A concentrator intelligent fusion terminal supporting remote power management

Abstract

This invention relates to the field of data management technology, specifically to a concentrator intelligent fusion terminal supporting remote power management. The concentrator intelligent fusion terminal includes a data filtering module, used to construct an electrical parameter data point matrix of the target cluster based on electrical parameter data points on the electrical parameter signals of service users in the target cluster. It adjusts a preset kernel bandwidth parameter based on the angle between the electrical parameter data points and neighboring signal data points, the differences in electrical parameter data, and the gradient differences between the electrical parameter data points and the matrix's neighboring data points to obtain the kernel bandwidth parameter of the electrical parameter data points. It then uses the kernel bandwidth parameter of the electrical parameter data points in the electrical parameter data point matrix to perform Gaussian filtering on the corresponding electrical parameter data points to obtain a filtering matrix. A compression management module uses the SVD compression algorithm to compress the filtering matrix to obtain compressed data. This invention not only improves the compression effect but also enhances the effectiveness of subsequent remote power management.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and more specifically to a concentrator intelligent fusion terminal that supports remote power management. Background Technology

[0002] To improve the effectiveness of remote power management, intelligent fusion terminals for concentrators with intelligent data processing capabilities have emerged. These terminals provide support for remote power management through functions such as power acquisition, communication management, remote control, and power quality monitoring.

[0003] In practical applications, the electrical parameter data collected by the intelligent converged terminal of the concentrator from service users is massive. Therefore, to alleviate the pressure on communication bandwidth and storage resources, and to enable more timely and effective power management, the SVD compression algorithm is generally used to compress the collected user electrical parameter data before transmission and storage. However, this method of directly compressing the collected user electrical parameter data results in poor compression performance. Poor compression performance negatively impacts the effectiveness of subsequent remote power management. For example, poor compression performance reduces the timeliness and effectiveness of power quality management, making it impossible to promptly capture abnormal power quality fluctuations such as voltage dips and harmonic exceedances. Therefore, improving the compression performance of the electrical parameter data collected by the intelligent converged terminal of the concentrator from service users is an urgent problem to be solved. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a concentrator intelligent fusion terminal that supports remote power management. The specific technical solution adopted is as follows:

[0005] One embodiment of the present invention provides a concentrator intelligent fusion terminal supporting remote power management, the concentrator intelligent fusion terminal supporting remote power management comprising:

[0006] The data acquisition module is used to acquire the electrical parameter signals of each service user of the concentrator intelligent fusion terminal, wherein the electrical parameter signals are composed of electrical parameter data points;

[0007] The clustering module is used to obtain the frequency domain signal corresponding to the electrical parameter signal, extract features from the frequency domain signal to obtain the feature vector of each service user, and perform hierarchical clustering of all service users according to the distance between the feature vectors of any two service users to obtain the target cluster.

[0008] The data filtering module is used to construct an electrical parameter data point matrix of the target cluster based on electrical parameter data points on the electrical parameter signals of the service users in the target cluster, obtain signal neighborhood data points and matrix neighborhood data points of the electrical parameter data points in the electrical parameter data point matrix, adjust a preset kernel bandwidth parameter based on the angle between the electrical parameter data points and the signal neighborhood data points, the difference in electrical parameter data, and the gradient difference between the electrical parameter data points and the matrix neighborhood data points to obtain the kernel bandwidth parameter of the electrical parameter data points, and perform Gaussian filtering on the corresponding electrical parameter data points using the kernel bandwidth parameter of the electrical parameter data points in the electrical parameter data point matrix to obtain a filtering matrix;

[0009] The compression management module uses the SVD compression algorithm to compress the filter matrix to obtain compressed data.

[0010] Beneficial effects: This invention includes a data acquisition module for acquiring electrical parameter signals of each service user in a concentrator intelligent fusion terminal; a clustering module for acquiring the frequency domain signal corresponding to the electrical parameter signal, extracting features from the frequency domain signal to obtain feature vectors of each service user, and performing hierarchical clustering of all service users based on the distance between the feature vectors of any two service users to obtain a target cluster; a data filtering module for constructing an electrical parameter data point matrix of the target cluster based on the electrical parameter data points on the electrical parameter signals of the service users in the target cluster, acquiring signal neighborhood data points and matrix neighborhood data points of the electrical parameter data points in the electrical parameter data point matrix, adjusting a preset kernel bandwidth parameter based on the angle between the electrical parameter data points and the signal neighborhood data points, the difference in electrical parameter data, and the gradient difference between the electrical parameter data points and the matrix neighborhood data points to obtain the kernel bandwidth parameter of the electrical parameter data points, and performing Gaussian filtering on the corresponding electrical parameter data points using the kernel bandwidth parameter of the electrical parameter data points in the electrical parameter data point matrix to obtain a filtering matrix; and a compression management module for compressing the filtering matrix using the SVD compression algorithm to obtain compressed data. Furthermore, this invention filters the electrical parameter data point matrix by dynamically adjusting the kernel bandwidth parameter and performs SVD compression on the filtered matrix. This not only preserves important information but also improves the compression ratio and efficiency. In other words, filtering the electrical parameter data point matrix by dynamically adjusting the kernel bandwidth parameter can improve the compression effect, thereby improving the effect of subsequent remote power management. Attached Figure Description

[0011] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a structural block diagram of a concentrator intelligent fusion terminal that supports remote power management according to the present invention. Detailed Implementation

[0013] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0015] This embodiment provides a concentrator intelligent fusion terminal that supports remote power management, detailed as follows:

[0016] like Figure 1 As shown in the figure, this embodiment provides a concentrator intelligent fusion terminal that supports remote power management, comprising:

[0017] Data acquisition module 01 is used to acquire electrical parameter signals of each service user of the concentrator intelligent fusion terminal.

[0018] This embodiment primarily involves pre-processing the electrical parameter data collected from service users by the concentrator's intelligent fusion terminal, followed by compression using a compression algorithm to improve subsequent compression effectiveness. The steps of electrical parameter data acquisition, pre-compression processing, and data compression in this embodiment are as follows: The concentrator's intelligent fusion terminal typically establishes a communication connection with each user's smart meter via a low-voltage power line or RS485 bus, periodically or in real-time collecting electrical parameter data from each meter. The concentrator's intelligent fusion terminal then performs pre-processing operations such as normalization and data cleaning on the collected data. Following this, it performs pre-compression processing on the pre-processed data, using the SVD compression algorithm to compress the data. The compressed data is then transmitted to a remote master station system or cloud platform via the concentrator's built-in communication module. This remote platform enables further storage, analysis, scheduling, and remote control operations to achieve centralized monitoring and management of regional power or remote power management of the power system. In addition, the compression algorithm in this embodiment mainly refers to the SVD compression algorithm, and the concentrator intelligent fusion terminal that appears later in this embodiment is the same concentrator intelligent fusion terminal.

[0019] This embodiment first acquires all users served by the concentrator intelligent fusion terminal and records them as service users of the concentrator intelligent fusion terminal. Then, it acquires all electrical parameter data collected by the concentrator intelligent fusion terminal for each service user during the target monitoring period. The data types of electrical parameter data collected by the concentrator intelligent fusion terminal for any service user include, but are not limited to, voltage, current, active power, reactive power, and energy from the meter of that service user collected by the concentrator intelligent fusion terminal. Next, according to the order of acquisition of the electrical parameter data, initial sequences corresponding to each service user during the target monitoring period are obtained. These initial sequences are then preprocessed, and the preprocessed sequences are recorded as the electrical parameter data sequences corresponding to the service users during the target monitoring period. Preprocessing includes, but is not limited to, data normalization or standardization, data denoising, and data cleaning. Furthermore, the specific acquisition process for each initial sequence corresponding to each service user during the target monitoring period is as follows: For any service user, the electrical parameter data of the same type collected by the concentrator intelligent fusion terminal during the target monitoring period are sorted according to the order of collection time, and the sorted sequence is recorded as each initial sequence corresponding to the service user during the target monitoring period.

[0020] Furthermore, the monitoring time period is the time period consisting of any adjacent data transmission moments of the concentrator intelligent fusion terminal, while the target monitoring time period in this embodiment is the time period formed from the previous data transmission moment to the current data transmission moment of the concentrator intelligent fusion terminal. In specific applications, implementers need to set the time interval for the concentrator intelligent fusion terminal to receive or collect electrical parameter data and the time interval for the concentrator intelligent fusion terminal to transmit electrical parameter data according to the actual situation. This embodiment requires that electrical parameter data of different service users but belonging to the same type be collected synchronously, and the time interval for the concentrator intelligent fusion terminal to receive or collect electrical parameter data should be much smaller than the time interval for the concentrator intelligent fusion terminal to transmit electrical parameter data. For example, in this embodiment, the time interval for the concentrator intelligent fusion terminal to receive or collect electrical parameter data and the time interval for the concentrator intelligent fusion terminal to transmit electrical parameter data are set to 1 minute, and the time interval for the concentrator intelligent fusion terminal to transmit electrical parameter data is set to 1 hour. In addition, this embodiment can also require that different types of electrical parameter data of the same service users be collected synchronously.

[0021] After obtaining all electrical parameter data sequences corresponding to each service user of the concentrator intelligent fusion terminal during the target monitoring time period, the electrical parameter signals of each electrical parameter data sequence corresponding to each service user of the concentrator intelligent fusion terminal during the target monitoring time period are obtained and recorded as the electrical parameter signals of each service user of the concentrator intelligent fusion terminal during the target monitoring time period. The process of obtaining the electrical parameter signal of any electrical parameter data sequence corresponding to any service user of the concentrator intelligent fusion terminal during the target monitoring time period is as follows: first, the type of electrical parameter data in the electrical parameter data sequence is obtained and recorded as type W; then, a two-dimensional space corresponding to type W is constructed, and... All electrical parameter data and their acquisition times in the electrical parameter data sequence are mapped to a two-dimensional space corresponding to type W, resulting in electrical parameter data points corresponding to each electrical parameter data in the sequence. Then, the electrical parameter data points corresponding to the electrical parameter data in the sequence are connected sequentially according to the acquisition time, and the curve signal of the connection is recorded as the electrical parameter signal of the sequence. The horizontal coordinate of any electrical parameter data point in the sequence is the acquisition time of the electrical parameter data, and the vertical coordinate is the electrical parameter data itself. The types of electrical parameter data in any electrical parameter data sequence are consistent.

[0022] Therefore, this embodiment can obtain the electrical parameter signals of each service user of the concentrator intelligent fusion terminal during the target monitoring time period through the above process. Since this embodiment needs to process the electrical parameter data points on the electrical parameter signals of all service users of the concentrator intelligent fusion terminal before compression, and then compress the processed data, and since this embodiment not only has the same pre-compression processing process for electrical parameter data points on each type of electrical parameter signal, but also the same subsequent compression process for each type of electrical parameter signal, for ease of understanding and analysis, this embodiment will describe the pre-compression processing process on any type of electrical parameter signal as an example. That is, the type of electrical parameter data points on all electrical parameter signals that appear in this embodiment is the same. For example, this embodiment can describe the pre-compression processing process and the subsequent compression process for electrical parameter data points on voltage electrical parameter signals.

[0023] Clustering module 02 is used to obtain the frequency domain signal corresponding to the electrical parameter signal, extract features from the frequency domain signal to obtain the feature vector of each service user, and perform hierarchical clustering of all service users according to the distance between the feature vectors of any two service users to obtain the target cluster.

[0024] In practical applications of intelligent fusion terminals for concentrators, the collected power data such as voltage, current, and load often exhibit significant regional and time-varying characteristics due to differences in regional type, electricity consumption habits, and load fluctuations. For example, industrial areas are prone to high-frequency disturbances, while residential areas show periodic power consumption fluctuations. Therefore, this embodiment requires classifying service users based on their characteristics, and then processing and compressing each cluster obtained from the classification separately. The specific process for classifying service users is as follows:

[0025] First, a Fast Fourier Transform is performed on the electrical parameter signals of each service user obtained above to obtain the corresponding frequency domain signals. Then, feature extraction is performed on the frequency domain signals corresponding to the electrical parameter signals of each service user. The vector constructed from all the feature values ​​extracted from the frequency domain signals corresponding to the electrical parameter signals of each service user is denoted as the feature vector of the corresponding service user. When extracting feature values ​​for each frequency domain signal, the extracted feature values ​​mainly include the mean, variance, dominant frequency, total harmonic distortion rate, and spectral entropy of the frequency domain signal. Therefore, the feature vector corresponding to each service user contains the electrical parameters of the corresponding service user. The signal is composed of the mean, variance, dominant frequency, total harmonic distortion, and spectral entropy of the corresponding frequency domain signal. Furthermore, the eigenvalues ​​at the same position in different eigenvectors have the same type. For example, if the c-th eigenvalue in a certain eigenvector is the dominant frequency of the frequency domain signal, then the c-th eigenvalue in all eigenvectors is the dominant frequency of the frequency domain signal. After obtaining the feature vectors of the service users, hierarchical clustering is performed on all service users based on the distance between the feature vectors of any two service users to obtain the target cluster. The specific process of hierarchical clustering on all service users based on the distance between the feature vectors of any two service users to obtain the target cluster is as follows:

[0026] First, calculate the distance between the feature vectors of any two service users, and use this distance as the metric distance between the corresponding service users. That is, the metric distance between service user w1 and service user w2 is the distance between the feature vectors of service user w1 and service user w2. Furthermore, the distance between the feature vectors of service user w1 and service user w2 can be either the Euclidean distance between their respective feature vectors or... , The cosine similarity between the feature vectors of service user w1 and service user w2 is given. A higher similarity indicates that the two service users are closer. Therefore, when the similarity is higher, the distance measurement should be lower.

[0027] Then, hierarchical clustering is performed on all service users based on the metric distance between any two service users to obtain a clustering tree. Since the process of obtaining the clustering tree by hierarchical clustering of all service users is well-known given the metric distance between service users, it will not be described in detail in this embodiment. After obtaining the clustering tree, this embodiment will analyze the clustering results of each layer of the clustering tree to determine the layer suitable for subsequent analysis, i.e., to determine the target layer. The reason for not using any layer of the clustering tree as the target layer at this time is that if the number of clusters in the selected layer is too small, the density of each cluster in that layer will be low. When the density is low, more singular values ​​need to be retained to maintain the reconstruction accuracy or compression accuracy when performing SVD compression on the clusters in that layer. More singular values ​​will lead to a lower compression ratio. The selected layer has too many clusters. Although the feature vectors of service users within each cluster in the layer are relatively similar, the large number of clusters and the subsequent compression of clusters in the target layer will lead to low compression efficiency. Therefore, this embodiment needs to obtain the clustering result representation value corresponding to each layer of the clustering tree based on the distance between clusters in each layer, the metric distance between service users in each cluster in each layer, and the number of clusters in each layer. The larger the clustering result representation value corresponding to each layer, the better the compression effect based on the clusters in the corresponding layer. Therefore, after obtaining the clustering result representation value corresponding to each layer of the clustering tree, this embodiment selects the layer with the largest clustering result representation value as the target layer and uses all clusters in the target layer as target clusters.

[0028] Furthermore, in this embodiment, the specific process of obtaining the clustering result representation value corresponding to each layer of the clustering tree based on the distance between clusters at each layer, the metric distance between service users in each cluster at each layer, and the number of clusters at each layer is as follows: For the i-th layer of the clustering tree:

[0029] First, in each cluster at level i, the service user closest to the cluster center of the corresponding cluster is obtained and denoted as the representative user of the corresponding cluster. Then, based on the metric distance between the representative users of the clusters at level i, the cluster distance representation value of level i is obtained. Based on the metric distance between the service users in each cluster at level i, the user distribution density representation value of each cluster at level i is obtained. The mean of the user distribution density representation values ​​of all clusters at level i is calculated and used as the cluster density representation value of level i. Next, the reciprocal of the number of clusters at level i is obtained. Finally, the product of the reciprocal of the number of clusters at level i, the cluster distance representation value of level i, and the cluster density representation value of level i is calculated and used as the clustering result representation value of level i.

[0030] In this embodiment, the process of obtaining the cluster distance representation value corresponding to the i-th layer based on the metric distance between representative users of the clusters in the i-th layer is as follows: For the j-th cluster in the i-th layer, firstly, in the i-th layer, the set of the remaining clusters excluding the j-th cluster is denoted as the set to be analyzed corresponding to the j-th cluster. Then, the distance between the j-th cluster and each cluster in the set to be analyzed corresponding to the j-th cluster is calculated and denoted as the cluster distance. The set of cluster distances between the j-th cluster and each cluster in the set to be analyzed corresponding to the j-th cluster is denoted as the cluster distance set corresponding to the j-th cluster. The distance between the a-th cluster in the cluster distance set corresponding to the cluster is the metric distance between the representative user of the j-th cluster and the representative user of the a-th cluster in the set to be analyzed corresponding to the j-th cluster. Then, the mean of the cluster distance set corresponding to the j-th cluster is calculated and denoted as the mean cluster distance of the j-th cluster. Finally, the mean of the mean cluster distances corresponding to all clusters in the i-th layer is calculated and denoted as the cluster distance representation value corresponding to the i-th layer. The larger the mean cluster distance of each cluster in the i-th layer, that is, the larger the cluster distance representation value corresponding to the i-th layer, the greater the difference or distance between the clusters in the i-th layer.

[0031] In this embodiment, the specific process of obtaining the user distribution density representation value corresponding to each cluster in the i-th layer based on the metric distance between any two service users in each cluster in the i-th layer is as follows: For the j-th cluster in the i-th layer, all service users in the j-th cluster are arranged in pairs without repetition to obtain all user combinations corresponding to the j-th cluster. The reciprocal of the metric distance between two service users in each user combination is recorded as the user density index value of the corresponding user combination. The mean of the user density index values ​​of all user combinations corresponding to the j-th cluster is calculated and used as the j-th cluster density index value. The user distribution density representation value corresponding to the cluster is as follows: the larger the user distribution density representation value corresponding to the j-th cluster, the closer the distribution of service users in the j-th cluster or the closer the distance between service users in the j-th cluster. In addition, if there is a case where the metric distance between two service users is 0, the reciprocal of the result of adding the metric distance between two service users in each user combination to the preset first constant can be used as the user density index value of the corresponding user combination. The preset first constant is to prevent the denominator from being 0, and can be set according to the actual situation, such as setting it to 0.01.

[0032] And the specific calculation expression for the clustering result representation value corresponding to the i-th layer is:

[0033]

[0034] in, This represents the clustering result characteristic value corresponding to the i-th layer in the clustering tree. Let be the number of clusters in the i-th layer. Let a be the mean cluster distance of the a-th cluster at the i-th layer. This represents the user distribution density characteristic value corresponding to the a-th cluster on the i-th layer. This represents the cluster distance characteristic of the i-th layer. Let be the cluster density representation value corresponding to the i-th layer. A larger cluster distance representation value and a larger cluster density representation value corresponding to the i-th layer, coupled with a smaller number of clusters in the i-th layer, indicates a better compression effect based on the clusters in the i-th layer. In other words, a larger clustering result representation value for the i-th layer indicates a better compression effect based on the clusters in the i-th layer, thus indicating a higher probability that the i-th layer is the target layer. Conversely, a smaller value indicates a lower probability that the i-th layer is the target layer. The smaller the ratio; in addition, when the cluster distance representation value corresponding to the i-th layer is larger, it indicates that the difference or distance between the clusters on the i-th layer is larger. And when the difference or distance between the clusters on the i-th layer is larger, it indicates that the compression effect based on the clusters on the i-th layer is better. When the cluster density representation value corresponding to the i-th layer is larger, it indicates that the distribution of service users in each cluster on the i-th layer is more compact. And when the density representation value of the clusters corresponding to the i-th layer is larger, it indicates that the distribution of service users in each cluster on the i-th layer is more compact. And when the density representation value of the clusters on the i-th layer is larger, it indicates that the compression effect based on the clusters on the i-th layer is better.

[0035] Therefore, this embodiment can obtain the target cluster through the above process.

[0036] The data filtering module 03 is used to construct an electrical parameter data point matrix of the target cluster based on the electrical parameter data points on the electrical parameter signals of the service users in the target cluster, obtain the signal neighborhood data points and matrix neighborhood data points of the electrical parameter data points in the electrical parameter data point matrix, adjust the preset kernel bandwidth parameter according to the angle between the electrical parameter data points and the signal neighborhood data points, the difference in electrical parameter data, and the gradient difference between the electrical parameter data points and the matrix neighborhood data points to obtain the kernel bandwidth parameter of the electrical parameter data points, and perform Gaussian filtering on the corresponding electrical parameter data points using the kernel bandwidth parameter of the electrical parameter data points in the electrical parameter data point matrix to obtain the filtering matrix.

[0037] After obtaining the target clusters, an electrical parameter data point matrix for each target cluster is constructed. The process for obtaining the electrical parameter data point matrix for each target cluster is as follows: Based on the number of service users in each target cluster and the electrical parameter data points on the electrical parameter signals of each service user in each target cluster, an electrical parameter data point matrix for each target cluster is constructed. Furthermore, all electrical parameter data points in the k-th row of the electrical parameter data point matrix of any target cluster are electrical parameter data points on the electrical parameter signals of the k-th service user in that target cluster. The electrical parameter data point matrix of any target cluster... The number of rows is consistent with the number of service users in the target cluster. The number of electrical parameter data points on the electrical parameter data point matrix of any target cluster is consistent with the number of electrical parameter data points on the electrical parameter signals of all service users in the target cluster. Moreover, the electrical parameter data point in the k-th row and s-th column of the electrical parameter data point matrix corresponding to any target cluster is the s-th electrical parameter data point on the electrical parameter signal of the k-th service user in the target cluster. That is, in the electrical parameter data point matrix of any target cluster, the acquisition time or horizontal coordinate value of all electrical parameter data points located in the same column is consistent.

[0038] Furthermore, since the core of SVD compression of a matrix is ​​to reconstruct an approximate version of the original data matrix with minimal information, this embodiment, in order to further improve the compression effect, does not directly compress the electrical parameter data point matrix after obtaining the electrical parameter data point matrix of the target cluster. Instead, it adaptively adjusts the kernel bandwidth parameter when performing Gaussian filtering on the corresponding electrical parameter data points by analyzing the importance of each electrical parameter data point in the matrix and the degree of perturbation of the matrix structure by each electrical parameter data point in the matrix. This allows the filtered matrix to retain key or important information while making the data distribution in the matrix more regular, thus making the filtered matrix easier to approximate with a low rank. In other words, it allows for reconstruction with fewer singular values ​​in subsequent SVD compression of the filtered matrix, thereby improving the compression quality and compression ratio. That is, the more similar the data in the matrix, the more regular the distribution, and the lower the rank of the matrix, the fewer singular values ​​are retained in the SVD compression of the matrix.

[0039] Therefore, based on the above analysis, this embodiment, after obtaining the electrical parameter data point matrix of each target cluster, adaptively adjusts the kernel bandwidth parameter when performing Gaussian filtering on the corresponding electrical parameter data points according to the importance of each electrical parameter data point in the matrix and the degree of local perturbation of the matrix structure by each electrical parameter data point in the matrix. Since the angle between each electrical parameter data point in the matrix and its corresponding signal neighbor data points, as well as the difference in electrical parameter data, can reflect the importance of the corresponding electrical parameter data points, the matrix neighbor of each electrical parameter data point in the matrix and its corresponding signal neighbor data points... The gradient difference between domain data points can reflect the degree of local perturbation of the matrix structure by the corresponding electrical parameter data points. Therefore, in this embodiment, it is necessary to first obtain the signal neighborhood data points and matrix neighborhood data points of each electrical parameter data point in the electrical parameter data point matrix. Then, based on the angle between each electrical parameter data point in the electrical parameter data point matrix and the signal neighborhood data points of the corresponding electrical parameter data point, the difference in electrical parameter data, and the gradient difference between each electrical parameter data point in the electrical parameter data point matrix and the matrix neighborhood data points of the corresponding electrical parameter data point, the preset kernel bandwidth parameters are adjusted to obtain the kernel bandwidth parameters of each electrical parameter data point in the electrical parameter data point matrix.

[0040] In this embodiment, the specific process for obtaining the signal neighborhood data points and matrix neighborhood data points of each electrical parameter data point in the electrical parameter data point matrix is ​​as follows: For any electrical parameter data point h on the electrical parameter data point matrix A corresponding to any target cluster:

[0041] First, obtain the electrical parameter signal to which the electrical parameter data point h belongs. Then, on the electrical parameter signal to which the electrical parameter data point h belongs, obtain the electrical parameter data points located to the left of the electrical parameter data point h, and record them as the left signal neighbor data points of the electrical parameter data point h. On the electrical parameter signal to which the electrical parameter data point h belongs, obtain the electrical parameter data points located to the right of the electrical parameter data point h, and record them as the right signal neighbor data points of the electrical parameter data point h. Both the left and right signal neighbor data points of the electrical parameter data point h belong to the signal neighbor data points of the electrical parameter data point h. Additionally, it should be noted that if there are no electrical parameter data points to the left or right of the electrical parameter data point h on the electrical parameter signal to which the electrical parameter data point h belongs... If the data point is an electrical parameter data point, then an interpolation algorithm is used to interpolate to the left or right of the electrical parameter data point h. The interpolation result is used as the signal neighborhood data point of the electrical parameter data point h. If the electrical parameter data point h is the first electrical parameter data point on the electrical parameter signal to which the electrical parameter data point belongs, then interpolation is performed to the left of the electrical parameter data point h, and the interpolated data point is used as the signal neighborhood data point to the left of the electrical parameter data point h. If the electrical parameter data point h is the last electrical parameter data point on the electrical parameter signal to which the electrical parameter data point h belongs, then interpolation is performed to the right of the electrical parameter data point h, and the interpolated data point is used as the signal neighborhood data point to the right of the electrical parameter data point h. The interpolation process is well known. On the electrical parameter data point matrix A, a rectangular window of a preset length is constructed with electrical parameter data point h as the center, and is denoted as the neighborhood window of electrical parameter data point h. Then, all electrical parameter data points located within the neighborhood window of electrical parameter data point h are acquired and denoted as matrix neighborhood data points of electrical parameter data point h. In specific applications, the implementer needs to set the preset length according to the actual situation such as the size of the electrical parameter data point matrix. For example, in this embodiment, the preset length is set to 3, so the size of the constructed neighborhood window of electrical parameter data points is 3×3.

[0042] In this embodiment, the process of adjusting the preset kernel bandwidth parameter based on the angle between each electrical parameter data point in the electrical parameter data point matrix and its corresponding signal neighbor data point, the difference in electrical parameter data, and the gradient difference between each electrical parameter data point in the electrical parameter data point matrix and its corresponding matrix neighbor data point, to obtain the kernel bandwidth parameter of each electrical parameter data point in the electrical parameter data point matrix, is as follows:

[0043] For any electrical parameter data point h on the electrical parameter data point matrix A corresponding to any target cluster:

[0044] First, based on the signal neighbor data points of electrical parameter data point h, the neighborhood angle of electrical parameter data point h is obtained. The specific process for obtaining the neighborhood angle of electrical parameter data point h is as follows: the line connecting electrical parameter data point h to the signal neighbor data point to the left of electrical parameter data point h is recorded as the first line segment of electrical parameter data point h, the line connecting electrical parameter data point h to the signal neighbor data point to the right of electrical parameter data point h is recorded as the second line segment of electrical parameter data point h, the second line segment of electrical parameter data point h is rotated counterclockwise, and the angle at which the second line segment of electrical parameter data point h is rotated counterclockwise to the first line segment of electrical parameter data point h is recorded as the neighborhood angle of electrical parameter data point h.

[0045] After obtaining the neighborhood angles of the electrical parameter data point h, the importance value of the electrical parameter data point h is obtained based on the differences in neighborhood angles between the electrical parameter data point h and other electrical parameter data points in the same row, as well as the difference between the mean of the y-coordinates of the electrical parameter data point h and the mean of the y-coordinates of the electrical parameter data points in the same row. The x-axis of the electrical parameter data point represents time, and the y-axis represents the electrical parameter data. Then, based on the gradient differences between the electrical parameter data point h and its matrix neighborhood data points, the matrix local perturbation value of the electrical parameter data point h is obtained. A larger importance value indicates that the electrical parameter data point h is more important and critical. Alternatively, the higher the value of the electrical parameter data point h, the more important it is for the subsequent remote power management. Conversely, the smaller the value of the importance of the electrical parameter data point h, the less important it is. The larger the value of the local perturbation of the matrix of the electrical parameter data point h, the greater the degree of local perturbation of the electrical parameter data point matrix A. A large degree of local perturbation of the electrical parameter data point matrix A means that a small change in the electrical parameter data point h will cause a significant change in the overall properties of the matrix. Then, the preset kernel bandwidth parameter is adjusted according to the importance value of the electrical parameter data point h and the local perturbation value of the matrix to obtain the kernel bandwidth parameter of the electrical parameter data point h.

[0046] In this embodiment, the specific process of obtaining the importance characterization value of electrical parameter data point h based on the difference in neighborhood angles between electrical parameter data point h and other electrical parameter data points in the same row as h, and the difference between the mean of the ordinate of electrical parameter data point h and the mean of the ordinates of electrical parameter data points in the same row as h, is as follows:

[0047] First, in the electrical parameter data point matrix A, obtain the row containing electrical parameter data point h, that is, the row to which electrical parameter data point h belongs. Then, in the row to which electrical parameter data point h belongs, obtain all electrical parameter data points except for electrical parameter data point h, and denote the set of all remaining electrical parameter data points in the row to which electrical parameter data point h belongs as the row set of electrical parameter data point h. Next, obtain the feature angle ratio between electrical parameter data point h and each electrical parameter data point in the row set, and the feature angle ratio between electrical parameter data point h and the g-th electrical parameter data point in the row set is... Rh is the neighborhood angle of the electrical parameter data point h. Let the angle be the neighborhood angle of the g-th electrical parameter data point in the same set. Let be the acquisition time interval between electrical parameter data point h and the g-th electrical parameter data point, which is also the absolute value of the difference between the x-coordinate value of electrical parameter data point h and the x-coordinate value of the g-th electrical parameter data point. Next, calculate the mean of the characteristic angle ratios between electrical parameter data point h and all electrical parameter data points in the same row set, and normalize the calculated mean of the characteristic angle ratios using the normalization function Norm(). Use the result of the normalization process as the normalized angle difference representation value. Then, obtain the mean of the y-coordinates of all electrical parameter data points in the row to which electrical parameter data point h belongs, and use this as the row average of electrical parameter data point h. First, the absolute value of the difference between the ordinate value and the row mean of the electrical parameter data point h is calculated. Then, the absolute value of the difference between the ordinate value and the row mean of the electrical parameter data point h is normalized using the normalization function Norm(). The result of the normalization is used as the deviation characterization value of the electrical parameter data point h. Finally, the mean of the normalized angle difference characterization value and the deviation characterization value of the electrical parameter data point h is calculated and recorded as the importance characterization value of the electrical parameter data point h. The specific formula for calculating the importance characterization value of the electrical parameter data point h is as follows:

[0048]

[0049] Where W is the importance value of electrical parameter data point h, Norm() is the normalization function, and G is the number of electrical parameter data points in the same row set of electrical parameter data point h. The ordinate value of the electrical parameter data point h. Let h be the row mean of the electrical parameter data points; and It can reflect the difference between electrical parameter data point h and the electrical parameter data points of the same service user to which h belongs, and when A larger value indicates that, within the electrical parameter signal of the service user to which electrical parameter data point h belongs, the difference in electrical parameter data between the electrical parameter data point and h is greater in time. Generally, data collected at similar times may have high similarity and correlation. Therefore, when... The larger the value, the more special the electrical parameter data point h is, indicating that the electrical parameter data point h is more likely to carry important information related to subsequent remote power management. A larger value indicates that the electrical parameter data point h is more important; while when A larger value indicates a greater difference in electrical parameter data between the electrical parameter data point h and all electrical parameter data points on the electrical parameter signal of the service user to which h belongs. This suggests a greater deviation of the electrical parameter data point h from the overall data size on the electrical parameter signal to which h belongs, indicating a greater importance of the electrical parameter data point h or a higher likelihood that the electrical parameter data point h carries important information related to subsequent remote power management. Therefore, when... The larger and When W is larger, it indicates that the electrical parameter data point h is more important or more likely to carry important information related to subsequent remote power management. Therefore, the retention of points with greater importance should be greater. That is, the kernel bandwidth parameter used for filtering should be smaller when filtering points with greater importance are used. A smaller kernel bandwidth parameter indicates a smaller smoothness, and a larger kernel bandwidth parameter indicates a greater smoothness.

[0050] In this embodiment, the specific process of obtaining the matrix local perturbation characterization value of electrical parameter data point h based on the gradient difference between electrical parameter data point h and its matrix neighborhood data points is as follows:

[0051] First, the set of all matrix neighborhood data points of electrical parameter data point h is denoted as the matrix neighborhood set of electrical parameter data point h. Then, the SOBEL operator is used to calculate the gradient magnitude and gradient direction between electrical parameter data point h and the electrical parameter data points in the matrix neighborhood set. The process of calculating gradient magnitude and gradient direction based on the SOBEL operator is well-known. Next, based on the calculated gradient magnitude and gradient direction of electrical parameter data point h and the gradient magnitude and gradient direction of the electrical parameter data points in the matrix neighborhood set, the gradient difference value between electrical parameter data point h and each matrix neighborhood data point in the matrix neighborhood set is calculated. The gradient difference value between electrical parameter data point h and the f-th matrix neighborhood data point in the matrix neighborhood set of electrical parameter data point h is... Norm() is the normalization function. Let h be the gradient magnitude of the electrical parameter data point. Let f be the gradient magnitude of the f-th matrix neighborhood data point in the matrix neighborhood set of electrical parameter data point h. The gradient direction of the electrical parameter data point h. Let f be the gradient direction of the f-th matrix neighborhood data point in the matrix neighborhood set of the electrical parameter data point h; then, calculate the mean of the gradient differences between the electrical parameter data point h and all matrix neighborhood data points in the matrix neighborhood set of the electrical parameter data point h, and use this as the matrix local perturbation characterization value of the electrical parameter data point h, that is, the matrix local perturbation characterization value of the electrical parameter data point h is... H represents the number of neighborhood data points in the matrix neighborhood set of the electrical parameter data point h. Furthermore, the greater the gradient difference between the electrical parameter data point h and the neighborhood data points in its matrix neighborhood set—that is, the greater the local perturbation value of the electrical parameter data point h—the greater the degree of local perturbation of the matrix containing the electrical parameter data point h. Conversely, the greater the degree of local perturbation of the matrix containing the electrical parameter data point h, the greater the local perturbation of the matrix containing the electrical parameter data point h. The rank of the data point h may be relatively high. That is, the greater the local perturbation of the matrix containing the data point h, the more likely the data point h is to be the reason why the matrix containing the data point h is not easily approximated by a low-rank matrix but is easily approximated by a high-rank matrix during compression. Therefore, in order to make the matrix containing the data point h more easily approximated by a low-rank matrix during compression, this embodiment should require that the data point with low importance and a larger kernel bandwidth parameter for the data point containing the data point h with a greater degree of local perturbation of the matrix is ​​further required.

[0052] In this embodiment, the specific process of adjusting the preset kernel bandwidth parameter based on the importance characterization value and matrix local perturbation characterization value of the electrical parameter data point h to obtain the kernel bandwidth parameter of the electrical parameter data point h is as follows: First, calculate the reciprocal of the result obtained by adding the importance characterization value of the electrical parameter data point h to a preset first constant, and record it as the first reciprocal value. Then, normalize the first reciprocal value and use it as the first adjustment factor of the electrical parameter data point h. Next, normalize the matrix local perturbation characterization value of the electrical parameter data point h and use it as the second adjustment factor of the electrical parameter data point h. Finally, adjust the first adjustment factor of the electrical parameter data point h and the kernel bandwidth parameter of the electrical parameter data point h. The second adjustment factor of the electrical parameter data point h is weighted and summed, and the weighted sum is used as the initial adjustment coefficient of the electrical parameter data point h. Then, the initial adjustment coefficient of the electrical parameter data point h is normalized and used as the kernel bandwidth parameter adjustment coefficient of the electrical parameter data point h. Finally, the product of the kernel bandwidth parameter adjustment coefficient of the electrical parameter data point h and the preset kernel bandwidth parameter is calculated and used as the kernel bandwidth parameter of the electrical parameter data point h. A smaller kernel bandwidth parameter adjustment coefficient indicates a greater degree of adjustment to the preset kernel bandwidth parameter, and a smaller adjusted kernel bandwidth parameter. Conversely, a larger kernel bandwidth parameter adjustment coefficient indicates a smaller degree of adjustment to the preset kernel bandwidth parameter, and a larger adjusted kernel bandwidth parameter. In this embodiment, the preset kernel bandwidth parameter is taken as an empirical value of 2, and the preset kernel bandwidth parameter can also be set according to actual conditions.

[0053] In this embodiment, the initial adjustment coefficient for electrical parameter data point h is calculated using the following expression: ,in, Here, W represents the importance of the electrical parameter data point h, U represents the local matrix perturbation of the electrical parameter data point h, and c1 is a preset first constant. Here, c1 is used to prevent the denominator from being zero. A smaller W and a larger U indicate a larger initial adjustment coefficient for the electrical parameter data point h, and vice versa. The decision weights mentioned above need to be set by the implementer according to the actual situation. For example, if a greater correlation is required between the adjustment of the kernel bandwidth parameter and the importance of the data point, then the decision weights can be set larger. Conversely, if a greater correlation is required between the adjustment of the kernel bandwidth parameter and the local matrix perturbation of the data point, then the decision weights can be set smaller. In this embodiment, the decision weights can be set as follows: The value is set to 0.65. Furthermore, the normalization of the initial adjustment coefficient for electrical parameter data point h mentioned above refers to the ratio of the initial adjustment coefficient of electrical parameter data point h to the sum of the initial adjustment coefficients of all electrical parameter data points in the electrical parameter data point matrix to which h belongs. Alternatively, the normalization function Norm() can be used directly to normalize the initial adjustment coefficient.

[0054] Therefore, this embodiment can obtain the kernel bandwidth parameter of each electrical parameter data point in the electrical parameter data point matrix of each target cluster through the above method. After obtaining the kernel bandwidth parameter of each electrical parameter data point in the electrical parameter data point matrix, a Gaussian filter with the kernel bandwidth parameter of the corresponding electrical parameter data point is used to filter the corresponding electrical parameter data point, and the filtered result is recorded as the filtered data point of the corresponding electrical parameter data point. The new matrix composed of the filtered data points of each electrical parameter data point in the electrical parameter data point matrix is ​​recorded as the filtering matrix of the corresponding electrical parameter data point matrix. That is, when performing Gaussian filtering on any electrical parameter data point in any electrical parameter data point matrix, the kernel bandwidth parameter used is the above... The kernel bandwidth parameter of the calculated electrical parameter data point is described above. The data point in the nth row and mth column of the filter matrix of any electrical parameter data point matrix is ​​the filtered data point of the electrical parameter data point in the nth row and mth column of the electrical parameter data point matrix. At this time, the filtered matrix of the electrical parameter data point matrix obtained at this time retains key information and the data distribution in the filtered matrix is ​​more regular than that in the corresponding electrical parameter data point matrix. Compared with the direct SVD compression of the electrical parameter data point matrix, the compression effect of the filtered matrix is ​​better. In addition, given that the kernel bandwidth parameters of each parameter in the matrix are known, the process of Gaussian filtering the matrix is ​​well known and therefore will not be described in detail.

[0055] Therefore, this embodiment can obtain the filtering matrix of the electrical parameter data point matrix of each target cluster through the above process.

[0056] Compression management module 04 uses the SVD compression algorithm to compress the filter matrix to obtain compressed data.

[0057] After obtaining the filtering matrix of the electrical parameter data point matrix of each target cluster, SVD compression is performed on the filtering matrix of the electrical parameter data point matrix of each target cluster to obtain compressed data. That is, in this embodiment, the SVD compression algorithm is used to compress each filtering matrix separately. Then, the concentrator intelligent fusion terminal transmits the compressed data to the remote power management platform for subsequent remote power management. Moreover, this embodiment can not only improve the compression effect, but also save storage space and reduce storage costs. In addition, based on the data with better compression effect, the effect of subsequent remote power management can also be improved.

[0058] Thus, this embodiment has completed the compression of electrical parameter data points on the electrical parameter signals of each service user of the concentrator intelligent fusion terminal, or completed the compression of the electrical parameter data of the service user; and the process of SVD compression and decompression of the matrix is ​​well known, so it will not be described in detail here.

[0059] In summary, this embodiment includes a data acquisition module for acquiring the electrical parameter signals of each service user in the concentrator's intelligent fusion terminal; a clustering module for acquiring the frequency domain signals corresponding to the electrical parameter signals, extracting features from the frequency domain signals to obtain feature vectors for each service user, and performing hierarchical clustering of all service users based on the distance between the feature vectors of any two service users to obtain a target cluster; a data filtering module for constructing an electrical parameter data point matrix of the target cluster based on the electrical parameter data points on the electrical parameter signals of the service users in the target cluster, acquiring signal neighborhood data points and matrix neighborhood data points of the electrical parameter data points in the electrical parameter data point matrix, adjusting a preset kernel bandwidth parameter based on the angle between the electrical parameter data points and the signal neighborhood data points, the difference in electrical parameter data, and the gradient difference between the electrical parameter data points and the matrix neighborhood data points to obtain the kernel bandwidth parameter of the electrical parameter data points, and performing Gaussian filtering on the corresponding electrical parameter data points using the kernel bandwidth parameter of the electrical parameter data points in the electrical parameter data point matrix to obtain a filtering matrix; and a compression management module for compressing the filtering matrix using the SVD compression algorithm to obtain compressed data. Furthermore, this embodiment filters the electrical parameter data point matrix by dynamically adjusting the kernel bandwidth parameter and performs SVD compression on the filtered matrix. This not only preserves important information but also improves the compression ratio and efficiency. In other words, filtering the electrical parameter data point matrix by dynamically adjusting the kernel bandwidth parameter can improve the compression effect, thereby improving the effect of subsequent remote power management.

[0060] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A concentrator intelligent fusion terminal supporting remote power management, characterized in that, The concentrator intelligent fusion terminal supporting remote power management includes: The data acquisition module is used to acquire the electrical parameter signals of each service user of the concentrator intelligent fusion terminal, wherein the electrical parameter signals are composed of electrical parameter data points; The clustering module is used to obtain the frequency domain signal corresponding to the electrical parameter signal, extract features from the frequency domain signal to obtain the feature vector of each service user, and perform hierarchical clustering of all service users according to the distance between the feature vectors of any two service users to obtain the target cluster. The data filtering module is used to construct an electrical parameter data point matrix of the target cluster based on electrical parameter data points on the electrical parameter signals of the service users in the target cluster, obtain signal neighborhood data points and matrix neighborhood data points of the electrical parameter data points in the electrical parameter data point matrix, adjust a preset kernel bandwidth parameter based on the angle between the electrical parameter data points and the signal neighborhood data points, the difference in electrical parameter data, and the gradient difference between the electrical parameter data points and the matrix neighborhood data points to obtain the kernel bandwidth parameter of the electrical parameter data points, and perform Gaussian filtering on the corresponding electrical parameter data points using the kernel bandwidth parameter of the electrical parameter data points in the electrical parameter data point matrix to obtain a filtering matrix; The compression management module uses the SVD compression algorithm to compress the filter matrix to obtain compressed data; The methods for obtaining signal neighborhood data points and matrix neighborhood data points for each electrical parameter data point include: For any electrical parameter data point in the electrical parameter data point matrix A corresponding to any target cluster: on the electrical parameter signal to which the electrical parameter data point belongs, the electrical parameter data points located to the left of the electrical parameter data point are recorded as the left signal neighbor data points of the electrical parameter data point, and the electrical parameter data points located to the right of the electrical parameter data point are recorded as the right signal neighbor data points of the electrical parameter data point. The left and right signal neighbor data points of the electrical parameter data point belong to the signal neighbor data points of the electrical parameter data point; on the electrical parameter data point matrix A, a rectangular window with a preset length is constructed with the electrical parameter data point as the center, and is recorded as the neighbor window of the electrical parameter data point. All data points located within the neighbor window are recorded as the matrix neighbor data points of the electrical parameter data point; The method for obtaining the kernel bandwidth parameter of the electrical parameter data point includes: Obtain the neighborhood angle of each electrical parameter data point on the electrical parameter data point matrix; For any electrical parameter data point in any electrical parameter data point matrix: Based on the difference in neighborhood angles between the electrical parameter data point and other electrical parameter data points in the same row, and the difference between the mean of the ordinate of the electrical parameter data point and the mean of the ordinates of the electrical parameter data points in the same row, the importance value of the electrical parameter data point is obtained; the set of all matrix neighborhood data points of the electrical parameter data point is denoted as the matrix neighborhood set; the gradient difference value between the electrical parameter data point and each matrix neighborhood data point in the matrix neighborhood set is obtained; the gradient difference value between the electrical parameter data point and the f-th matrix neighborhood data point in the matrix neighborhood set is... Norm() is the normalization function. The gradient magnitude of the electrical parameter data points. Let f be the gradient magnitude of the f-th matrix neighborhood data point in the matrix neighborhood set. The gradient direction of the electrical parameter data points. The gradient direction of the f-th matrix neighborhood data point in the matrix neighborhood set is defined as follows: the mean of the gradient differences between the electrical parameter data point and all matrix neighborhood data points in the matrix neighborhood set is used as the matrix local perturbation characterization value of the electrical parameter data point; the preset kernel bandwidth parameter is adjusted according to the importance characterization value and the matrix local perturbation characterization value of the electrical parameter data point to obtain the kernel bandwidth parameter of the electrical parameter data point.

2. The concentrator intelligent fusion terminal supporting remote power management as described in claim 1, characterized in that, Methods for obtaining target clusters by hierarchical clustering of all service users based on the distance between the feature vectors of any two service users include: Obtain the distance between the feature vectors of any two service users and record it as the metric distance between the corresponding service users. The feature vector of the service user is composed of the mean, variance, main frequency, total harmonic distortion rate and spectral entropy of the frequency domain signal corresponding to the electrical parameter signal of the corresponding service user. All service users are hierarchically clustered based on the metric distance between any two service users to obtain a clustering tree. Based on the distance between clusters at each level of the clustering tree, the metric distance between service users in each cluster at each level, and the number of clusters at each level, the clustering result representation value corresponding to each level of the clustering tree is obtained. The level corresponding to the largest clustering result representation value is selected as the target level, and all clusters at the target level are recorded as target clusters.

3. A concentrator intelligent fusion terminal supporting remote power management as described in claim 2, characterized in that, Methods for obtaining the representation values ​​of clustering results include: For the i-th layer of the clustering tree: In each cluster of the i-th layer, the service user closest to the cluster center of the corresponding cluster is obtained and recorded as the representative user of the corresponding cluster; the cluster distance representation value corresponding to the i-th layer is obtained based on the metric distance between the representative users of the clusters of the i-th layer; the user distribution density representation value corresponding to each cluster of the i-th layer is obtained based on the metric distance between any two service users in each cluster of the i-th layer; the mean of the user distribution density representation values ​​corresponding to all clusters of the i-th layer is used as the cluster density representation value corresponding to the i-th layer; the product of the reciprocal of the number of clusters of the i-th layer, the cluster distance representation value corresponding to the i-th layer, and the cluster density representation value corresponding to the i-th layer is used as the clustering result representation value corresponding to the i-th layer.

4. A concentrator intelligent fusion terminal supporting remote power management as described in claim 3, characterized in that, The methods for obtaining the cluster distance representation value corresponding to the i-th layer include: Obtain the cluster distance set corresponding to each cluster in the i-th layer. The a-th cluster distance in the cluster distance set corresponding to the j-th cluster in the i-th layer is the metric distance between the representative user of the j-th cluster and the representative user of the a-th cluster in the set to be analyzed corresponding to the j-th cluster. The set to be analyzed corresponding to the j-th cluster consists of the remaining clusters in the i-th layer except for the j-th cluster. The mean of the cluster distance set corresponding to each cluster in the i-th layer is denoted as the mean cluster distance of the corresponding cluster, and the mean of the mean cluster distances corresponding to all clusters in the i-th layer is denoted as the cluster distance representation value of the i-th layer.

5. A concentrator intelligent fusion terminal supporting remote power management as described in claim 3, characterized in that, The methods for obtaining the user distribution density representation values ​​corresponding to each cluster in the i-th layer include: For any cluster in the i-th layer, all service users in the cluster are arranged in pairs without repetition to obtain all user combinations corresponding to the cluster. The reciprocal of the metric distance between two service users in the user combination is recorded as the user tightness index value of the corresponding user combination. The mean of the user tightness index values ​​of all user combinations corresponding to the cluster is used as the user distribution tightness characterization value of the cluster.

6. A concentrator intelligent fusion terminal supporting remote power management as described in claim 1, characterized in that, The method for obtaining the neighborhood angle of each electrical parameter data point includes: For any electrical parameter data point h, the line connecting the electrical parameter data point h to the signal neighbor data point to the left of the electrical parameter data point h is denoted as the first line segment, the line connecting the electrical parameter data point h to the signal neighbor data point to the right of the electrical parameter data point h is denoted as the second line segment, and the angle at which the second line segment is rotated counterclockwise to the first line segment is denoted as the neighborhood angle of the electrical parameter data point h.

7. A concentrator intelligent fusion terminal supporting remote power management as described in claim 1, characterized in that, The method for obtaining the importance characterization values ​​of the electrical parameter data points includes: In the row to which the electrical parameter data point belongs, the set of all electrical parameter data points other than the stated electrical parameter data point is denoted as the row set. The characteristic angle ratio between the stated electrical parameter data point and each electrical parameter data point in the row set is obtained. The characteristic angle ratio between the stated electrical parameter data point and the g-th electrical parameter data point in the row set is... Rh is the included angle of the neighborhood of the electrical parameter data point. Let the angle be the neighborhood angle of the g-th electrical parameter data point in the same set. The acquisition time interval between the electrical parameter data point and the g-th electrical parameter data point; the mean of the characteristic angle ratio between the electrical parameter data point and all electrical parameter data points in the same row set is normalized and used as the normalized angle difference characterization value; the mean of the ordinates of all electrical parameter data points in the row to which the electrical parameter data point belongs is used as the row mean of the electrical parameter data point; the absolute value of the difference between the ordinate value of the electrical parameter data point and the row mean is normalized and used as the deviation characterization value; the mean of the normalized angle difference characterization value and the deviation characterization value is recorded as the importance characterization value of the electrical parameter data point.

8. A concentrator intelligent fusion terminal supporting remote power management as described in claim 1, characterized in that, A method for obtaining the kernel bandwidth parameters of the electrical parameter data points by adjusting preset kernel bandwidth parameters based on the importance characterization values ​​and matrix local perturbation characterization values ​​of the electrical parameter data points includes: The reciprocal of the result obtained by adding the importance characterization value to a preset first constant and then normalizing the result is used as the first adjustment factor for the electrical parameter data point. The result of normalizing the matrix local perturbation characterization value of the electrical parameter data point is used as the second adjustment factor for the electrical parameter data point. The weighted sum of the first adjustment factor and the second adjustment factor is calculated and then normalized, and this result is used as the kernel bandwidth parameter adjustment coefficient for the electrical parameter data point. The product of the kernel bandwidth parameter adjustment coefficient and a preset kernel bandwidth parameter is used as the kernel bandwidth parameter for the electrical parameter data point.

Images