A method and system for estimating and judging operation error of electric meter

Abstract

The present invention provides an electric meter operation error estimation and judgment method and system, which relates to the field of data processing technology. The method comprises: obtaining electric meter operation data; performing EMD decomposition on the electric meter operation data to obtain a series of intrinsic mode functions; according to the characteristics of the intrinsic mode functions, identifying and removing the mode functions related to noise to obtain the residual mode functions; reconstructing the residual mode functions to obtain the denoised electric energy meter operation data; analyzing the denoised electric energy meter operation data to obtain the minimum roughness; constructing a decision tree for estimating the electric energy meter operation error according to the minimum roughness; judging and calculating the roughness attribute according to the constructed decision tree to realize the operation error estimation of the electric energy meter. The present invention constructs a decision tree for estimating the electric energy meter operation error based on the minimum roughness, and can quickly and accurately divide the data subsets and assign corresponding error estimation values ​​to each subset.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for estimating and judging an operation error of an electric meter. Background Art

[0002] With the rapid development of the power industry and the increasing maturity of intelligent technology, the status of electric energy meters in the power system has become increasingly prominent and has become an indispensable metering device. The operating accuracy and stability of electric energy meters are directly related to the ability to ensure power supply, the effect of promoting energy conservation, and the fairness of power market transactions. However, in actual application, electric energy meters are often interfered by various internal and external factors, such as electromagnetic interference, mechanical vibration, temperature changes, etc. The existence of these factors causes errors in the operation of electric energy meters, which in turn affects the accuracy of the measurement results.

[0003] Electromagnetic interference may come from strong electromagnetic fields in the surroundings or the switching operation of electrical equipment. These interference signals will penetrate into the metering circuit of the energy meter, causing the metering results to deviate from the true value. Mechanical vibration may be transmitted to the energy meter due to unstable operation of the equipment itself or vibration of the external environment, causing relative displacement of internal components, thereby affecting the accuracy of measurement. In addition, temperature change is also an important influencing factor. Extreme temperature or temperature fluctuations may cause changes in the performance of internal components of the energy meter, thereby causing operating errors. The existence of these errors will not only affect the stability and reliability of power supply, but may also cause energy waste and unreasonable fee settlement. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for estimating and judging the operation error of an electric meter, which constructs a decision tree for estimating the operation error of an electric meter based on the minimum roughness, and can quickly and accurately divide data subsets and assign corresponding error estimation values ​​to each subset.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a method for estimating and judging an operation error of an electric meter is provided, the method comprising:

[0007] Obtain meter operation data;

[0008] Perform EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions;

[0009] According to the characteristics of the intrinsic mode function, the mode function related to the noise is identified and removed to obtain the residual mode function;

[0010] Reconstruct the residual mode function to obtain the denoised electric energy meter operation data;

[0011] Analyze the denoised electric energy meter operation data to obtain the minimum roughness;

[0012] According to the minimum roughness, a decision tree for estimating the operation error of the electric meter is constructed;

[0013] According to the constructed decision tree, the roughness attributes are judged and calculated to achieve the operation error estimation of the electric energy meter.

[0014] Furthermore, the EMD decomposition is performed on the meter operation data to obtain a series of intrinsic mode functions, including:

[0015] By traversing the meter operation data, all local maximum points and local minimum points are obtained;

[0016] According to the local extreme points, the upper envelope and the lower envelope are fitted respectively by cubic spline interpolation method to obtain two envelopes;

[0017] According to the two envelopes, The mean envelope is calculated, where is the mean envelope at time point t The value of t It's the time point. It is in The original data value at each time point; is a time point index; n is the size of the sliding window, It's at the time t The maximum value in the sliding window of It's at the time t The minimum value in the sliding window of is the weight coefficient of the upper envelope, is the weight coefficient of the lower envelope;

[0018] The mean envelope is subtracted from the original meter operation data to obtain a series of intrinsic mode functions.

[0019] Furthermore, according to the characteristics of the intrinsic mode function, the mode function related to the noise is identified and removed to obtain the residual mode function, including:

[0020] By analyzing the characteristics of each intrinsic mode function, the characteristics of the noise can be identified;

[0021] Based on the characteristic analysis of noise, each modal function is checked one by one by setting a threshold to identify the noise modal function;

[0022] The noise mode function is removed from the original set of intrinsic mode functions to obtain the residual mode functions.

[0023] Furthermore, the residual mode function is reconstructed to obtain the denoised electric energy meter operation data, including:

[0024] Decomposing the residual mode function into a plurality of mode function components to obtain mode function components;

[0025] The modal function components;

[0026] pass Perform an addition operation to perform the reconstruction operation, where is the signal after reconstruction; represents the amplitude coefficient; represents the frequency parameter; Indicates phase shift; Represents the value of the input signal; is a constant term; represents the Gaussian function; Indicates time; is the Gaussian center position; Indicates the width of the distribution; is an overall offset constant; is the control offset term with the total number of mode functions exponential decay rate of change; represents the base of natural logarithms;

[0027] According to the reconstruction operation, a quality check is performed on the obtained denoised electric energy meter operation data to obtain the denoised electric energy meter operation data.

[0028] Furthermore, the denoised electric energy meter operation data is analyzed to obtain the minimum roughness, including:

[0029] The denoised electric energy meter operation data is sorted to obtain a sorted data set;

[0030] Based on the sorted data set, an information table is constructed;

[0031] According to the information table, simplify the data set by reducing the attributes;

[0032] Based on the simplified data set;

[0033] pass Calculate the kernel attributes and classify them according to the kernel attributes to obtain data subsets of each attribute, where: Representation sample and The similarity between and Represents two different sample features in the data set is the width parameter of the Gaussian kernel; is the number of dimensions of the feature; Indicates Features The variance of It is for Nonlinear transformation of the distance between two samples in feature dimensions; Indicates the data set The sample The value of a feature; Indicates the data set The sample The value of a feature; Indicates the first Features is an index variable used to traverse all features in the calculation of feature weights; Represents the total number of features in the dataset; is an index variable used to traverse all features in the calculation of the kernel function; is the index of the sample; is the index of the sample;

[0034] According to the data subset of each attribute, The roughness of the data subsets of each attribute is calculated to evaluate the roughness of the data, where: is the roughness of the data subset, is the number of samples in the data subset, Is the first in the data subset i The attribute value of each sample; is the mean of the data subset;

[0035] The minimum roughness is obtained by comparing the roughness corresponding to different attribute subsets.

[0036] Furthermore, based on all the nodes after inspection, a decision tree for estimating the operation error of the electric meter is constructed, including:

[0037] The minimum roughness is used as the root node of the decision tree, and the meter operation data is divided into different subsets according to the minimum roughness, and each subset corresponds to a branch;

[0038] For each subset after division, create a new child node and add it to the child node list of all nodes after verification to recursively build the subtree;

[0039] When the process of recursively constructing the subtree is terminated, some leaf nodes are generated, and error estimation values ​​are assigned to the leaf nodes to construct a decision tree for the error estimation of the electric meter operation.

[0040] Furthermore, according to the constructed decision tree, the roughness attributes are judged and calculated to realize the operation error estimation of the electric energy meter, including:

[0041] According to the constructed decision tree, the roughness attribute of the operating data is traversed from the root node along the path of the decision tree until a leaf node is reached to obtain a leaf node, wherein the value corresponding to the leaf node is the error estimation result of the operating data.

[0042] In a second aspect, a system for estimating and judging an operation error of an electric meter includes:

[0043] A collection module is used to obtain the operation data of the electric meter;

[0044] A processing module is used to perform EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions; according to the characteristics of the intrinsic mode functions, the mode functions related to noise are identified and removed to obtain the residual mode functions; the residual mode functions are reconstructed to obtain the denoised meter operation data;

[0045] An analysis module, used for analyzing the denoised electric energy meter operation data to obtain the minimum roughness;

[0046] The operation module is used to construct a decision tree for estimating the operation error of the electric energy meter according to the minimum roughness; and to judge and calculate the roughness attribute according to the constructed decision tree to realize the operation error estimation of the electric energy meter.

[0047] According to a third aspect, a computing device includes:

[0048] one or more processors;

[0049] The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method described.

[0050] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described is implemented.

[0051] The above solution of the present invention includes at least the following beneficial effects:

[0052] Through the EMD decomposition technology, this method can effectively decompose the meter operation data into a series of intrinsic mode functions, and then identify and remove the mode functions related to noise, which significantly improves the accuracy and reliability of the data. The advanced denoising technology is adopted to reconstruct the residual mode function to obtain the denoised meter operation data, which helps to reduce the interference of noise on the estimation of the meter operation error, making the estimation result more accurate. The denoised meter operation data is deeply analyzed, and the roughness of the data is more accurately evaluated by calculating the minimum roughness, which helps to find outliers and fluctuations in the data; the decision tree for the estimation of the meter operation error is constructed based on the minimum roughness. This method can quickly and accurately divide the data subsets and assign corresponding error estimation values ​​to each subset, which improves the efficiency and accuracy of the estimation of the meter operation error. The constructed decision tree can intuitively display the estimation results of the meter operation error. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of a method for estimating and judging an operation error of an electric meter provided by an embodiment of the present invention.

[0054] Figure 2 It is a schematic diagram of an electric meter operation error estimation and judgment system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in a form and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0056] like Figure 1 As shown, an embodiment of the present invention provides a method for estimating and judging an operation error of an electric meter, the method comprising the following steps:

[0057] Step 11, obtaining the meter operation data;

[0058] Step 12, performing EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions;

[0059] Step 13, according to the characteristics of the intrinsic mode function, identifying and removing the mode function related to the noise to obtain the residual mode function;

[0060] Step 14, reconstructing the residual mode function to obtain the denoised electric energy meter operation data;

[0061] Step 15, analyzing the denoised electric energy meter operation data to obtain the minimum roughness;

[0062] Step 16, constructing a decision tree for estimating the operation error of the electric meter according to the minimum roughness;

[0063] Step 17, judging and calculating the roughness attribute according to the constructed decision tree, so as to realize the operation error estimation of the electric energy meter.

[0064] In the embodiment of the present invention, the original data of the electric meter in actual operation can be collected to provide a basis for data processing and analysis. Through EMD decomposition, the complex electric meter operation data can be decomposed into a series of intrinsic mode functions, which is helpful to understand the intrinsic structure and characteristics of the data more deeply. Effectively removing the noise mode function can improve the signal-to-noise ratio of the data, reduce the impact of noise on subsequent analysis, and make the analysis results more accurate and reliable. By reconstructing the residual mode function, the denoised electric energy meter operation data can be obtained, which can reflect the real operation state of the electric meter. By analyzing the denoised data and calculating the minimum roughness, the hidden laws and potential problems in the data can be revealed, providing an important basis for the performance evaluation and optimization of the electric meter. A decision tree for estimating the operation error of the electric meter is constructed. By constructing a decision tree, an intuitive and easy-to-understand electric meter operation error estimation model can be formed, which is helpful to quickly and accurately judge the operation state of the electric meter and take corresponding measures. Finally, by judging and calculating the roughness attribute, the accurate estimation of the operation error of the electric energy meter can be achieved, providing strong support for the stable operation and energy conservation and emission reduction of the power system.

[0065] In a preferred embodiment of the present invention, in step 11, by setting the communication parameters of the electric meter, it is possible to ensure smoother communication between the data acquisition tool and the electric meter, reduce data transmission delay, and thus improve the efficiency of data acquisition. Correctly setting the trigger conditions for data acquisition can ensure that the electric meter data is collected at the set time point, avoid collecting data in abnormal working conditions or under abnormal circumstances, and thus ensure the accuracy of the collected data. By flexibly adjusting the communication parameters of the electric meter and the trigger conditions for data acquisition, the data acquisition system can be made more flexible.

[0066] In a specific embodiment of the present invention, according to the specification of the electric meter, the communication protocol, interface type and communication parameter range information supported by the electric meter are obtained, and the communication baud rate, data bit, stop bit, check mode and other parameters of the electric meter are set using the electric meter configuration tool or software to ensure that they match the communication settings of the data acquisition tool, and the frequency of data acquisition is determined, such as collecting by minute, hour or day. If the electric meter supports event-triggered acquisition, specific events (such as voltage fluctuations, sudden changes in power, etc.) can be configured as trigger conditions so that data can be collected in time when an abnormality occurs. Open the data acquisition software or device, create a new data acquisition task in the data acquisition tool, enter the communication address, communication protocol and other information of the electric meter, configure the frequency or trigger conditions of data acquisition, start the data acquisition task, and the data acquisition tool will automatically obtain operating data from the electric meter according to the configured communication parameters and trigger conditions.

[0067] In a preferred embodiment of the present invention, the above step 12 may include:

[0068] By traversing the meter operation data, all local maximum points and local minimum points are obtained;

[0069] According to the local extreme points, the upper envelope and the lower envelope are fitted respectively by cubic spline interpolation method to obtain two envelopes;

[0070] According to the two envelopes, The mean envelope is calculated, where is the mean envelope at time point t The value of t It's the time point. It is in The original data value at each time point; is a time point index; n is the size of the sliding window, It's at the time t The maximum value in the sliding window of It's at the time t The minimum value in the sliding window of is the weight coefficient of the upper envelope, is the weight coefficient of the lower envelope;

[0071] The mean envelope is subtracted from the original meter operation data to obtain a series of intrinsic mode functions.

[0072] In the embodiment of the present invention, by traversing the meter operation data to obtain all local maximum points and local minimum points, the fluctuation and change trend in the data can be captured more accurately; by fitting the upper envelope and the lower envelope using the cubic spline interpolation method, the data can be effectively smoothed to reduce the impact of noise and outliers on data analysis. By calculating the mean envelope and allowing the weight coefficients of the upper and lower envelopes to be adjusted, the meter operation data can be flexibly processed according to actual needs, thereby improving the applicability and accuracy of data processing. By subtracting the mean envelope to obtain a series of intrinsic mode functions, the inherent structure and characteristics of the meter operation data can be more deeply understood, which is helpful to identify and analyze abnormal conditions or fault modes in the operation of the meter. The entire processing flow provides a systematic and structured method for the preprocessing of the meter operation data. From the identification of local extreme points to the fitting of the envelope, to the calculation of the mean envelope and the extraction of the intrinsic mode function, each step has a clear purpose and effect, which helps to improve the efficiency and accuracy of data preprocessing.

[0073] In a specific embodiment of the present invention, the data set of the electric meter operation obtained according to the above steps is usually time series data, which records the readings of the electric meter at different time points; starting from the first data point of the data set, check the size relationship between each data point and its adjacent data points one by one, if a data point is larger than the adjacent data points before and after it, then mark the point as a local maximum point, if a data point is smaller than the adjacent data points before and after it, then mark the point as a local minimum point; record the positions and values ​​of all identified local maximum points and local minimum points. Using all the identified local maximum points, a curve passing through these maximum points is fitted by the cubic spline interpolation method, which is the upper envelope; similarly, using all the local minimum points, a curve passing through these minimum points is fitted by the cubic spline interpolation method, which is the lower envelope. Select a sliding window size, for each time point in the data set t , find out from tn arrive t The maximum and minimum values ​​during this period are calculated according to the given formula, combined with the weight coefficients of the upper and lower envelopes, and the maximum and minimum values ​​in each sliding window to calculate the mean envelope at each time point. t For each time point in the data set, the mean envelope value at the corresponding time point is subtracted from the original meter operation data. The result of the above subtraction operation is a series of intrinsic mode functions, which reflect the oscillation components of different frequencies in the original data. Ensure that each IMF meets the definition of the intrinsic mode function, that is, the number of extreme points and zero crossing points in the entire data set does not differ by more than one.

[0074] For example, suppose there is a set of time series data on electricity meter operation, recording meter readings for several consecutive days.

[0075] Identify local extreme points:

[0076] Traverse the entire data set and find all local maxima and local minima; for example, if the data set is [10, 15, 12, 17, 14, 19, 16], then 15, 17, and 19 are local maxima, and 12 and 14 are local minima.

[0077] Fitting envelope:

[0078] Using the cubic spline interpolation method, the upper envelope and the lower envelope are fitted through all the local maximum points and local minimum points respectively. These two envelopes wrap the data to form a "channel" for data changes.

[0079] Compute the mean envelope:

[0080] For each time point, the average of the upper envelope and the lower envelope is calculated to obtain the mean envelope, which reflects the overall trend of the data.

[0081] Extract the intrinsic mode functions:

[0082] Subtracting the mean envelope from the original data results in the first intrinsic mode function (IMF1), which reflects the highest frequency oscillation component in the data. The IMF obtained in the previous step is used as new data, and the above operation is repeated until all IMFs are extracted. Each iteration will extract the second highest frequency oscillation component in the data. Finally, the original data is decomposed into a series of intrinsic mode functions, each of which represents an oscillation component of a specific frequency in the data. By analyzing these IMFs, we can have a deeper understanding of the internal structure and characteristics of the meter operation data.

[0083] In a preferred embodiment of the present invention, the above step 13 may include:

[0084] By analyzing the characteristics of each intrinsic mode function, the characteristics of the noise can be identified;

[0085] Based on the characteristic analysis of noise, each modal function is checked one by one by setting a threshold to identify the noise modal function;

[0086] The noise mode function is removed from the original set of intrinsic mode functions to obtain the residual mode functions.

[0087] In an embodiment of the present invention, by analyzing the characteristics of each intrinsic modal function to identify noise, the signal can be effectively distinguished from the noise, the signal-to-noise ratio of the data can be significantly improved, and the useful information in the meter operation data can be more prominent. Based on the characteristic analysis of noise, by setting a reasonable threshold to check each modal function one by one, it is possible to accurately identify which modal functions are caused by noise. The precise identification ability helps to avoid misjudging noise as a useful signal or misjudging a useful signal as noise, thereby ensuring the accuracy and reliability of the data. Removing the identified noise modal function from the original intrinsic modal function set can further purify the data and obtain a clearer and more accurate residual modal function, which can better reflect the actual state of the meter operation. Since the meter operation data in the actual environment may be interfered by a variety of noises, removing the noise modal function by this method can enhance the robustness of data processing. The residual modal function after noise removal is more concise and clear, which helps to reduce the complexity and uncertainty in the analysis process and improve the efficiency and accuracy of decision-making.

[0088] In a specific embodiment of the present invention, a set of intrinsic mode functions is obtained according to the previous steps, and each mode function is time series data obtained after the original meter operation data is processed in a specific manner (such as subtracting the mean envelope). For each mode function, its key statistical features are calculated, such as mean, standard deviation, kurtosis, and skewness.

[0089] Each modal function is converted from the time domain to the frequency domain by the frequency domain analysis method of Fourier transform. In the frequency domain, the energy distribution, main frequency components and periodic characteristics of the modal function can be observed. The autocorrelation function of each modal function and the cross-correlation function with other modal functions are calculated. The typical characteristics of noise are determined, for example, the noise is manifested as high-frequency, low-amplitude fluctuations, or there is no obvious periodic structure in the frequency domain. According to the determined noise characteristics, a series of thresholds are set to identify noise modal functions. The thresholds include the maximum allowable value of the standard deviation and the energy proportion of high-frequency components in the frequency domain. Starting from the first modal function, each modal function is checked one by one. For each modal function, its statistical characteristics, frequency domain characteristics, and autocorrelation or cross-correlation analysis results are compared with the set thresholds. If the characteristics of a modal function exceed or fall below the set threshold, it is marked as a noise modal function. Initialize an empty set to store the remaining modal functions after screening, traverse all modal functions, and add the modal functions that are not marked as noise to the remaining modal function set.

[0090] In a preferred embodiment of the present invention, the above step 14 may include:

[0091] Decomposing the residual mode function into a plurality of mode function components to obtain mode function components;

[0092] The modal function components;

[0093] pass Perform an addition operation to perform the reconstruction operation, where is the signal after reconstruction; represents the amplitude coefficient; represents the frequency parameter; Indicates phase shift; Represents the value of the input signal; is a constant term; represents the Gaussian function; Indicates time; is the Gaussian center position; Indicates the width of the distribution; is an overall offset constant; is the control offset term with the total number of mode functions exponential decay rate of change; represents the base of natural logarithms;

[0094] According to the reconstruction operation, a quality check is performed on the obtained denoised electric energy meter operation data to obtain the denoised electric energy meter operation data.

[0095] In an embodiment of the present invention, the residual mode function is decomposed into multiple mode function components, which helps to understand the internal structure and characteristics of the electric energy meter operation data in more detail. The mode function components are reconstructed through a specific addition operation formula, and the denoised electric energy meter operation data can be accurately restored. In the reconstruction process, the reconstruction effect can be optimized by adjusting parameters such as weight coefficients and filter responses. The denoised electric energy meter operation data after quality inspection is cleaner and more accurate, and the interference of noise and outliers is effectively removed. This provides a high-quality data foundation for subsequent data analysis, fault diagnosis and prediction, and improves the reliability and accuracy of the analysis. Based on the denoised electric energy meter operation data, the operating status and performance of the electric energy meter can be more accurately evaluated, providing strong decision support for the stable operation of the power system and energy conservation and emission reduction.

[0096] In a specific embodiment of the present invention, commonly used modal decomposition methods include empirical mode decomposition or its variants, such as ensemble empirical mode decomposition; these methods can decompose complex signals into several simple modal function components. The residual mode function is used as input, and the selected modal decomposition algorithm is applied. The algorithm iteratively separates different frequency components in the signal to form multiple modal function components. Each modal function satisfies two conditions: in the entire data set, the number of extreme values ​​(including maximum and minimum values) differs from the number of zero crossings by at most one, and at any point, the average value of the upper envelope defined by the local maximum value and the lower envelope defined by the local minimum value is zero; check whether the modal function components obtained by decomposition meet the above conditions. Based on the energy of the function component, a weight coefficient is assigned to each modal function component, and a filter response function is selected or designed, which can adjust each modal function component according to its frequency characteristics, wherein the filter can be a low-pass, high-pass, band-pass or band-stop type.

[0097] The reconstructed signal is obtained by using the formula for reconstruction. The original electricity meter operation data is compared with the reconstructed data by drawing charts, calculating error indicators (such as mean square error MSE), etc., observing the differences between the two, ensuring that the reconstructed data does not lose any important information, such as key timestamps, electricity readings, etc. At the same time, check whether there are outliers or unreasonable data points in the data. According to the specific application requirements, evaluate whether the denoising effect meets expectations, for example, through quantitative indicator evaluation (such as statistical analysis of noise levels) or qualitative visual inspection, such as comparing waveforms to check whether the reconstructed data is smoother and the noise level is reduced. If the denoising effect is found to be unsatisfactory, you can go back to the previous steps to adjust the parameters, such as changing the algorithm parameters of the modal decomposition, adjusting the weight coefficient, or modifying the filter response function, and find the best denoising solution through repeated iterations and optimization.

[0098] In a preferred embodiment of the present invention, the above step 15 may include:

[0099] The denoised electric energy meter operation data is sorted to obtain a sorted data set;

[0100] Based on the sorted data set, an information table is constructed;

[0101] According to the information table, simplify the data set by reducing the attributes;

[0102] Based on the simplified data set;

[0103] pass Calculate the kernel attributes and classify them according to the kernel attributes to obtain data subsets of each attribute, where: Representation sample and The similarity between and Represents two different sample features in the data set is the width parameter of the Gaussian kernel; is the number of dimensions of the feature; Indicates Features The variance of It is for Nonlinear transformation of the distance between two samples in feature dimensions; Indicates the data set The sample The value of a feature; Indicates the data set The sample The value of a feature; Indicates the first Features is an index variable used to traverse all features in the calculation of feature weights; Represents the total number of features in the dataset; is an index variable used to traverse all features in the calculation of the kernel function; is the index of the sample; is the index of the sample;

[0104] According to the data subset of each attribute, The roughness of the data subsets of each attribute is calculated to evaluate the roughness of the data, where: is the roughness of the data subset, is the number of samples in the data subset; Is the first in the data subset i The attribute value of each sample; is the mean of the data subset;

[0105] The minimum roughness is obtained by comparing the roughness corresponding to different attribute subsets.

[0106] In the embodiment of the present invention, the denoised electric energy meter operation data is sorted to obtain a data set with a clear structure and easy management, which helps to improve the efficiency of data retrieval, storage and processing. By constructing an information table, the sorted data set can be presented in a more intuitive way, which is convenient for users to understand and analyze. Attribute reduction can remove redundant attributes in the data set, simplify the data set structure, reduce the amount of calculation in the data processing process, and improve the efficiency and accuracy of data processing. By calculating the core attributes and classifying them, data subsets of each attribute can be obtained, which can more accurately divide the data category and improve the performance of data classification. The classification based on the core attributes can also help to discover the similarity and correlation between data and provide support for more in-depth data analysis. By calculating the roughness of the data subsets of each attribute, the roughness of the data can be quantitatively evaluated, which helps to understand the distribution and fluctuation range of the data, and provides a strong basis for data cleaning, preprocessing and subsequent analysis. By comparing the roughness corresponding to different attribute subsets, the minimum roughness can be obtained, thereby determining the optimal data processing strategy.

[0107] In a specific embodiment of the present invention, the de-noised electric energy meter operation data is collected from different data sources to ensure that the collected data contains all necessary fields, such as timestamp, electric energy reading, voltage value, current value, etc. Duplicate records are identified and deleted, missing values ​​are processed or filled, and outliers are detected and corrected. Non-numeric fields in the data are converted into numerical form (such as converting categorical data into unique hot encoding), standardization or normalization techniques are applied to eliminate dimensional differences and make different features comparable, and the data is divided into training sets, validation sets, and test sets; the structure of the information table is defined using a database or data processing framework, including column names, data types, etc.; the pre-processed data is imported into the information table, and storage media and index strategies are selected according to the access frequency and importance of the data to optimize query performance. Feature selection techniques (such as filtering, packaging, or embedding) are applied to evaluate the importance of each attribute to the target variable, remove those attributes that are low in correlation or redundant with the target variable, and use attribute reduction algorithms such as those in rough set theory to further reduce the feature set by calculating the dependency of the attributes, and ensure that the reduced attribute set can still retain the main information of the original data set. For each pair of samples in the simplified data set, the Gaussian kernel function is used to calculate the similarity between them, and the width parameter σ of the Gaussian kernel is adjusted to find the best similarity measure. The calculated similarity values ​​are filled into a matrix to form a similarity matrix, which will represent the similarity relationship between the samples in the data set. The K-means clustering algorithm is applied to cluster the samples in the data set into different clusters based on the similarity matrix. Each cluster represents a data subset with similar characteristics. For each data subset obtained by clustering, its roughness is calculated using the formula. The roughness formula measures the degree of discreteness of the attribute values ​​in the data subset. The smaller the value, the more concentrated and smoother the data. Compare the roughness of different data subsets to find the data subset with the smallest roughness.

[0108] In a preferred embodiment of the present invention, the above step 16 may include:

[0109] The minimum roughness is used as the root node of the decision tree, and the meter operation data is divided into different subsets according to the minimum roughness, and each subset corresponds to a branch;

[0110] For each subset after division, create a new child node and add it to the child node list of all nodes after verification to recursively build the subtree;

[0111] When the process of recursively constructing the subtree is terminated, some leaf nodes are generated, and error estimation values ​​are assigned to the leaf nodes to construct a decision tree for the error estimation of the electric meter operation.

[0112] In an embodiment of the present invention, by using the minimum roughness as the partitioning criterion, it can be ensured that each data partition is based on the most discriminative features, and the partitioning strategy based on the minimum roughness can minimize the category mixing within each subset, which means that the data purity is higher at each branch and leaf node. Therefore, when predicting the operation error of the meter, the decision tree is known for its intuitive and easy-to-understand tree structure, which can easily understand how the model makes predictions based on the input features. In the scenario of meter operation error estimation, it helps to quickly locate the factors that may cause the error and take corresponding corrective measures. By recursively constructing subtrees, the model can automatically explore the deeper structure of the data without manual intervention. This automated process not only improves the efficiency of modeling, but also ensures that the model can capture the complex patterns and relationships in the data. The error estimate is assigned to the leaf node, providing additional confidence information for the prediction result. This enables users to not only obtain the predicted value of the meter operation error, but also understand the degree of uncertainty of the predicted value, so as to make more informed decisions in practical applications. By understanding the estimated value of the prediction error, power system operators can better manage potential risks.

[0113] In a specific embodiment of the present invention, for each feature in the meter operation data, its roughness is calculated. Roughness is an indicator to measure the purity of data partitioning, and can be calculated using methods such as information entropy and Gini index. The feature with the minimum roughness is selected as the partition attribute of the root node. A root node is created, and the entire meter operation data set is assigned to the node. The root node will be partitioned according to the selected minimum roughness feature. For the selected minimum roughness feature, one or more partition thresholds are determined. These thresholds will be used to partition the data into different subsets. For numerical features, the median, mean, etc. can be selected as thresholds; for categorical features, the partitioning is based on the category value. According to the determined partition threshold, the meter operation data is divided into different subsets. Each subset corresponds to a branch under the root node and contains sample data with the same feature value. For each divided subset, a new child node is created and added to the child node list of the verified node.

[0114] The child node will inherit the data subset and partitioning conditions of its parent node. For each newly generated child node, the above is repeated, that is, further data partitioning and child node generation are performed according to the minimum roughness until the recursive termination condition is met. These conditions may include that the amount of data in the child node is less than a preset threshold, the output values ​​of all samples are the same (i.e., the purity requirement is met), or the preset maximum tree depth is reached. When the process of recursively constructing a subtree terminates, a series of leaf nodes will be generated. These leaf nodes correspond to the end of the decision tree and no further data partitioning is performed. For each leaf node, an error estimate is calculated and assigned based on the meter operation data samples it contains. This error estimate can be calculated based on the difference between the actual operation data and the predicted value, for example, using indicators such as mean square error and mean absolute error. After the decision tree is built, some post-processing operations can be performed to optimize the performance, such as pruning the tree to reduce the risk of overfitting, or smoothing the error estimate of the leaf node to improve the prediction stability.

[0115] In a preferred embodiment of the present invention, the above step 17 may include:

[0116] According to the constructed decision tree, the roughness attribute of the operating data is traversed from the root node along the path of the decision tree until a leaf node is reached to obtain a leaf node, wherein the value corresponding to the leaf node is the error estimation result of the operating data.

[0117] In the embodiment of the present invention, the new meter operation data can be quickly traversed by using the already constructed decision tree. The path from the root node to the leaf node directly corresponds to the result of the error estimation, which helps to find problems and handle them in time. The traversal process of the decision tree is fully automated and does not require human intervention. Once the decision tree is built, it can automatically process the new data and output the error estimation result, which greatly improves the processing efficiency and accuracy. Compared with other complex machine learning models, the computational complexity of the decision tree in the traversal process is relatively low. This means that when processing a large amount of meter operation data, the decision tree can provide faster response time and lower computing resource consumption. The traversal process of the decision tree is intuitive and easy to understand, which is convenient for engineers to understand and implement. This simplicity makes the decision tree easier to be accepted and deployed in practical applications, and reduces the technical threshold. Since the decision tree is built based on data features, it has a certain robustness to data outliers and noise. In the meter operation data, this robustness helps to reduce the error estimation bias caused by data quality problems. When new meter operation data is added, the decision tree can be easily updated and adjusted without rebuilding the entire model. This incremental learning capability enables decision trees to continuously adapt to changes in data and maintain the accuracy of error estimates.

[0118] In a specific embodiment of the present invention, it is ensured that there is already a constructed decision tree, which is a tree structure including a root node, an internal node (branch node) and a leaf node. Each node includes a partition attribute (such as roughness) and a partition threshold (for numerical attributes) or a partition category (for category attributes). Prepare the running data to be traversed. These data should contain the same features used when building the decision tree, especially the roughness attribute, because it is the key in the traversal process. Start traversal from the root node of the decision tree. The root node is the starting point of the decision tree and contains the first partition of the entire data set. During the traversal process, first check the type of the current node. If the current node is a leaf node, directly read the error estimate of the node and end the traversal. If the current node is an internal node (branch node), proceed to the next step. Obtain the partition attribute and the corresponding partition threshold (for numerical attributes) or partition category (for category attributes) from the current internal node. This information will be used to determine the traversal direction of the next step. Compare the roughness attribute of the running data with the partition attribute of the current node. If the roughness attribute value is less than the partition threshold (for numerical attributes), or is equal to a partition category (for categorical attributes), the corresponding left subtree or right subtree (or a specific subtree, which may have multiple branches for categorical attributes) is selected as the direction of the next traversal. According to the decision result of the previous step, move to the next node and repeat steps 2 and 3 until a leaf node is reached. During the traversal process, the path from the root node to the leaf node can be recorded, and the traversal stops when the traversal reaches a leaf node. Read the error estimate from the leaf node. This value is calculated based on the training data contained in the leaf node during the decision tree construction process, and represents the expected error level when the roughness attribute of the running data meets the conditions of the leaf node. The read error estimate is returned as the result of the traversal. This value can be used to evaluate the accuracy of the running data or perform other subsequent analysis.

[0119] like Figure 2 As shown, an embodiment of the present invention further provides an electric meter operation error estimation and judgment system, comprising:

[0120] A collection module is used to obtain the operation data of the electric meter;

[0121] A processing module is used to perform EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions; according to the characteristics of the intrinsic mode functions, the mode functions related to noise are identified and removed to obtain the residual mode functions; the residual mode functions are reconstructed to obtain the denoised meter operation data;

[0122] An analysis module, used for analyzing the denoised electric energy meter operation data to obtain the minimum roughness;

[0123] The operation module is used to construct a decision tree for estimating the operation error of the electric energy meter according to the minimum roughness; and to judge and calculate the roughness attribute according to the constructed decision tree to realize the operation error estimation of the electric energy meter.

[0124] It should be noted that the system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0125] The embodiment of the present invention further provides a computing device, comprising: a processor, a memory storing a computer program, wherein when the computer program is executed by the processor, the method described above is executed. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0126] The embodiment of the present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

Claims

1. A method for estimating and judging an electric meter operation error, characterized in that: The method comprises: Obtain meter operation data; Perform EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions; According to the characteristics of the intrinsic mode function, the mode function related to the noise is identified and removed to obtain the residual mode function; Reconstruct the residual mode function to obtain the denoised electric energy meter operation data; Analyze the denoised electric energy meter operation data to obtain the minimum roughness; According to the minimum roughness, a decision tree for estimating the operation error of the electric meter is constructed; According to the constructed decision tree, the roughness attributes are judged and calculated to achieve the operation error estimation of the electric energy meter; Perform EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions, including: By traversing the meter operation data, all local maximum points and local minimum points are obtained; According to the local extreme points, the upper envelope and the lower envelope are fitted respectively by cubic spline interpolation method to obtain two envelopes; According to the two envelopes, The mean envelope is calculated, where is the mean envelope at time point t The value of t It's the time point. It is in The original data value at each time point; is a time point index; n is the size of the sliding window, It's at the time t The maximum value in the sliding window of It's at the time t The minimum value in the sliding window of is the weight coefficient of the upper envelope, is the weight coefficient of the lower envelope; Subtract the mean envelope from the original meter operation data to obtain a series of intrinsic mode functions; According to the characteristics of the intrinsic mode function, the mode function related to the noise is identified and removed to obtain the remaining mode function, including: By analyzing the characteristics of each intrinsic mode function, the characteristics of the noise can be identified; Based on the characteristic analysis of noise, each modal function is checked one by one by setting a threshold to identify the noise modal function; The noise mode function is removed from the original set of intrinsic mode functions to obtain the residual mode function; Reconstruct the residual mode function to obtain the denoised electric energy meter operation data, including: Decomposing the residual mode function into a plurality of mode function components to obtain mode function components; The modal function components are passed through Perform an addition operation to perform the reconstruction operation, where is the signal after reconstruction; represents the amplitude coefficient; represents the frequency parameter; Indicates phase shift; Represents the value of the input signal; is a constant term; represents the Gaussian function; Indicates time; is the Gaussian center position; Indicates the width of the distribution; is an overall offset constant; is the total number of control offset terms with the modal function exponential decay rate of change; represents the base of natural logarithms; According to the reconstruction operation, quality inspection is performed on the obtained denoised electric energy meter operation data to obtain the denoised electric energy meter operation data; The denoised electric energy meter operation data is analyzed to obtain the minimum roughness, including: The denoised electric energy meter operation data is sorted to obtain a sorted data set; Based on the sorted data set, an information table is constructed; According to the information table, simplify the data set by reducing the attributes; According to the simplified data set, Calculate the kernel attributes and classify them according to the kernel attributes to obtain data subsets of each attribute, where: Representation sample and The similarity between and Represents two different sample characteristics in the data set; is the width parameter of the Gaussian kernel; is the number of dimensions of the feature; Indicates Features The variance of It is for Nonlinear transformation of the distance between two samples in feature dimensions; Indicates the data set The sample The value of a feature; Indicates the data set The sample The value of a feature; Indicates the first Features is an index variable used to traverse all features in the calculation of feature weights; is an index variable used to traverse all features in the calculation of the kernel function; is the index of the sample; is the index of the sample; According to the data subset of each attribute, The roughness of the data subset of each attribute is calculated to evaluate the roughness of the data, where: is the roughness of the data subset, is the number of samples in the data subset, Is the first in the data subset i The attribute value of each sample; is the mean of the data subset; By comparing the roughness corresponding to different attribute subsets, the minimum roughness is obtained; Based on all the nodes after inspection, a decision tree for estimating the operation error of the electric meter is constructed, including: The minimum roughness is used as the root node of the decision tree, and the meter operation data is divided into different subsets according to the minimum roughness, and each subset corresponds to a branch; For each partitioned subset, create a new child node and add it to the child node list of all nodes after verification to recursively build the subtree; When the process of recursively constructing the subtree is terminated, some leaf nodes are generated, and error estimation values ​​are assigned to the leaf nodes to construct a decision tree for estimating the operation error of the electric meter; According to the constructed decision tree, the roughness attributes are judged and calculated to realize the operation error estimation of the electric energy meter, including: According to the constructed decision tree, the roughness attribute of the operating data is traversed from the root node along the path of the decision tree until a leaf node is reached to obtain a leaf node, wherein the value corresponding to the leaf node is the error estimation result of the operating data.

2. A system for estimating and judging an electric meter operation error, characterized in that: Applied in the method as claimed in claim 1, comprising: A collection module is used to obtain the operation data of the electric meter; A processing module is used to perform EMD decomposition on the meter operation data to obtain a series of intrinsic mode functions; according to the characteristics of the intrinsic mode functions, the mode functions related to noise are identified and removed to obtain the residual mode functions; the residual mode functions are reconstructed to obtain the denoised meter operation data; An analysis module, used for analyzing the denoised electric energy meter operation data to obtain the minimum roughness; The operation module is used to construct a decision tree for estimating the operation error of the electric energy meter according to the minimum roughness; and to judge and calculate the roughness attribute according to the constructed decision tree to realize the operation error estimation of the electric energy meter.

3. A computing device, characterized in that: include: one or more processors; A storage device, used for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in claim 1.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which implements the method according to claim 1 when executed by a processor.

Images