A data feature analysis method and system for a smart meter

Abstract

The application discloses a data feature analysis method and system for a smart meter, and relates to the technical field of data processing and analysis.The method comprises the following steps: collecting daily power consumption data of all power users in a target area within a preset time period and preprocessing the daily power consumption data to obtain corresponding daily power consumption time series and a daily power consumption curve; calculating the similarity between adjacent daily power consumption curves of each user to obtain a stability characteristic value of the power consumption change of the user; combining the stability characteristic value of the power consumption change of the user, and calculating the optimal Euclidean distance between the daily power consumption time series of different users; clustering the users in the target area based on the optimal Euclidean distance; and analyzing the power consumption behavior of the users based on the clustering result. By measuring the stability of the power consumption behavior of the users, the application solves the clustering deviation problem caused by the traditional method which ignores the time sequence fluctuation, improves the accuracy of clustering, and effectively improves the precision of power consumption behavior analysis and the reliability of abnormality detection.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and analysis, and in particular to a data feature analysis method and system for a smart meter. Background Art

[0002] With the global energy transition and the accelerated development of smart grids, the deployment of electricity meters, a key terminal device connecting electricity users to the grid, has seen explosive growth. Furthermore, with the continuous development of the power industry and the advancement of smart grids, the performance and functionality requirements for electricity meters are becoming increasingly stringent. Compared to traditional meters, which only provide periodic, cumulative electricity usage data, smart meters offer powerful data acquisition, storage, and communication capabilities. They can continuously and densely record multi-dimensional time-series data, including active power, reactive power, voltage, current, power factor, and frequency, along with associated timestamps and device status information such as switch status and fault alarms, at a frequency of minutes or even seconds. This creates a massive, high-dimensional, and dynamically updated dataset of user electricity usage behavior. By analyzing smart meter data corresponding to electricity users in a target area, it is possible to effectively monitor user electricity usage within the target area, providing technical support for the advancement of smart grids.

[0003] Traditional electricity meter monitoring mainly focuses on measuring electrical parameters such as current, voltage, and power. When analyzing user electricity usage behavior, existing technologies use clustering algorithms based on different users' daily electricity consumption to cluster users with similar electricity usage characteristics into one category, so as to analyze user electricity usage characteristics and detect electricity anomalies. However, this technology ignores the changes in users' daily electricity usage, which will affect the subsequent clustering and the process and results of user electricity usage behavior analysis. Summary of the Invention

[0004] To effectively separate users with unstable electricity usage data when their daily electricity usage behavior changes, and to address the clustering bias problem caused by traditional methods that ignore time series fluctuations, the present invention provides a data feature analysis method and system for smart meters. The technical solution is as follows:

[0005] In a first aspect, the present invention provides a data feature analysis method for a smart meter, comprising the following steps: collecting daily electricity consumption data of all electricity users in a target area within a preset time period and preprocessing the data to obtain corresponding daily electricity consumption time series and daily electricity consumption curves; calculating the similarity between adjacent daily electricity consumption curves of each user to obtain a stability characteristic value of the user's electricity consumption changes; calculating the optimal Euclidean distance between the daily electricity consumption time series of different users based on the stability characteristic value of the user's electricity consumption changes; clustering the users in the target area based on the optimal Euclidean distance; and analyzing the users' electricity consumption behavior based on the clustering results.

[0006] The calculation process of the optimal Euclidean distance is as follows: using the standard Euclidean distance calculation formula, the basic Euclidean distance of the corresponding daily electricity consumption time series between two users in the target area is calculated; based on the similarity between the users' adjacent daily electricity consumption curves, the stability characteristic value of each user is obtained; the minimum value of the stability characteristic values ​​of the two users is negatively normalized and multiplied by the corresponding basic Euclidean distance as the optimal Euclidean distance between the corresponding two users.

[0007] Preferably, the periodic interval and time period for collecting data are preset according to the actual application scenario, all electricity users in the target area are numbered, and the smart meter data corresponding to all electricity users in the target area are periodically collected to obtain the daily electricity consumption data of the users within the preset time period; the collected electricity consumption data are aligned according to the timestamp of the data collection period and normalized to obtain the daily electricity consumption time series of each user, and the normalized daily electricity consumption data are plotted into curves that change with time to obtain the daily electricity consumption curves of all electricity users within the preset time period.

[0008] Preferably, the DTW algorithm is used to match the data points on two adjacent daily electricity consumption curves, obtain the number of matching point pairs on the two curves and the Euclidean distance between the matching point pairs, and calculate the similarity between the two adjacent daily electricity consumption curves of each user. The similarity calculation formula is:

[0009]

[0010] Where, Represents a user During the preset period Article and The similarity between the daily electricity consumption curves, represents the number of matching point pairs, Indicates the Article and The daily electricity consumption curve The Euclidean distance between pairs of matching points, Indicates the Article and The daily electricity consumption curve The time difference between the matching point pairs, is the setting parameter for normalizing the time difference, Indicates the Article and The standard deviation of the time difference between all matching point pairs on the daily electricity consumption curve, The value range of is related to the preset time period and is a positive integer, represents the natural exponential function.

[0011] Preferably, the stability characteristic value of the power consumption change of all power users in the target area is calculated by the similarity between the adjacent daily power consumption curves, wherein the user The stability characteristic value of the power consumption change within the preset time period is , the calculation formula is:

[0012]

[0013] Where, Represents a user The standard deviation of the similarity between two adjacent daily electricity consumption curves within the preset time period, Represents a user The average similarity between two adjacent daily electricity consumption curves within a preset time period.

[0014] Preferably, the optimal Euclidean distance of the daily power consumption time series between all electricity users in the target area is calculated in sequence; wherein, the user With users The optimal Euclidean distance between the corresponding daily electricity consumption time series is , the calculation formula is:

[0015]

[0016] Where, Represents a user With users The basic Euclidean distance between the corresponding daily electricity consumption time series, Indicates user and users The minimum value among the stability eigenvalues.

[0017] Preferably, the DBSCAN clustering algorithm is used to cluster the data points in the user's daily electricity consumption time series, where the minimum number of points is 3.

[0018] Preferably, users belonging to the same cluster are combined and analyzed to extract the electricity usage pattern behaviors of users in the same cluster, and the differences between the electricity usage behaviors of users in different clusters are compared and analyzed; for users who do not belong to any cluster, their electricity usage behaviors are marked as abnormal.

[0019] In a second aspect, the present invention provides a data feature analysis system for a smart meter, which is used to implement the above-mentioned data feature analysis method, including: a processor, a memory and a communication interface, and the processor stores computer program instructions for implementing the data feature analysis method.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] The application calculates the similarity of two adjacent daily electricity consumption curves of a user by introducing a DTW algorithm, quantifies the volatility of the user's own electricity consumption, so that the similarity calculation can reflect the curve shape and the time misalignment degree at the same time, and then converts the time misalignment degree and the curve shape difference into a calculable stability index, which can effectively handle the misalignment problem of time series and more accurately measure the stability of the user's electricity consumption behavior; at the same time, the stability feature is used to improve the Euclidean distance between users, so that the distance between users with low stability and other users is enlarged, avoiding the clustering of users with low stability and users with high stability into one category, separating users with unstable electricity consumption behavior, avoiding the interference of users with unstable electricity consumption behavior on the clustering result of normal users, solving the clustering deviation problem caused by ignoring time series fluctuations in traditional methods, improving the accuracy of clustering, and thus effectively improving the accuracy of electricity consumption behavior analysis and the reliability of anomaly detection. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 An implementation flowchart of the data feature analysis method is shown in

[0023] Figure 2 An implementation flowchart of the data feature analysis method is shown in DETAILED DESCRIPTION

[0024] The technical features of the application will be further described in detail below in combination with the drawings so that those skilled in the related art can understand.

[0025] A data feature analysis method for a smart meter, the implementation flowchart is shown in Figure 1 The specific implementation steps are as follows:

[0026] Step S1: Collecting daily electricity consumption data of all electricity users in a target area within a preset time period and pre-processing to obtain corresponding daily electricity consumption time series and daily electricity consumption curves;

[0027] Specifically, according to the actual application scenario, the periodic interval and time period of data collection are preset, all electricity users in the target area are numbered, and the corresponding smart meter data of all electricity users in the target area are periodically collected to obtain the daily electricity consumption data of the users within the preset time period; the collected electricity consumption data is aligned according to the time stamp of the data collection period, and normalized to obtain the daily electricity consumption time series of each user, and the normalized daily electricity consumption data is plotted into a curve varying with time, and the daily electricity consumption curves of all electricity users within the preset time period are obtained;

[0028] The period interval of collecting data is usually set as half an hour, the preset time period is 10 days, the daily electricity consumption data of each user is normalized, and the normalized daily electricity consumption data is used as the daily electricity consumption time sequence of the corresponding user in the preset time period, so that 10 columns of daily electricity consumption time sequences and 10 daily electricity consumption curves are obtained.

[0029] The normalization processing of the collected electricity consumption data is as follows: obtaining the maximum value in all electricity consumption data, and taking the ratio of the value of each electricity consumption data to the maximum value of the electricity consumption data as the corresponding normalized value, so that the electricity consumption data values of all electricity users in the target area are normalized to the range of 0 to 1.

[0030] The purpose of the normalization processing is to scale the electricity consumption data to a unified scale, eliminate the dimension influence, and ensure the fairness of the Euclidean distance calculation.

[0031] Step S2: calculating the similarity between the adjacent daily electricity consumption curves of each user, and obtaining the stability feature value of the daily electricity consumption change of the user.

[0032] In the process of analyzing the user electricity consumption behavior, the prior art usually only depends on the total electricity consumption or the daily average electricity consumption of the user in a period of time to realize clustering analysis of different users, and then analyzes the fault through other power parameters such as voltage, current and power. However, the prior art cannot accurately and effectively cluster and analyze the changes in the daily electricity consumption behavior of the user, and ignores the influence of the stability of the daily electricity consumption change of the user on the subsequent clustering and user electricity consumption behavior analysis. Therefore, it is necessary to analyze the stability of the electricity consumption of the user to identify and separate the users with unstable electricity consumption behavior and improve the accuracy of the subsequent clustering results.

[0033] The stability feature value of the daily electricity consumption change of all electricity users in the target area in the preset time period is obtained, and the specific process is as follows:

[0034] The DTW algorithm is used to match the data points on the adjacent two daily electricity consumption curves, obtain the number of matching point pairs and the Euclidean distance between the matching point pairs on the two curves, and calculate the similarity between the adjacent two daily electricity consumption curves of each user. The DTW algorithm, i.e. dynamic time warping method, can calculate the optimal matching path of two time sequence curves, solve the time misalignment problem such as electricity consumption peak shift, and thus quantify the similarity between the two curves. The similarity between the The similarity between the th and the th daily electricity consumption curves in the preset time period is , and the calculation formula is as follows:

[0035]

[0036] Where, represents the number of matching point pairs, Indicates the Article and The daily electricity consumption curve The Euclidean distance between pairs of matching points, Indicates the Article and The daily electricity consumption curve The time difference between the matching point pairs, is the setting parameter for normalizing the time difference, Indicates the Article and The standard deviation of the time difference between all matching point pairs on the daily electricity consumption curve, The value range is related to the preset time period and is a positive integer, Represents a natural exponential function. When the preset time period is 10 days, The value range of is 1 to 9;

[0037] Among them, the Euclidean distance between matching point pairs is weighted by the time difference between them, which can quantify the impact of the degree of time misalignment on the distance between two adjacent daily electricity consumption curves of users. The larger the time difference, the larger the Euclidean distance between the corresponding matching point pairs and the lower the similarity.

[0038] The larger the value, the more likely the user The more similar the daily electricity consumption characteristics are and the more stable the changes in electricity consumption are; The smaller the value of , the closer the distance between the matching point pairs is and the higher the similarity is. Indicates the time difference between the matching point pairs, which is the time offset of the matching point pairs on the two curves. For example, the difference between 08:00 and 08:30 is 0.5 hours, which is used for weighted Euclidean distance. The larger the value of , the higher the degree of misalignment between the corresponding matching point pairs, the greater the difference between the two adjacent daily electricity consumption curves of the corresponding user, and the more unstable the electricity consumption change of the user; Used to standardize the time difference, which can be selected according to the actual situation. Usually, the definition For two hours, set The value of is 2, and the unit is hour; The larger the value of , the greater the numerical difference in the time difference between all matching point pairs on the two adjacent daily electricity consumption curves of the corresponding user, that is, the lower the similarity between the two daily electricity consumption curves, and the lower the stability of the corresponding user's electricity consumption changes; The natural constant The exponential function with base is negatively correlated with the similarity value, and can map the similarity value to between 0 and 1.

[0039] In addition, the stability characteristic value of the electricity consumption change of all electricity users in the target area is calculated by the similarity between the adjacent daily electricity consumption curves. The higher the stability characteristic value, the more similar the daily electricity consumption curves of the user in the preset period are, the smaller the difference in electricity consumption change is, and the more stable the electricity consumption behavior of the corresponding user is. The stability characteristic value of the power consumption change within the preset time period is , the calculation formula is:

[0040]

[0041] Where, Represents a user The standard deviation of the similarity between two adjacent daily electricity consumption curves within the preset time period, Represents a user The average similarity between each two adjacent daily electricity consumption curves within the preset time period; the stability characteristic value is inversely proportional to the stability of the user's electricity consumption behavior. The larger the stability characteristic value, the more severe the corresponding user's electricity consumption fluctuations and the worse the stability of the electricity consumption behavior.

[0042] Step S3: Calculate the optimal Euclidean distance between different users’ daily electricity consumption time series based on the stability characteristic value of the user’s electricity consumption changes;

[0043] In traditional clustering algorithms, when calculating the Euclidean distance between different user data points, users with different stability tend to be grouped together. Therefore, it is necessary to improve the Euclidean distance calculation formula by utilizing the stability characteristics of user power consumption changes to avoid grouping users with low stability and users with high stability together. This effectively separates users with unstable power consumption behavior, avoids interfering with the clustering results of normal users, and improves clustering accuracy.

[0044] Specifically, the calculation process of the Euclidean distance in the traditional clustering algorithm is improved to be applicable to the application scenario of the present invention. The calculation process of the optimal Euclidean distance is as follows:

[0045] The standard Euclidean distance calculation formula is used to calculate the basic Euclidean distance of the daily electricity consumption time series between two users in the target area. The minimum value of the stability characteristic value between the two users is taken and negative correlation normalization is performed on the minimum value, that is, the natural exponential function is used for reverse mapping to obtain the mapping parameter of the stability characteristic. The product of the mapping parameter and the basic Euclidean distance is used as the optimal Euclidean distance between the corresponding two users. The optimal Euclidean distance of the daily electricity consumption time series between all electricity users in the target area is calculated in sequence.

[0046] Among them, users With users The optimal Euclidean distance between the corresponding daily electricity consumption time series is , and its calculation formula is:

[0047]

[0048] Where, Represents a user With users The basic Euclidean distance between the corresponding daily electricity consumption time series is calculated as follows:

[0049]

[0050] The total number of data elements representing the electricity consumption of each user, Represents a user The normalized value of the mth data element, Represents a user The normalized value of the mth data element, when the data collection interval is set to half an hour and the preset time period is 10 days, The value of is 480. The basic Euclidean distance can measure the difference in average daily electricity consumption between users in the traditional clustering method;

[0051] Indicates user and users Taking the minimum value of the stability characteristic values ​​between two users can ensure that the Euclidean distance is amplified when any user is unstable, and widen the Euclidean distance between the daily power consumption time series of users with low stability and other users, which can avoid classifying users with low stability and other users into the same category, thereby preventing users with unstable electricity consumption behavior from interfering with subsequent analysis; represents the natural exponential function.

[0052] Step S4: Clustering users in the target area based on the optimal Euclidean distance;

[0053] Specifically, the DBSCAN clustering algorithm is used to cluster the data points in the user's daily electricity consumption time series based on the corresponding optimal Euclidean distance. This density-based spatial clustering method can automatically identify clusters and exclude noise points, that is, unstable users, without the need to preset the number of clusters; wherein, the minimum number of points is set to 3, and the cluster radius is selected according to the actual situation; since the electricity consumption data is normalized in step S1, the value of the cluster radius should be selected according to the actual situation. Under normal circumstances, when the preset time period is 10 days, the cluster radius can be set to 0.1.

[0054] Step S5: Analyze the user's electricity usage behavior based on the clustering results;

[0055] Specifically, users belonging to the same cluster are combined for analysis, their electricity usage patterns are extracted, and differences in electricity usage behaviors between users in different clusters are compared and analyzed. For users who do not belong to any cluster, their electricity usage behaviors are marked as abnormal. After further investigation and confirmation of the corresponding user's electricity failure, an alarm is issued.

[0056] Since it avoids clustering users with low stability and users with high stability into one category, the accuracy of clustering is improved. When subsequently analyzing the data characteristics of electricity users in the target area, the electricity behavior analysis based on the clustering results can be more accurate and effective. Moreover, by separating users with unstable electricity behavior, users with abnormal electricity usage can be quickly located, and rapid and effective anomaly detection and analysis of electricity failures can be achieved, thereby improving the accuracy of electricity behavior analysis and the reliability of anomaly detection.

[0057] The present invention also discloses a data feature analysis system for smart meters, the structure of which is as follows: Figure 2 As shown, the method for implementing the above-mentioned data feature analysis method includes: a processor, a memory and a communication interface. The processor stores computer program instructions for implementing the above-mentioned data feature analysis method, the memory contains several databases for storing data, and the communication interface is used to receive and transmit data.

[0058] The embodiments included in the present invention are merely descriptions of preferred implementation methods of the present invention and are not limited to the precise structures described above and shown in the accompanying drawings. Various modifications and changes can be made without departing from the scope of protection of the present invention. Without departing from the design concept of the present invention, various variations and improvements made to the technical solutions of the present invention by engineers and technicians in this field should fall within the scope of protection of the present invention.

Claims

1. A data feature analysis method for a smart meter, characterized by: Daily electricity consumption data for all electricity users in the target area within a preset time period is collected and preprocessed to obtain corresponding daily electricity consumption time series and daily electricity consumption curves. The similarity between adjacent daily electricity consumption curves of each user is calculated to obtain the stability characteristic value of the user's electricity consumption changes. Combined with the stability characteristic value of the user's electricity consumption changes, the optimal Euclidean distance between the daily electricity consumption time series of different users is calculated. Based on the optimal Euclidean distance, users in the target area are clustered. Based on the clustering results, the users' electricity consumption behavior is analyzed. The optimal Euclidean distance is calculated as follows: using the standard Euclidean distance calculation formula, the basic Euclidean distance of the corresponding daily electricity consumption time series between two users in the target area is calculated; based on the similarity between the user's adjacent daily electricity consumption curves, the stability characteristic value of each user is obtained; the minimum value of the stability characteristic values ​​of the two users is normalized for negative correlation and multiplied by the corresponding basic Euclidean distance as the optimal Euclidean distance between the corresponding two users; The DTW algorithm is used to match the data points on two adjacent daily electricity consumption curves. The number of matching point pairs on the two curves and the Euclidean distance between the matching point pairs are obtained. The similarity between the two adjacent daily electricity consumption curves of each user is calculated. The similarity calculation formula is: Where, Represents a user During the preset period Article and The similarity between the daily electricity consumption curves, represents the number of matching point pairs, Indicates the Article and The daily electricity consumption curve The Euclidean distance between pairs of matching points, Indicates the Article and The daily electricity consumption curve The time difference between the matching point pairs, is the setting parameter for normalizing the time difference, Indicates the Article and The standard deviation of the time difference between all matching point pairs on the daily electricity consumption curve, The value range of is related to the preset time period and is a positive integer, represents the natural exponential function; The stability characteristic value of the power consumption change of all power users in the target area is calculated by the similarity between the adjacent daily power consumption curves, wherein the user The stability characteristic value of the power consumption change within the preset time period is , the calculation formula is: Where, Represents a user The standard deviation of the similarity between two adjacent daily electricity consumption curves within the preset time period, Represents a user The average similarity between two adjacent daily electricity consumption curves within a preset time period; The optimal Euclidean distance of the daily electricity consumption time series between all electricity users in the target area is calculated in sequence; With users The optimal Euclidean distance between the corresponding daily electricity consumption time series is , the calculation formula is: Where, Represents a user With users The basic Euclidean distance between the corresponding daily electricity consumption time series, Indicates user and users The minimum value among the stability eigenvalues.

2. The data feature analysis method for a smart meter according to claim 1, characterized in that: The method collects daily electricity consumption data of all electricity users in the target area within a preset time period and pre-processes it to obtain a corresponding daily electricity consumption time series and a daily electricity consumption curve, including: pre-setting the periodic interval and time period for data collection according to the actual application scenario, numbering all electricity users in the target area, periodically collecting smart meter data corresponding to all electricity users in the target area, and obtaining the daily electricity consumption data of the users within the preset time period; aligning the collected electricity consumption data according to the timestamp of the data collection period and normalizing it to obtain a daily electricity consumption time series for each user, and plotting the normalized daily electricity consumption data into curves that change with time to obtain daily electricity consumption curves for all electricity users in the preset time period.

3. The data feature analysis method for a smart meter according to claim 2, characterized in that: Clustering users in the target area based on the optimal Euclidean distance includes: clustering data points in the user's daily electricity consumption time series using a DBSCAN clustering algorithm, wherein the minimum number of points is 3.

4. The data feature analysis method for a smart meter according to claim 3, characterized in that: The analysis of the user's electricity usage behavior based on the clustering results includes: merging and analyzing users belonging to the same cluster, extracting the electricity usage pattern behavior of users in the same cluster, and comparing and analyzing the differences between the electricity usage behaviors of users in different clusters; for users who do not belong to any cluster, marking their electricity usage behavior as abnormal.

5. A data feature analysis system for a smart meter, characterized by: The invention comprises a processor, a memory and a communication interface, wherein the processor stores computer program instructions for implementing the data feature analysis method according to any one of claims 1 to 4.

Images