the multiple time-series datasets are analyzed to deduce the error growth trend and error reduction trend. The current environmental conditions are matched with the multiple environmental category groups based on similarity to determine the best matching category group. The error growth trend and error reduction trend are adjusted according to the degree of difference between the current environmental conditions and the reference environmental conditions of the best matching category group, and the adjusted growth trend and reduction trend are output. A verification strategy is selected based on the adjusted growth trend and reduction trend, a verification guidance scheme is generated, and prompts are executed.

[0041] Example 1: Referring to Figure 2, the instantaneous energy value of the meter is read in real time during operation. Simultaneously, ambient temperature, humidity, and grid frequency parameters are collected to determine the current environmental conditions. This process is accomplished through a sensor network and data acquisition unit deployed at the metering site. The data acquisition unit continuously captures the cumulative energy value output from the meter's pulse output or digital communication interface at a preset sampling frequency. Ambient temperature and humidity sensors are installed inside or near the meter box to accurately reflect the microenvironment in which the meter operates. Grid frequency measurement is obtained through monitoring equipment connected to the same busbar segment. Using the instantaneous energy value as a keyword, the historical database is searched for past calibration records of the same meter. The historical database stores all periodic calibrations, temporary calibration events, and daily inspection records throughout the meter's lifecycle. Each past calibration record includes the standard instrument reading recorded during calibration, the meter reading, the calculated error value, and environmental parameters such as ambient temperature, humidity, and grid frequency at the time of calibration. The database index is built on the meter's unique identifier and the energy value timestamp for rapid retrieval.

[0042] Cluster analysis was performed on various environmental condition parameters stored in past calibration records to divide them into multiple environmental category groups. The K-means algorithm was used to process the historical environmental condition data. The input feature vector consisted of values ​​in three dimensions: temperature, humidity, and power grid frequency. The data for each dimension was normalized before clustering to eliminate the influence of dimensional differences on the clustering results. The number of clusters was determined using the elbow rule or silhouette coefficient method to achieve a grouping effect with high similarity within groups and large differences between groups. For each environmental category group, the standard error range recorded in the historical calibration archives was retrieved to generate a historical deviation data set. The standard error range is defined by the metrological regulations or the technical specifications of the gate electricity meter. The historical deviation data set contains the error values ​​that were judged to be qualified in all historical calibration records corresponding to that environmental category group, forming a statistical distribution of errors. Extract multiple sets of actual error values ​​and multiple sets of verification time points submitted by operators under the corresponding environment category group. The multiple sets of actual error values ​​come from the error data actually observed and entered into the system by operators during the verification process. These data may exceed the standard error range. The multiple sets of verification time points accurately record the year, month, day, hour, minute and second information of each verification operation, reflecting the historical time distribution of the verification activities.

[0043] Multiple basic evaluation index sets are obtained by grouping the deviations of multiple sets of actual error values ​​with historical deviation data sets. The deviation is calculated using the absolute value of the difference between the average of each set of actual error values ​​and the average of the historical deviation data sets, while also considering the ratio of the standard deviations of the two sets for standardization, making the deviation a dimensionless value that facilitates cross-group comparisons. Multiple historical evaluation index sets are obtained by weighted fusion of these basic evaluation index sets based on a preset time window and the time span ratio of multiple verification time point sets. The preset time window is a configurable parameter, such as 30 days or 90 days, used to define an analysis period. The time span ratio is the ratio of the actual number of days covered by all verification time points under a certain environmental category group to the number of days in the preset time window. During the weighted fusion process, each basic evaluation index set is assigned a weight coefficient, which is positively correlated with the time span ratio. A larger time span ratio means that the data in that group is more comprehensive in the time dimension, and its weight is higher. The fused historical evaluation index sets comprehensively reflect the historical performance and temporal distribution characteristics of the meter's error under specific environmental conditions.

[0044] Example 2: Select a set from multiple historical evaluation indicator sets as the processing object. Within this set, designate a central evaluation indicator. The selection criteria for the central evaluation indicator are its position in the time series and the central tendency of its values. Typically, the evaluation indicator whose timestamp is located at the median of the timestamps of all evaluation indicators in the set is selected. If multiple candidates exist, the one whose value is closest to the average of all evaluation indicators in the set is further selected as the central evaluation indicator. Measure the dispersion of other evaluation indicators within the historical evaluation indicator set relative to the central evaluation indicator. The dispersion is quantified using the average absolute deviation method, i.e., calculating the absolute value of the difference between the value of each evaluation indicator in the set and the value of the central evaluation indicator. Then, sum all these absolute values ​​and divide by the total number of evaluation indicators to obtain a scalar value characterizing the overall volatility level within the set. An initial probability distribution is formed by allocating sampling probabilities based on the degree of dispersion. The degree of dispersion is inversely proportional to the sampling probability. The greater the degree of dispersion of an evaluation index, the greater its difference from the center point. In the initial screening, the probability of retaining it is set to be lower because overly discrete points may represent noise or outliers. Evaluation indices with smaller degrees of dispersion are considered more representative, and their sampling probability is correspondingly increased. The initial probability distribution ensures that the sampling process tends to retain evaluation indices that can represent the core characteristics of the set.

[0045] A compressed dataset is formed by randomly sampling the historical evaluation indicator set according to the initial probability distribution, retaining key evaluation indicators. A subset of evaluation indicators is then randomly selected from the historical evaluation indicator set as a candidate compressed set according to a preset sampling rate (a system parameter, such as 30% or 50%), which defines the ratio between the size of the compressed dataset and the original set. The dispersion of evaluation indicators within the candidate compressed set and the central evaluation indicator is calculated, updating the sampling probabilities to obtain the candidate probability distribution. This dispersion calculation is repeated within the new candidate compressed set. Based on the new dispersion value, a new probability is assigned to each evaluation indicator in the candidate compressed set according to the same inverse proportionality rule, forming a candidate probability distribution. The degree of overlap between the candidate probability distribution and the initial probability distribution is used as the compression quality score. The overlap is measured using statistical correlation calculations, such as calculating the cosine similarity of the two probability distribution vectors. The cosine similarity ranges from 0 to 1; the closer the value is to 1, the more similar the two distributions are, and the higher the compression quality score, meaning the candidate compressed set better preserves the probabilistic characteristics of the original historical evaluation indicator set. A new compressed quality score is calculated by randomly selecting a subset of evaluation metrics. The sampling process is repeated until the quality score stabilizes. The subset with the highest quality score is selected as the compressed dataset. The repeated sampling process is implemented through loop iteration. Each iteration generates a new candidate compressed set and calculates its compressed quality score. The quality score is stable when the change in the compressed quality score is less than a preset convergence threshold in several consecutive iterations. After the iteration is completed, the subset with the highest compressed quality score is selected from all generated candidate compressed sets as the final compressed dataset.

[0046] A time-series dataset is generated by sorting the evaluation indicators in the compressed dataset according to their time labels. Each evaluation indicator is accompanied by a precise timestamp during generation. The sorting algorithm arranges the evaluation indicators in the compressed dataset in ascending order according to the order of the timestamps, forming a strictly ordered sequence in the time dimension. This sequence is the time-series dataset used for subsequent trend analysis. The compression and sorting operations are iteratively performed on the remaining historical evaluation indicator sets to obtain multiple time-series datasets. The iterative process means that for each historical evaluation indicator set generated by different environmental category groups, the above operation process is independently and completely executed, including specifying the central evaluation indicator, calculating the dispersion, assigning initial probabilities, randomly sampling to generate candidate sets, calculating the compression quality score, iteratively selecting the optimal compression set, and sorting by time. The final output multiple time-series datasets correspond completely to the input historical evaluation indicator sets in terms of quantity. Each time-series dataset represents the sequence of the evolution of the metering error evaluation indicator of the gate energy meter over time under specific environmental category conditions after information compression and purification. The core purpose of information compression is to reduce the amount of data while preserving the main characteristics of the historical evaluation index set. This reduces the computational complexity of subsequent trend analysis algorithms and improves their sensitivity to long-term pattern identification. Organizing the data in chronological order ensures consistency between the data format and the input requirements of the time series analysis model. The operations described in this embodiment constitute a data preprocessing flow that transforms a scattered set of historical evaluation indicators, which may contain redundant information, into multiple refined, time-aligned time series datasets. This provides structured input data for accurately deriving error growth and reduction trends.

[0047] Referring to Figure 3, in the implementation of a metering verification method for energy meters, the effectiveness of information compression is presented by comparing the original data volume and the compressed data volume under different environmental category groups. Specifically, the horizontal axis of the bar chart represents the environmental category groups (groups 1 to 5), and the vertical axis represents the data volume. Gray bars represent the original evaluation index volume, and dark bars represent the compressed evaluation index volume. The figure indicates the compression rate for each group; for example, group 1 has a compression rate of 40.0%, quantifying the efficiency of data compression. In terms of technical details, the information compression process calculates the sampling probability distribution based on the dispersion of the historical evaluation index set, retains key indicators through random sampling, and forms a compressed dataset. The compression quality is evaluated by the overlap between the candidate probability distribution and the initial probability distribution, ensuring the consistency of topological features.

[0048] Example 3: Referring to Figure 4, a sample database is constructed by collecting time-series data from multiple sets of electricity meters. The database relies on the historical data warehouse of the power company's metering automation system. The data covers various types of metering meters installed on multiple substations and feeders. The length of the sample time-series data varies from several months to several years to ensure complete coverage of seasonal climate change cycles and equipment aging trends. Error change patterns are identified in each sample time series, and sample growth and decline trends are extracted. The pattern recognition process uses a sliding window combined with least squares linear fitting. The width of the sliding window is dynamically set according to the data sampling interval, for example, with a window period of thirty days. Linear regression analysis is performed on the time-series data within each window. The slope of the regression line is used to quantify the average rate of error change. Intervals with a positive and statistically significant slope are marked as a sample growth trend segment, and their slope value, start timestamp, and duration are recorded. Intervals with a negative and statistically significant slope are marked as a sample decline trend segment, and their corresponding feature values ​​are recorded. Using sample time-series data as training input and sample growth and decline trends as training outputs, a trend prediction model is constructed. This model employs a deep learning framework based on a long short-term memory (LSTM) neural network. The network input layer is designed to receive fixed-length, standardized time-series data segments. The input data is normalized before being fed into the network, ensuring its values ​​are between zero and one. The network output layer contains two independent neurons corresponding to the confidence levels of the growth and decline trends, respectively, with output confidence values ​​also ranging from zero to one. The LSM neural network internally includes multiple layers of memory units and dropout layers. Memory units capture long-term dependencies in the time series, while dropout layers prevent overfitting. Network training utilizes an adaptive moment estimation algorithm to iteratively optimize network parameters by minimizing the binary cross-entropy loss function between the predicted output and the true label.

[0049] A trend prediction model is applied to pattern recognition on multiple time series datasets, outputting multiple categories of increasing and decreasing trends. The time series datasets are ordered sequences after prior information compression, each corresponding to an environmental category group. Each time series dataset is segmented to the required length for the model input. Each segment is fed into the trained trend prediction model for forward propagation. The model outputs an increasing trend confidence value and a decreasing trend confidence value for each segment. The arithmetic mean of the confidence values ​​output by all data segments within the same environmental category group is calculated to obtain the category increasing trend value and category decreasing trend value, representing the overall error change direction of that environmental category group. The association strength between the current environmental condition and multiple environmental category groups is calculated. The association strength is based on the normalized similarity measure between the current environmental condition vector and the central environmental vectors of each environmental category group. Each dimension of the current environmental condition vector and the central environmental vectors is normalized before calculation to eliminate the influence of dimensions. Normalization transforms the value of each environmental parameter to the range of zero to one. Association Strength The calculation formula is as follows:

[0050] ;

[0051] in: This indicates the strength of the association between the current environmental conditions and the k-th environmental category group. This represents the Euclidean distance between the normalized current environment condition vector and the normalized center environment vector of the k-th environment category group.

[0052] The error growth trend and error reduction trend are obtained by linearly combining the growth trends and reduction trends of multiple categories based on the association strength. The linear combination operation uses a weighted average algorithm, where the category growth trend value contributed by each environmental category group is multiplied by the association strength of that group. As weights, all weights are added together to obtain the total weight. The sum of the weighted category growth trend values ​​divided by the total weight yields the final error growth trend value. The error reduction trend value is calculated using the same weighted averaging process and the same association strength. The trend values ​​of each category are fused using these values ​​as weights. This correlation strength-based weighting method ensures that the historical trend information contained in environmental category groups that are more similar to the current actual environmental conditions has a greater influence on the final result, thus making the derived error growth and error reduction trends more consistent with the current operating environment. The entire process learns common patterns from large-scale historical samples, establishes a mapping relationship from data to trends, applies the model to specific, compressed field data, and finally synthesizes the final trend prediction through an environmentally adaptive weighting mechanism, forming a data-driven path from historical experience to current decision-making. The introduction of the trend prediction model improves the automation and accuracy of extracting effective features from complex time series, while the correlation strength-weighted fusion strategy enhances the environmental relevance and practicality of the trend inference results.

[0053] Example 4: The environmental category group with the strongest correlation is selected as the best matching category group. The correlation strength is calculated based on the formula defined in the previous examples. The system iterates through the calculated correlation strength values ​​between the current environmental conditions and each environmental category group, and finds the environmental category group with the largest value through a comparison algorithm. This environmental category group represents the set of conditions in historical data that are most similar to the current environment. Reference environmental conditions for the best matching category group are obtained. The reference environmental conditions are not a single historical record point, but are obtained by calculating the arithmetic mean of each dimension of all historical environmental condition data points within the best matching category group. For example, the reference temperature is the average of all recorded temperatures within the group, the reference humidity is the average of all recorded humidity within the group, and the reference power grid frequency is the average of all recorded power grid frequencies within the group, thus forming a representative comprehensive environmental vector.

[0054] The Euclidean distance between the current environmental conditions and the reference environmental conditions is measured. The Euclidean distance is calculated in a normalized data space. The three parameters of the current environmental conditions—temperature, humidity, and power grid frequency—as well as the corresponding parameters of the reference environmental conditions, have all been pre-normalized to the [0,1] interval. The distance calculation follows the Euclidean distance formula in three-dimensional space, which is the square root of the sum of the squares of the differences in each dimension. An adjustment factor is set based on the distance. The adjustment factor is designed as a monotonically decreasing function of the Euclidean distance, and its specific numerical mapping relationship is determined by a predefined configuration table. This configuration table defines the adjustment factor values ​​corresponding to different distance ranges, aiming to map continuous distance values ​​to continuous adjustment factor values, ensuring the smoothness of trend adjustments.

[0055] Table 1: Mapping Relationship between Euclidean Distance and Adjustment Factor

[0056]

[0057] The scaling operation uses an adjustment factor to scale the error growth and reduction trends, generating adjusted growth and reduction trends. The scaling operation employs a multiplicative model: the initially derived error growth trend value is multiplied by the adjustment factor to obtain the adjusted growth trend value, and the initially derived error reduction trend value is multiplied by the adjustment factor to obtain the adjusted reduction trend value. An adjustment factor value less than or equal to 1 indicates a greater difference between the current environmental conditions and the reference environmental conditions of the best-matching category group, a larger Euclidean distance, and a smaller adjustment factor. This results in a stronger attenuation effect on the original trend estimate, reflecting the increased uncertainty caused by environmental differences and a corresponding decrease in the confidence level of the trend prediction. This scaling is a linear adjustment, simple and direct in calculation, and can proportionally correct the magnitude of the trend prediction based on the environmental matching degree.

[0058] The entire environmental matching and trend adjustment process constructs a feedback mechanism that transforms abstract environmental similarity metrics into specific numerical correction coefficients. Euclidean distance quantifies the degree of difference between the current environment and historical typical environments, while the adjustment factor translates this difference into the strength of the correction to the trend prediction results. Obtaining the adjustment factor by querying a mapping table avoids complex real-time calculations, ensuring the efficiency of the system response. The adjusted growth and decline trends serve as inputs for subsequent verification strategy selection; their values ​​already contain information about the environmental matching degree. This ensures that the final verification guidance scheme is not only based on historical trends but also fully considers the current actual environmental conditions, enhancing the adaptability and rationality of the decision. From selecting the best matching category group to obtaining reference conditions, from calculating distances to querying adjustment factors, and finally completing trend scaling, this series of steps is interconnected, advancing the impact of environmental factors from qualitative judgment to quantitative adjustment, achieving data-driven adaptive trend correction. The tabular mapping relationship makes the adjustment strategy clear, explicit, and easy to configure and maintain, while the multiplicative scaling model ensures the simplicity and interpretability of the adjustment logic.

[0059] Referring to Figure 5, the correlation strength and Euclidean distance of different environmental categories (groups A to E) are quantitatively displayed during the metering and verification process of the energy meter at the checkpoint. This is used to support environmental similarity matching and trend adjustment decisions. In the specific analysis, the correlation strength is calculated by a comparison algorithm, reflecting the degree of similarity between the current environmental conditions and the historical conditions of each environmental category group. Its value is directly used to determine the best matching category group. The Euclidean distance is a three-dimensional spatial measure based on normalized environmental parameters, quantifying the degree of difference between the current environment and the reference environmental conditions. It is converted into an adjustment factor through a preset mapping table to achieve scaling correction of error trends. In the chart, the bar chart represents the correlation strength, and the line chart represents the Euclidean distance. It can be seen that group C has the highest correlation strength (0.9), indicating that its historical data is most similar to the current environment, but the Euclidean distance is large (0.7), indicating that the environmental differences are significant and a strong attenuation of trend prediction is required through adjustment factors. Groups D and E have lower correlation strength and smaller Euclidean distances, which reflects the characteristics of weak similarity but low difference. This visualization structure provides a data-driven basis for the selection of verification strategies, ensuring that environmental factors are quantitatively integrated into the adaptive decision-making process.

[0060] Example 5: A sample solution library is formed by summarizing adjusted growth and decline trend samples from historical operations. Historical operations refer to the complete records generated by the system executing this verification method on all monitored energy meters over a past period. Each record contains a set of final adjusted growth and decline trend values, as well as the actual verification operation instructions taken for that energy meter at that time. Sample guidance schemes are defined based on the numerical relationships of each sample. A sample guidance scheme is a structured set of instructions, including suggested verification cycles, suggested verification item types, and prompt text to be displayed to operators. The prompt text is divided into different prompt levels based on the numerical relationships, such as "Normal Monitoring," "Suggested Scheduled Verification," and "Suggested Immediate On-Site Verification." The sample numerical value is negatively correlated with the prompt level. The smaller the absolute value of the adjusted growth and decline trend values, the more stable the energy meter's error state, the lower the predicted risk, and the lower the corresponding prompt level, such as "Normal Monitoring." Larger values ​​indicate a significant error change trend, high uncertainty, or large potential deviation, requiring more proactive intervention, and the higher the corresponding prompt level, such as "Suggested Immediate On-Site Verification." Using adjusted growth and decline trend samples as input features and sample guidance schemes as output labels, a decision engine is trained. The decision engine employs a classifier model based on the decision tree algorithm. The input feature is a two-dimensional vector containing adjusted growth and decline trend values, and the output label is a discrete category corresponding to different sample guidance schemes. During training, the decision tree selects the optimal feature split point by calculating information gain or Gini impurity, recursively dividing the feature space into different regions. Each leaf node corresponds to a specific sample guidance scheme category. The trained decision tree model can learn the complex nonlinear mapping relationship between continuous numerical features and discrete operational schemes.

[0061] The decision engine is invoked to classify the current adjusted growth and decline trends to generate verification guidance schemes. When the system calculates the current adjusted growth and decline trend values ​​for a specific energy meter, these two values ​​form a feature vector, which is then input into the trained decision engine. The decision engine traverses the decision tree through a series of conditional judgments starting from the root node, eventually reaching a leaf node. The sample guidance scheme represented by this leaf node is the generated verification guidance scheme. The verification guidance scheme is output in a structured data format, containing clear action instructions, suggested time windows, and detailed prompts. For example, the scheme might instruct, "Please arrange an accuracy verification within fifteen working days, focusing on checking the voltage loop connection," and this information is pushed to the maintenance workbench. After generating the verification guidance scheme, its consistency with real-time metering values ​​is verified. This consistency verification is achieved by comparing the error expectation implicit in the verification guidance scheme with the current operating status reflected by the real-time metering values ​​read from the energy meter. The system calculates a consistency score based on the degree of agreement between the real-time metering values ​​and the predicted range based on historical data. If the consistency score falls below a set threshold, the information compression and time-series dataset generation process will be restarted. The set threshold is a configurable threshold; for example, a consistency score below 0.7 will trigger a recalculation process. Restarting means the system will ignore previous trend analysis results based on compressed data and start from an earlier set of original historical evaluation indicators. It will then re-execute information compression to form a new time-series dataset and re-perform the entire process of trend derivation, environmental matching adjustments, and decision classification, aiming to obtain a new verification guideline that is more consistent with the current real-time data.

[0062] For example, suppose the system generates a verification guidance plan for an outdoor energy meter, with the following content: "Alert Level: Normal Monitoring; Recommendation: Perform routine checks in the next planned verification cycle." At this time, the system reads the meter's real-time readings and finds that the pulse count fluctuations have increased abnormally in a short period, inconsistent with the stable state expected by the "normal monitoring" plan. The consistency score is calculated to be 0.5, below the set threshold of 0.7. The system then triggers a recalculation process, retracing back to the meter's historical evaluation index set from the previous week, recompressing the information, constructing a new time series, and analyzing trends that may indicate potential accelerated error growth. After environmental matching adjustments, the new trend values ​​are input into the decision engine, potentially generating a completely new verification guidance plan, such as "Alert Level: Recommendation to arrange planned verification; Recommendation: Conduct on-site inspection within seven days, focusing on the temperature compensation module." This new plan has higher consistency with the observed real-time reading fluctuations. The entire process demonstrates the system's self-verification and iterative optimization capabilities; the generation of the verification guidance plan is not a one-way endpoint but rather incorporates a feedback loop. The application of the decision engine transforms continuous trend values ​​into explicit operational instructions, reducing the complexity of manual judgment. The consistency verification acts as a safety valve, ensuring that the system's recommended operations are consistent with the latest actual state of the equipment. When significant inconsistencies occur, the system can automatically backtrack and recalculate, effectively avoiding the risk of making erroneous decisions based on outdated or distorted trend information. From historical sample learning to real-time decision classification, and from consistency verification to possible process restarts, the method described in this embodiment constructs a dynamically adaptable intelligent verification decision support system. The establishment of a sample solution library allows decisions to be based on a wealth of historical experience, the implementation of the decision engine ensures the objectivity and automation of the decision-making process, and the consistency verification mechanism introduces necessary fault tolerance and adaptive capabilities, collectively improving the accuracy and reliability of metering verification management at key points.

[0063] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for metering and verifying a gated energy meter, characterized in that, The method includes: monitoring the meter readings and current environmental conditions of the energy meter in real-time operation; querying past verification records associated with the meter readings; classifying and organizing the environmental condition data in the past verification records into multiple environmental category groups; extracting historical deviation data sets and historical verification time sets from each environmental category group; calculating multiple historical evaluation index sets; performing information compression processing on each historical evaluation index set to obtain a compressed evaluation index sequence, and organizing it into multiple time series datasets in chronological order; analyzing the development patterns of the multiple time series datasets to deduce the error growth trend and error reduction trend; performing similarity matching between the current environmental conditions and the multiple environmental category groups to determine the best matching category group; and adjusting the error growth trend and error reduction trend based on the degree of difference between the current environmental conditions and the reference environmental conditions of the best matching category group, outputting the adjusted growth trend. The process involves: identifying and adjusting the growth and decline trends; selecting a verification strategy based on the adjusted growth and decline trends; generating a verification guidance scheme and executing prompts; and compressing information for each historical evaluation indicator set to obtain a compressed evaluation indicator sequence, which is then organized into multiple time-series datasets in chronological order. Specifically, this includes: selecting one set from multiple historical evaluation indicator sets as the processing object, specifying a central evaluation indicator within that set; measuring the dispersion of other evaluation indicators within the set relative to the central evaluation indicator, allocating sampling probabilities based on the dispersion, and forming an initial probability distribution, where the dispersion is inversely proportional to the sampling probability; randomly sampling the set according to the initial probability distribution, retaining key evaluation indicators to form a compressed dataset; sorting the evaluation indicators in the compressed dataset according to their time labels to generate a time-series dataset; and iteratively performing compression and sorting operations on the remaining historical evaluation indicator sets to obtain multiple time-series datasets.

2. The metering and verification method for a gated energy meter according to claim 1, characterized in that, The monitoring of the meter readings and current environmental conditions of the energy meter at the monitoring point is carried out in real time. The query of past verification records associated with the meter readings includes: reading the instantaneous power value of the energy meter at the monitoring point in real time, and simultaneously collecting ambient temperature, humidity and power grid frequency parameters as the current environmental conditions; using the instantaneous power value as the keyword, searching for past verification files of the same energy meter in the historical database to obtain past verification records.

3. The metering and verification method for a gated energy meter according to claim 1, characterized in that, The environmental condition data in the past verification records are classified and organized into multiple environmental category groups. From each environmental category group, historical deviation data sets and historical verification time sets are extracted, and multiple historical evaluation index sets are calculated. Specifically, this includes: performing cluster analysis on various environmental condition parameters stored in the past verification records to divide them into multiple environmental category groups; for each environmental category group, retrieving the standard error range recorded in the historical verification archives to generate a historical deviation data set, and extracting multiple sets of actual error values ​​and multiple sets of verification time points submitted by operators under the corresponding environmental category group; calculating the deviation between the multiple sets of actual error values ​​and the historical deviation data sets, and grouping them to obtain multiple basic evaluation index sets; and weighting and fusing the multiple basic evaluation index sets according to the time span ratio of a preset time window and the multiple sets of verification time points to obtain multiple historical evaluation index sets.

4. The metering and verification method for a gated energy meter according to claim 1, characterized in that, Random sampling is performed on the set according to the initial probability distribution, retaining key evaluation indicators to form a compressed dataset. Specifically, this includes: randomly selecting a subset of evaluation indicators from the set according to a preset sampling rate as a candidate compressed set; calculating the dispersion of other evaluation indicators in the candidate compressed set relative to the central evaluation indicator, updating the sampling probability, and obtaining a candidate probability distribution; comparing the overlap between the candidate probability distribution and the initial probability distribution as the compression quality score; randomly selecting a subset of evaluation indicators again and calculating a new compression quality score; repeating the sampling process until the quality score stabilizes, and selecting the subset of evaluation indicators with the highest quality score as the compressed dataset.

5. The metering and verification method for a gated energy meter according to claim 1, characterized in that, The analysis of the development patterns of the multiple time-series datasets derives the error growth trend and error reduction trend. Specifically, this includes: collecting sample time-series data from multiple sets of electricity meters, constructing a sample database, identifying error change patterns from each sample time series, and extracting sample growth and reduction trends; constructing a trend prediction model using the sample time-series data as training input and the sample growth and reduction trends as training output; applying the trend prediction model to perform pattern recognition on multiple time-series datasets, outputting multiple category growth trends and multiple category reduction trends; calculating the correlation strength between the current environmental conditions and multiple environmental category groups, and linearly combining the multiple category growth trends and multiple category reduction trends based on the correlation strength to obtain the error growth trend and error reduction trend.

6. The metering and verification method for a gated energy meter according to claim 5, characterized in that, The process involves: matching the current environmental conditions with multiple environmental category groups based on similarity to determine the best matching category group; adjusting the error growth trend and error reduction trend based on the degree of difference between the current environmental conditions and the reference environmental conditions of the best matching category group; and outputting the adjusted growth trend and reduction trend. Specifically, this includes: selecting the environmental category group with the strongest correlation as the best matching category group and obtaining the reference environmental conditions of the best matching category group; measuring the Euclidean distance between the current environmental conditions and the reference environmental conditions and setting an adjustment factor based on the distance; and using the adjustment factor to scale the error growth trend and error reduction trend to generate the adjusted growth trend and reduction trend.

7. The metering and verification method for a gated energy meter according to claim 1, characterized in that, The verification strategy is selected based on the adjusted growth and decline trends, a verification guidance scheme is generated, and a prompting action is executed. Specifically, this includes: summarizing the adjusted growth and decline trend samples from historical operations; defining sample guidance schemes based on the numerical relationship of each sample to form a sample scheme library, where the sample guidance schemes include prompt levels, and the sample numerical size is negatively correlated with the prompt level; training the decision engine using the adjusted growth and decline trend samples as input features and the sample guidance schemes as output labels; and calling the decision engine to classify the current adjusted growth and decline trends to generate a verification guidance scheme.

8. The metering and verification method for a gated energy meter according to claim 1, characterized in that, The method further includes: after generating the verification guidance scheme, checking the consistency between the scheme and the real-time measurement values; if the consistency is lower than a set threshold, restarting the information compression processing and time series dataset generation process.

9. A metering and verification system for a gated energy meter, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the metering and verification method for a gate energy meter as described in any one of claims 1 to 8.

Images