Abnormal waveform matching method for voltage abnormity identification of electric energy meter

Abstract

The invention discloses an abnormal waveform matching method for electric energy meter voltage abnormity identification, and relates to the technical field of waveform matching, and the method comprises the steps: collecting a historical abnormal waveform, extracting a feature vector, and constructing an abnormal waveform dictionary through a dynamic dictionary learning algorithm; performing coverage analysis on historical abnormal waveforms, and screening core features to establish a simplified feature set; collecting a to-be-recognized voltage waveform in real time, extracting a feature vector to be matched with the simplified feature set, and primarily judging whether dominant anomaly exists or not; if yes, candidate template types are screened through the characteristic deviation ratio, and abnormal types are judged; if the dominant anomaly does not exist, a normal waveform feature baseline model is constructed, deviation analysis is carried out to judge whether the implicit anomaly exists or not, and if the implicit anomaly exists, the abnormal waveform dictionary is iteratively optimized in combination with the to-be-recognized waveform features; the method can effectively identify the voltage abnormity of the electric energy meter, balance the identification precision and the calculation efficiency, and adapt to a novel abnormal mode and equipment gradual change fault characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of waveform matching, and in particular to an abnormal waveform matching method for identifying abnormal voltage of an electric energy meter. Background Art

[0002] In the field of smart grid, accurate recognition of voltage anomalies by electric energy meters is the key to ensuring stable operation of distribution networks. In the prior art, the methods for identifying voltage anomalies by electric energy meters have the following problems that need to be solved.

[0003] Traditional methods mostly make anomaly judgments based on a single type of feature (such as only the frequency domain harmonic amplitude or the time domain waveform amplitude), which makes it difficult to characterize the characteristics of complex voltage anomalies, such as the time domain morphology of transient voltage surges or dips, the frequency domain distribution of harmonic distortion, and the coupling characteristics of time-frequency domain energy changes. As a result, the existing technology has a high missed judgment rate for complex anomalies or atypical anomalies.

[0004] Existing anomaly recognition relies on fixed thresholds or static templates, which cannot adapt to the waveform abnormality patterns caused by high-frequency oscillations and non-integer harmonic new voltages introduced by new equipment in the distribution network (such as new energy inverters and energy storage devices). It lacks the ability to autonomously learn and iterate dictionaries for unknown anomalies, resulting in a lag in the recognition of new anomalies.

[0005] To this end, the present invention provides an abnormal waveform matching method for identifying abnormal voltage of an electric energy meter. Summary of the invention

[0006] The object of the present invention is to provide an abnormal waveform matching method for identifying abnormal voltage of an electric energy meter, so as to solve at least one of the above-mentioned problems of the prior art.

[0007] An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter comprises the following steps: Collect historical abnormal waveforms of the electric energy meter, perform feature decomposition on the historical abnormal waveforms and construct an abnormal waveform dictionary; Perform coverage analysis on historical abnormal waveforms, obtain feature priority values, and filter and process them to obtain a streamlined feature set; The voltage waveform to be identified of the electric energy meter is collected in real time, the feature extraction of the voltage waveform to be identified is used to construct a waveform feature vector, the waveform feature vector is matched and calculated with the simplified feature set, and it is initially determined whether there is an obvious abnormality in the voltage waveform to be identified; If there is an obvious abnormality, the abnormality type is screened to obtain a candidate template type, and an abnormality determination analysis is performed on the candidate template waveform prototype and the voltage waveform to be identified to obtain the abnormality determination coefficient and finally determine the abnormality type of the electric energy meter; If it is initially determined that there is no obvious abnormality in the voltage waveform to be recognized, it is determined whether there is a hidden abnormality in the voltage waveform to be recognized. If so, the abnormal waveform dictionary is iteratively optimized in combination with the characteristics of the voltage waveform to be recognized.

[0008] As a further technical solution of the present invention, the method for constructing the abnormal waveform dictionary is as follows: Collect historical abnormal waveforms of the electric energy meter, perform feature decomposition on the historical abnormal waveforms, and obtain an identification feature vector including frequency domain features, time-frequency domain features, and morphological features; Based on the dynamic dictionary learning algorithm, construct an abnormal waveform dictionary including an identification feature vector, an abnormal type, and a waveform prototype.

[0009] As a further technical solution of the present invention, the method for obtaining the reduced feature set is as follows: Take the electric energy meter to be recognized as a sample for abnormal identification, obtain historical abnormal waveforms from the abnormal waveform dictionary, and classify and correspond the historical abnormal waveforms according to the sample; Obtain the feature coverage and feature discriminability of each sample and perform a summation process to obtain a feature priority value; Sort the sub-features based on the feature priority value, extract the core features from the abnormal waveform dictionary, and construct a reduced feature set based on the feature vectors corresponding to the core features.

[0010] As a further technical solution of the present invention, the method for obtaining the feature coverage and feature discriminability is as follows: Calculate the occurrence frequency of different sub-features of each sample to obtain the feature coverage; Obtain the between-class scatter and within-class scatter of the sample, and perform a ratio calculation to obtain the feature discriminability.

[0011] As a further technical solution of the present invention, the method for initially determining whether there is an obvious abnormality in the voltage waveform to be recognized is as follows: Perform waveform preprocessing on the voltage waveform to be recognized collected in real time, perform feature extraction on the waveform after waveform preprocessing, and construct a waveform feature vector; Extract the identification feature vector of each abnormal type from the reduced feature set, calculate the matching degree between the identification feature vector and the waveform feature vector of each abnormal type to obtain a feature matching degree; Compare the feature matching degree with a preset feature matching degree threshold to obtain a result of determining that there is an obvious abnormality in the voltage waveform to be recognized.

[0012] As a further technical solution of the present invention, the method for performing the matching degree calculation is as follows: Calculate the cosine similarity between the identification feature vector and the waveform feature vector to obtain the feature matching degree.

[0013] As a further technical solution of the present invention, the method for finally determining the abnormal type of the electric energy meter is as follows: If there is a corresponding abnormal type, calculate the deviation ratio of the feature matching degree of each abnormal type to the feature matching degree threshold to obtain the feature deviation ratio; Based on the feature deviation ratio, screen the abnormal types to obtain the candidate template types; Obtain the timing matching degree and feature matching degree of the candidate template type, and perform weighted summation processing on the timing matching degree and the feature matching degree to obtain the abnormal decision coefficient; Based on the abnormal decision coefficient, obtain the abnormal type of the candidate template type, and finally determine whether there is a dominant abnormal of the abnormal type in the voltage waveform to be recognized.

[0014] As a further technical solution of the present invention, the method for obtaining the timing matching degree is as follows: Extract the waveform prototype of the candidate template type from the abnormal waveform dictionary and the voltage waveform to be recognized collected by the electric energy meter, and perform dynamic time warping calculation on the waveform prototype and the voltage waveform to be recognized to obtain the timing matching degree.

[0015] As a further technical solution of the present invention, the method for obtaining the weight coefficient of the abnormal decision coefficient is as follows: Perform transformation processing on the waveform prototype and the voltage waveform to be recognized to generate a time-frequency spectrogram, and extract the transient energy distribution entropy; Calculate the absolute deviation ratio of the transient energy distribution entropy of the waveform prototype and the voltage waveform to be recognized as the weight coefficient.

[0016] As a further technical solution of the present invention, the method for determining whether there is a hidden abnormality in the voltage waveform to be recognized is as follows: Construct a normal waveform feature baseline model; Perform multi-dimensional feature deviation calculation to mark potential abnormal features; Perform dynamic trend analysis on the potential abnormal features to determine whether there is a trend abnormality; If there is a trend abnormality, combine the feature discriminability and the standardized deviation to determine the hidden abnormality.

[0017] Advantages of the present invention: 1. Through multi-scale variational mode decomposition, morphological filtering and S transform, extract the frequency domain, time-frequency domain and morphological composition recognition feature vectors from historical abnormal waveforms, and combine the dynamic dictionary learning algorithm for clustering analysis to construct an abnormal waveform dictionary including feature vectors, waveform prototypes and abnormal types, so as to realize the standardized modeling of multiple types of abnormalities such as overvoltage, undervoltage, harmonic distortion, voltage sag, etc., covering common and transient voltage abnormal scenarios.

[0018] 2. By calculating the feature coverage (frequency of occurrence in fault samples) and feature discriminability (ratio of between-class difference to within-class fluctuation), and summing them after normalization to obtain the feature priority value, the core features that appear frequently and are easy to distinguish are screened to construct a reduced feature set. Calculating the core features on the electricity meter is beneficial to improving the abnormal recognition efficiency of the electricity meter, realizing the lightweight deployment of complex feature vectors, and adapting to the scenario of limited resources of edge devices.

[0019] 3. Coarse screening is carried out by calculating the cosine similarity through the reduced feature set. For samples with insufficient matching degree, the complete dictionary is called for dynamic time warping (DTW) and time-frequency feature difference degree constraint, and the weight is dynamically adjusted according to the abnormal type, which is beneficial to improving the accuracy of abnormal recognition; if there is no obvious abnormality, a normal waveform feature baseline model is constructed to identify the early abnormality of continuous deviation of feature values, realizing the extension from threshold-based explicit detection to trend-based implicit warning. For the detected implicit abnormality or unknown abnormality, it is compared with the existing template through the density clustering algorithm. If it is a known abnormal variation, the template parameters are updated, and if it is a new type of abnormality, a new template entry is added; a closed loop of data acquisition - abnormal detection - dictionary update is formed to improve the adaptability of the electricity meter to the dynamic changes of abnormal waveforms. Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0021] Figure 1 It is a flowchart of an abnormal waveform matching method for voltage abnormality recognition of an electricity meter provided by the present invention; Figure 2 It is a flowchart of implicit recognition provided by the present invention; Figure 3 It is a module schematic diagram of an abnormal waveform matching method for voltage abnormality recognition of an electricity meter provided by the present invention. Detailed Embodiments

[0022] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] Embodiment 1: As Figure 1As shown in the figure, an abnormal waveform matching method for identifying abnormal voltage of an electric energy meter provided by an embodiment of the present invention specifically includes the following steps: Step 1: Collect historical abnormal waveforms of the electric energy meter, perform feature decomposition on the historical abnormal waveforms, and construct an abnormal waveform dictionary; Among them, the method of performing feature decomposition on the historical abnormal waveforms and constructing an abnormal waveform dictionary is as follows: Preferably, read the historical abnormal voltage waveforms from the log of the voltmeter, perform multi-scale variational mode decomposition on the original waveforms, decompose the complex voltage signals into multiple intrinsic mode functions, extract the eigenmode components of the transient voltage sudden rise and sudden drop abnormal waveforms, and use morphological filtering to enhance the waveform edge features (such as repairing depressions and sharpening mutation points); Generate a time-frequency spectrogram from the waveform after morphological filtering, and extract an identification feature vector including frequency domain features, time-frequency domain features, and morphological features: Those skilled in the art can understand that, first perform morphological filtering on the voltage waveform of the electric energy meter, perform opening and closing operations by selecting appropriate structural elements (such as rectangles, Gaussian kernels, etc.) to filter out noise and retain the waveform contour features; then perform S transform on the filtered waveform, and convert the time-domain signal into a time-frequency spectrogram by constructing a Gaussian window function that adaptively adjusts with frequency to characterize the energy distribution of the signal at different time-frequency points; finally, extract frequency domain features (such as the amplitude, phase, and energy ratio of each frequency component), time-frequency domain features (such as time-frequency energy entropy, time-frequency distribution variance, and frequency change rate over time) from the time-frequency spectrogram, and extract morphological features (such as the distance between the peak and valley, the slope of the rising or falling edge, and the number of zero-crossing points) from the waveform after morphological filtering, and splice the three types of features in a preset order to form an identification feature vector containing multi-dimensional information; It should be explained that the frequency domain features, time-frequency domain features, and morphological features include multiple sub-features; The frequency domain features are composed of three sub-features: fundamental wave amplitude, harmonic amplitude ratio, and total harmonic distortion; The time-frequency domain features are composed of two sub-features: transient energy distribution entropy and instantaneous frequency fluctuation coefficient; The morphological features are composed of three sub-features: peak slope, phase jump point position, and offset; Use the dynamic dictionary learning algorithm to perform clustering analysis on the identification feature vectors of the same type of abnormality, screen each clustering center as a standard feature template, and synchronously retain the time series data containing the complete voltage waveform, the abnormality type, and the time-domain curve of the corresponding prototype, and construct an abnormal waveform dictionary containing the identification feature vector corresponding to the abnormality type, the abnormality type, and the waveform prototype; Those skilled in the art can understand that when constructing an abnormal waveform dictionary using the dynamic dictionary learning algorithm, first, the recognition feature vectors of the same type of abnormalities are subjected to clustering analysis using a dynamic clustering method (such as the K-means++ algorithm). The sample assignment and center update are achieved by iteratively calculating the distance between the feature vectors and the clustering centers, and the number of clusters is dynamically adjusted according to the clustering quality to reduce local optimality. Then, each stable clustering center is selected from the clustering results as a standard feature template. At the same time, the robustness of the template is ensured through outlier detection, and a template update mechanism triggered regularly or by a threshold is established; The selected standard feature templates are associated and stored with the corresponding complete voltage waveform time series data, abnormal type labels, and time domain curves, constructing an abnormal waveform dictionary that contains multi-dimensional information such as recognition feature vectors, abnormal types, and waveform prototypes; Among them, the role of constructing the abnormal waveform dictionary is as follows: Function 1. Provide a standardized feature template library: By dynamically clustering and selecting each clustering center as a standard feature template, the recognition feature vectors of the same type of abnormalities are abstracted and standardized, providing a unified feature matching benchmark for the initial judgment of obvious abnormalities and the final judgment of abnormal types of real-time waveforms, ensuring the comparability of abnormal features in different time periods and different scenarios; Function 2. Support dynamic learning and iterative optimization: Synchronously store the complete waveform time series data and time domain curves corresponding to the abnormal types, not only retaining the abstract information of the feature vectors but also retaining the detailed features of the original waveforms, providing a data basis for the dictionary iteration module when detecting latent abnormalities. By dynamically updating the dictionary by combining new abnormal features, continuous learning of new abnormal patterns and expansion of the knowledge base are realized; Function 3. Realize multi-dimensional abnormal information association mapping: Structurally associate and store the recognition feature vectors, abnormal types, and waveform prototypes, constructing a mapping relationship of "feature - type - waveform", which not only supports fast matching calculations based on feature vectors (such as feature vector matching in obvious initial judgment), but also can trace back to the specific waveform prototype for detailed comparison during the final judgment of abnormalities (such as abnormal decision analysis between the candidate template and the waveform to be recognized), providing cross-dimensional data support for the multi-level abnormal recognition process and improving the recognition accuracy and decision reliability.

[0024] Step 2. Conduct a coverage analysis of historical abnormal waveforms, obtain the feature priority values, screen the core features of the abnormal waveform dictionary, and establish a refined feature set; Among them, the method of conducting a coverage analysis of historical abnormal waveforms and obtaining the feature priority values is as follows: Take the electricity meter to be recognized as a sample for abnormal recognition, obtain historical abnormal waveforms from the abnormal waveform dictionary, and classify and correspond the historical abnormal waveforms according to the sample; Calculate the occurrence frequencies of different sub-features of each sample to obtain the feature coverage; By formula: Get feature separability K f ; Where c represents the total number of abnormal types, that is, the number of fault categories covered in the abnormal waveform dictionary, such as overvoltage, undervoltage, harmonic distortion, and voltage sag; n i The number of samples representing the i-th abnormal type. For example, the i-th type “voltage sag” contains n i Historical abnormal waveform samples; It represents the mean of the i-th sample on the f-th sub-feature, for example, the mean of the third harmonic amplitude ratio of the i-th sample. f is the sub-feature number, which is used to traverse all sub-features in the frequency domain, time-frequency domain, and morphology. For example, f=1 corresponds to the fundamental amplitude, and f=2 corresponds to the harmonic amplitude ratio. It represents the global mean of all samples on the f-th sub-feature, that is, the f-th seed feature mean of all historical samples, which is used as a benchmark for measuring inter-class differences; Represents the specific value of a single sample on the fth sub-feature, such as the measured value of the peak slope of a fault waveform, belonging to the i-th sample set c i Elements in c i represents the i-th type abnormal sample set, that is, the subset of all historical waveform data marked as the i-th type abnormality, which is used to calculate the intra-class dispersion; It needs to be explained that , Respectively represent the inter-class dispersion and intra-class dispersion of samples; Inter-class dispersion reflects the mean difference of different abnormal categories on the f-th seed feature. The larger the difference, the more effective the sub-feature is in distinguishing categories. Intra-class dispersion reflects the degree of fluctuation of samples in the same class on the fth sub-feature (summing up the deviations of samples in all classes. The smaller the fluctuation, the higher the consistency of the sub-feature within the class. Feature separability K f The larger the value is, the higher the separability of the f-th sub-feature is (significant differences between classes and small differences within classes), and the greater its contribution to distinguishing fault types in anomaly recognition. Among them, the method of screening the core features of the abnormal waveform dictionary is: Obtain the feature coverage and feature separability of all samples and perform Min-Max normalization processing on them respectively; The normalized feature coverage and feature separability are summed to obtain the feature priority value; It can be understood that the physical meaning of the feature priority value is that the feature priority value is the result of summing the normalized feature coverage and feature discriminability; The feature priority value reflects the general frequency of occurrence in the sample and embodies universality. The feature discriminability reflects the ability to distinguish fault categories and embodies discriminative power; The feature priority value lies in quantifying the comprehensive value of features in fault identification: it not only reflects the wide coverage of features for common faults (such as the high-frequency occurrence of THD in scenarios such as harmonics and overvoltage), but also embodies the distinction between different fault types (such as the significant mean difference of THD in harmonic faults compared with other faults); Through the feature priority value, core features that "appear frequently and are easy to distinguish" can be preferentially screened (such as THD in the frequency domain, peak slope in morphology); The feature priority value guides the lightweighting of the model, the dynamic optimization of the dictionary, and the tuning of the matching algorithm, and finally realizes an efficient data-driven mapping from historical abnormal waveform data to real-time abnormal identification. In the scenario where the resources of edge devices are limited, it balances the recognition accuracy and computational efficiency, providing a basis for feature screening and application with both excellent universality and discriminative power for the voltage anomaly detection of electric energy meters; Sort the sub-features based on the feature priority value, extract the TOP-F features from the abnormal waveform dictionary as core features, and extract the feature vectors of the core features to construct a refined feature set; Those skilled in the art can understand that the implementation process of constructing a refined feature set based on the feature priority value is as follows: First, calculate the feature priority value of each sub-feature in the abnormal waveform dictionary, then sort the sub-features from high to low according to the feature priority value, extract the top F (TOP-F) sub-features as core features, and finally extract the feature vectors corresponding to these core features from all abnormal types in the abnormal waveform dictionary, and splice these vectors in the preset dimension order to form the final refined feature set; Among them, the role of constructing the refined feature set is: Function 1. Improve real-time matching efficiency: Screen the TOP-F core features through the feature priority value, eliminate redundant or low-discriminability sub-features, which is beneficial to reducing the dimension and computational complexity of the feature vector, enabling the explicit preliminary judgment module to quickly complete the calculation and comparison of the waveform feature vector and the refined feature set during real-time matching, and meeting the real-time requirements of electric energy meter anomaly recognition; Function 2. Strengthen the discriminant ability of key features: Sort based on the ability of features to distinguish abnormal types (such as indicators such as information gain and mutual information), ensure that only the most representative core features are retained in the refined feature set, reduce the interference of noise features, improve the recognition accuracy of the explicit preliminary judgment for explicit anomalies, and reduce misjudgment or missed judgment; Function 3. Support hierarchical recognition process for hierarchical processing: The refined feature set serves as a fast screening layer for explicit preliminary judgment, forming a hierarchical architecture with the in-depth matching layer of the abnormal final judgment module - the preliminary judgment quickly locates potential abnormalities through core features, and the final judgment discriminates the abnormal type through the detailed features of the complete candidate template, realizing optimized resource allocation and improving the overall processing efficiency of the system while ensuring the recognition accuracy.

[0025] The technical solution of this embodiment is: Collect the historical abnormal waveforms of the electric energy meter, perform feature decomposition on the historical abnormal waveforms and construct an abnormal waveform dictionary; perform coverage analysis on the historical abnormal waveforms, obtain feature priority values and perform screening processing to obtain a refined feature set; which is beneficial to improving the abnormal recognition efficiency of the electric energy meter, realizing the lightweight deployment of complex feature vectors, and adapting to the scenario of limited resources of edge devices.

[0026] Embodiment 2: As Figure 1 shown, an abnormal waveform matching method for voltage abnormality recognition of an electric energy meter further includes the following steps: Step 3. Real-time collect the voltage waveform to be recognized of the electric energy meter, extract features from the voltage waveform to be recognized to construct a waveform feature vector, and perform matching calculation on the waveform feature vector and the refined feature set to initially determine whether there is an explicit abnormality in the voltage waveform to be recognized; Among them, the method of extracting features from the voltage waveform to be recognized to construct a waveform feature vector is: Perform waveform preprocessing on the voltage waveform to be recognized collected in real time, extract features from the waveform after waveform preprocessing, and construct a waveform feature vector; It should be explained that the implementation process of preprocessing and feature extraction for the voltage waveform collected in real time is: First, perform morphological filtering and Z-score normalization on the original waveform to eliminate noise interference and dimensionality effects; extract features from the preprocessed waveform, including time-domain features (such as mean, variance, number of zero crossings), frequency-domain features (amplitude and phase of each frequency component obtained by S transform), and time-frequency domain features (such as time-frequency energy entropy, frequency change rate); finally, splice the extracted multi-dimensional features into a fixed-dimensional feature vector in a preset order, and perform dimensionality reduction optimization (such as principal component analysis) to construct the final waveform feature vector. The waveform feature vector retains the key features of the waveform and reduces redundant information, providing a standard input for subsequent matching calculations; Extract the recognition feature vector of each abnormal type from the refined feature set, perform matching degree calculation on the recognition feature vector and the waveform feature vector of each abnormal type to obtain the feature matching degree; Those skilled in the art can understand that the feature matching degree is obtained by calculating the cosine similarity between the recognition feature vector and the waveform feature vector; Compare the feature matching degree with a preset feature matching degree threshold. If the feature matching degree is higher than or equal to the preset feature matching degree threshold, initially determine that there is an obvious abnormality in the voltage waveform to be recognized; If the feature matching degree is lower than the preset feature matching degree threshold, extract the recognition feature vectors of all abnormal types from the abnormal waveform dictionary, calculate the feature matching degrees of all abnormal types, and compare and process them with the feature matching degree threshold, initially determining that there is an obvious abnormality in the voltage waveform to be recognized; It can be understood that when the feature matching degree is higher than the threshold, it indicates that the core features of the waveform to be recognized highly coincide with the refined feature set. It is determined as an obvious abnormality and quickly enters the final judgment process, reducing the full feature matching for all abnormal types, reducing the real-time calculation overhead, and meeting the real-time requirements of abnormal recognition of electric energy meters; When the matching degree is lower than the threshold, there may be two situations. One is that the refined feature set (TOP-F core features) fails to cover the key abnormal features of the waveform (such as insufficient core feature discrimination or subtle differences in abnormal features), and the other is that the waveform itself may belong to marginal abnormalities or new types of abnormalities; By extracting the complete feature vectors of all abnormal types in the abnormal waveform dictionary and recalculating the matching degree, it can reduce the missed judgment caused by core feature screening, effectively identify "atypical obvious abnormalities", significantly improve the recall rate of abnormal detection at the expense of a small amount of computing resources, and achieve the complementarity of "fast preliminary screening" and "comprehensive verification"; If there is an obvious abnormality, obtain the corresponding abnormal type; If the feature matching degrees of all abnormal types are still lower than the preset feature matching degree threshold, continuously monitor the voltage waveform of the electric energy meter; It should be explained that the role of initially determining that there is an obvious abnormality in the voltage waveform to be recognized is as follows: Function 1. Realize fast pre-screening of abnormalities: Through the efficient matching calculation of the refined feature set and the waveform feature vector, quickly locate the abnormal waveforms with obvious feature matching, reduce the full feature in-depth analysis of all waveforms, improve the real-time performance of abnormal recognition, and meet the requirements of real-time monitoring of the operating status of electric energy meters; Function 2. Provide a candidate range for precise judgment of abnormal types: The obvious initial judgment result serves as a "candidate abnormal set" for preliminary screening, providing a clear candidate template type (such as preset abnormal types like overvoltage and undervoltage) for the subsequent abnormal final judgment module, narrowing the matching range during final judgment, reducing calculation redundancy, enabling the final judgment link to focus on the comparison of detailed features between the candidate template and the waveform to be recognized (such as the time-domain curve fitting degree of the waveform prototype and the calculation of the abnormal determination coefficient), and improving the efficiency and accuracy of abnormal type determination.

[0027] Step 4: If there is an obvious anomaly, screen and process the anomaly type to obtain candidate template types, perform anomaly decision analysis on the candidate template waveform prototype and the voltage waveform to be recognized, obtain the anomaly decision coefficient, and finally determine the anomaly type existing in the electric energy meter; If there is a corresponding anomaly type, calculate the deviation ratio of the feature matching degree of each anomaly type to the feature matching degree threshold to obtain the feature deviation ratio; Based on the feature deviation ratio, screen the anomaly types to obtain candidate template types; It can be understood that the feature deviation ratios are sorted from small to large, and the first several (such as the first 3) anomaly types with the smallest deviation are retained as candidate templates, that is, it is considered that these templates are the most similar to the features of the waveform to be recognized and are used as the candidate range for further recognition; Extract the waveform prototype of the candidate template type from the abnormal waveform dictionary, as well as the voltage waveform to be recognized collected by the electric energy meter, and perform dynamic time warping (DTW) calculation on the waveform prototype and the voltage waveform to be recognized to obtain the timing matching degree; Preferably, through the formula: Obtain the timing matching degree ; It should be explained that D min : represents the dynamic time warping (DTW) distance between the voltage waveform to be recognized and the waveform of a certain anomaly type template in the abnormal waveform dictionary, measuring the difference degree of their timing characteristics, and the smaller the value, the closer the timing shape; max_d represents the preset maximum allowable distance threshold, which is used for normalization to ensure that the calculation result of the matching degree is within the range of [0,1], representing the maximum possible distance between this anomaly type template and other waveforms, and is used as the denominator to unify the dimension; The larger the value of the timing matching degree, the higher the matching degree of the timing characteristics between the waveform to be recognized and the template waveform. When D min = 0, the timing matching degree = 1, indicating a perfect match; Obtain the timing matching degree and feature matching degree of the candidate template type, and perform weighted summation processing on the timing matching degree and feature matching degree to obtain the anomaly decision coefficient; Based on the anomaly decision coefficient, obtain the anomaly type of the candidate template type, and finally determine whether there is an obvious anomaly of the anomaly type in the voltage waveform to be recognized; Among them, the method for finally determining whether there is an obvious anomaly of the anomaly type in the voltage waveform to be recognized is: Compare the anomaly decision coefficient with the preset anomaly decision threshold. If the anomaly decision coefficient is higher than or equal to the preset anomaly decision threshold, it is considered that there is an obvious anomaly of the anomaly type in the voltage waveform to be recognized; Among them, the method for obtaining the weight coefficient of the abnormal decision coefficient is as follows: Perform an S-transform on the waveform prototype and the voltage waveform to be recognized, generate a time-frequency spectrogram, and extract the transient energy distribution entropy; Normalize the transient energy distribution entropy, and calculate the proportion of the absolute deviation of the transient energy distribution entropy of the waveform prototype and the voltage waveform to be recognized as the weight coefficient.

[0028] It should be explained that the transient energy distribution entropy values of all waveform prototypes in the abnormal waveform dictionary and the transient energy distribution entropy value of the voltage waveform to be recognized are collected, the minimum and maximum values of the entropy values in the dictionary are calculated, and each entropy value (including the prototype and the waveform to be recognized) is subjected to Min-Max normalization processing; Calculate the absolute deviation between the normalized entropy value of the waveform to be recognized and the normalized entropy value of each waveform prototype to obtain the proportion of the absolute deviation; Finally, convert the proportion of the absolute deviation into a weight coefficient, so that the weight coefficient is in the range of [0,1]. The larger the weight value, the more similar the transient energy distributions of the two are, and the higher the contribution degree of this feature in the matching. Thus, the influence degree of the transient energy feature is dynamically adjusted during multi-feature fusion to improve the accuracy of abnormal recognition.

[0029] The technical solution of this embodiment is as follows: Real-time collect the voltage waveform of the electricity meter to be recognized, extract features from the voltage waveform to be recognized to construct a waveform feature vector, perform a matching calculation on the waveform feature vector and the refined feature set, and initially determine whether there is an obvious abnormality in the voltage waveform to be recognized; if there is an obvious abnormality, screen the abnormal type to obtain a candidate template type, perform an abnormal decision analysis on the candidate template waveform prototype and the voltage waveform to be recognized, obtain an abnormal decision coefficient, and finally determine the abnormal type existing in the electricity meter; it is beneficial to form a closed loop of data collection - abnormal detection - dictionary update, and improve the adaptability of the electricity meter to the dynamic changes of abnormal waveforms.

[0030] Embodiment 3: As Figure 1 shown, an abnormal waveform matching method for voltage abnormality recognition of an electricity meter further includes the following steps: Step 5: If it is initially determined that there is no obvious abnormality in the voltage waveform to be recognized, determine whether there is a hidden abnormality in the voltage waveform to be recognized. If so, iteratively optimize the abnormal waveform dictionary in combination with the characteristics of the voltage waveform to be recognized; As Figure 2 shown, the method for determining whether there is a hidden abnormality in the voltage waveform to be recognized is as follows: S1. Construct a normal waveform feature baseline model; Preferably, collect the historical normal waveforms of the same electricity meter, that is, there is no obvious abnormal record and the obvious abnormal record is higher than the preset fixed period, and extract the frequency domain, time-frequency domain, and morphological features of the electricity meter; Calculate the mean, standard deviation, and 95% quantile of each sub-feature in the frequency domain, time-frequency domain, and morphology of the electricity meter, and construct a dynamic baseline interval of normal features as the normal waveform feature baseline model; Adopt the sliding window algorithm to dynamically update the baseline interval to adapt to the periodic changes of the power grid load (such as the voltage fluctuation difference between morning and evening peaks); S2. Perform multi-dimensional feature deviation calculation to mark potential abnormal features; Preferably, if the sub-feature of the electricity meter deviates from the dynamic baseline interval, calculate the standardized deviation between the sub-feature and the dynamic baseline interval; Extract high-priority features based on the standardized deviation and mark the high-priority features as potential abnormal features; S3. Conduct dynamic trend analysis on potential abnormal features to determine whether there are trend abnormalities; Preferably, for the characteristic values of consecutive N periods (such as N = 10), use linear regression to calculate the trend slope. If the absolute value of the slope exceeds 3 times the baseline fluctuation rate, it is determined that there is a trend abnormality; S4. If there is a trend abnormality, combine the feature discriminability and the standardized deviation to conduct a hidden abnormality determination; It should be explained that if there is a trend abnormality (such as the characteristic value deviating in a single direction for multiple consecutive periods), the method of conducting hidden abnormality determination by calculating the feature discriminability and the standardized deviation is as follows: for the core features with high feature discriminability (such as THD, peak slope), set a lower standardized deviation threshold (such as Z≥1.5). When the standardized deviation of such features continuously exceeds the threshold and the trend slope exceeds the normal fluctuation range (such as an offset of ≥0.5% per week), combined with the weight given by the feature discriminability (the higher the discriminability, the greater the weight), if the weighted comprehensive score exceeds the preset threshold, it is determined as a hidden abnormality to achieve the identification of early or progressive faults; Among them, the method of iteratively optimizing the abnormal waveform dictionary by combining the voltage waveform features to be identified is as follows: Preferably, when a hidden abnormality or unknown abnormality is detected, first extract the frequency domain, time-frequency domain, and morphological feature vectors of the waveform to be identified, and compare them with the existing templates in the dictionary through the density clustering algorithm to determine whether a new abnormal clustering is formed; If it is an early or variant form of a known abnormality, update the feature mean, variation range, and temporal evolution pattern of the corresponding template (such as the trend feature of slow harmonic growth). If it is a new type of abnormality, add a new template entry (including feature vectors, waveform prototypes, and abnormal type labels); Recalculate the feature coverage and feature discriminability based on new samples, dynamically adjust the sorting of feature priority values, optimize and streamline the feature set, and eliminate abnormal samples from the normal baseline data to update the statistical model, which is conducive to achieving continuous adaptation of the dictionary to new abnormal patterns and device gradual fault characteristics in power grid operation.

[0031] Embodiment 4: As Figure 3 shown, an abnormal waveform matching system for identifying voltage abnormalities of an electric energy meter includes the following modules: Dictionary construction module: used to collect historical abnormal waveforms of the electric energy meter, decompose the features of the historical abnormal waveforms, and construct an abnormal waveform dictionary; Streamlining and splitting module: perform coverage analysis on the historical abnormal waveforms, obtain feature priority values, screen the core features of the abnormal waveform dictionary, and establish a streamlined feature set; Dominant initial judgment module: used to collect the voltage waveform to be identified of the electric energy meter in real time, extract the features of the voltage waveform to be identified to construct a waveform feature vector, perform matching calculation on the waveform feature vector and the streamlined feature set, and initially judge whether there is a dominant abnormality in the voltage waveform to be identified; Abnormal final judgment module: If there is a dominant abnormality, screen and process the abnormal types to obtain candidate template types, perform abnormal decision analysis on the candidate template waveform prototype and the voltage waveform to be identified, obtain an abnormal decision coefficient, and finally determine the abnormal type existing in the electric energy meter; Dictionary iteration module: If it is initially judged that there is no dominant abnormality in the voltage waveform to be identified, judge whether there is a recessive abnormality in the voltage waveform to be identified. If so, iteratively optimize the abnormal waveform dictionary in combination with the features of the voltage waveform to be identified.

[0032] The above has described an embodiment of the present invention in detail, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter, characterized in that, It includes the following steps: Collect the historical abnormal waveforms of the electric energy meter, perform feature decomposition on the historical abnormal waveforms, and construct an abnormal waveform dictionary; Perform coverage analysis on the historical abnormal waveforms, obtain feature priority values, and conduct screening and processing to obtain a refined feature set; Collect the voltage waveforms to be recognized of the electric energy meter in real time, extract features from the voltage waveforms to be recognized to construct waveform feature vectors, perform matching calculations on the waveform feature vectors and the refined feature set, and initially determine whether there are obvious abnormalities in the voltage waveforms to be recognized; If there are obvious abnormalities, conduct screening and processing on the abnormal types to obtain candidate template types, conduct abnormal decision analysis on the candidate template waveform prototypes and the voltage waveforms to be recognized, obtain abnormal decision coefficients, and finally determine the abnormal types existing in the electric energy meter; If it is initially determined that there are no obvious abnormalities in the voltage waveforms to be recognized, determine whether there are hidden abnormalities in the voltage waveforms to be recognized. If so, iteratively optimize the abnormal waveform dictionary in combination with the features of the voltage waveforms to be recognized.

2. The abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 1, characterized in that, The method for constructing the abnormal waveform dictionary is as follows: Collect the historical abnormal waveforms of the electric energy meter, perform feature decomposition on the historical abnormal waveforms, and obtain recognition feature vectors including frequency domain features, time-frequency domain features, and morphological features; Based on the dynamic dictionary learning algorithm, construct an abnormal waveform dictionary including recognition feature vectors, abnormal types, and waveform prototypes.

3. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 1, characterized in that The method for obtaining the refined feature set is as follows: Use the electric energy meter to be recognized as a sample for abnormal recognition, obtain historical abnormal waveforms from the abnormal waveform dictionary, and classify and correspond the historical abnormal waveforms according to the samples; Obtain the feature coverage and feature discriminability of each sample, and perform summation processing to obtain feature priority values; Sort the sub-features based on the feature priority values, extract the core features from the abnormal waveform dictionary, and construct a refined feature set based on the feature vectors corresponding to the core features.

4. The abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 3, characterized in that, The method for obtaining the feature coverage and feature discriminability is as follows: Calculate the occurrence frequencies of different sub-features of each sample to obtain the feature coverage; Obtain the between-class scatter and within-class scatter of the samples, and perform ratio calculation to obtain the feature discriminability.

5. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 1, characterized in that, The method for initially determining whether there are obvious abnormalities in the voltage waveforms to be recognized is as follows: Perform waveform preprocessing on the voltage waveforms to be recognized collected in real time, extract features from the waveforms after waveform preprocessing, and construct waveform feature vectors; Extract the recognition feature vectors of each abnormal type from the refined feature set, and perform matching degree calculations on the recognition feature vectors and the waveform feature vectors of each abnormal type to obtain feature matching degrees; Compare the feature matching degrees with a preset feature matching degree threshold for processing to obtain the result of determining that there are obvious abnormalities in the voltage waveforms to be recognized.

6. The abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 5, characterized in that The method for performing the matching degree calculation is as follows: Calculate the cosine similarity between the recognition feature vectors and the waveform feature vectors to obtain the feature matching degree.

7. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 1, characterized in that, The method for finally determining the abnormal types existing in the electric energy meter is as follows: If there are corresponding abnormal types, calculate the proportion of the deviation between the feature matching degree of each abnormal type and the feature matching degree threshold to obtain the feature deviation ratio; Screen the abnormal types based on the feature deviation ratio to obtain candidate template types; Obtain the timing matching degree and feature matching degree of the candidate template type, perform weighted summation processing on the timing matching degree and the feature matching degree to obtain an anomaly decision coefficient; Based on the anomaly decision coefficient, obtain the anomaly type of the candidate template type, and finally determine whether there is an obvious anomaly of the anomaly type in the voltage waveform to be recognized.

8. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 7, characterized in that, The method for obtaining the timing matching degree is as follows: Extract the waveform prototype of the candidate template type from the abnormal waveform dictionary, as well as the voltage waveform to be recognized collected by the electric energy meter, and perform dynamic time warping calculation on the waveform prototype and the voltage waveform to be recognized to obtain the timing matching degree.

9. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 7, characterized in that, The method for obtaining the weight coefficient of the anomaly decision coefficient is as follows: Perform transformation processing on the waveform prototype and the voltage waveform to be recognized to generate a time-frequency spectrogram, and extract the transient energy distribution entropy; Calculate the proportion of the absolute deviation of the transient energy distribution entropy of the waveform prototype and the voltage waveform to be recognized as the weight coefficient.

10. An abnormal waveform matching method for identifying abnormal voltage of an electric energy meter according to claim 1, characterized in that, The method for determining whether there is a hidden anomaly in the voltage waveform to be recognized is as follows: Construct a normal waveform feature baseline model; Perform multi-dimensional feature deviation calculation to mark potential abnormal features; Perform dynamic trend analysis on the potential abnormal features to determine whether there is a trend anomaly; If there is a trend anomaly, combine the feature discriminability and the standardized deviation to determine the hidden anomaly.

Images