Electric energy meter with intelligent monitoring function for user electricity consumption
By filtering and clustering recent historical monitoring data of electricity meters and dynamically adjusting the monitoring frequency, the problem of monitoring accuracy of electricity meters during load fluctuations is solved, and the effect of timely monitoring during faults and anomalies is achieved rather than reducing redundant data during anomalies.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG DEYUAN POWER TECHNOLOGY CORP LTD
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-21
AI Technical Summary
Existing electricity meters have insufficient data resolution when the load fluctuates drastically, making it impossible to capture faults and anomalies in a timely manner. When the load is stable, they generate redundant data, resulting in insufficient monitoring accuracy.
By analyzing recent historical monitoring data of electricity meters for dramatic changes and outliers, suspected abnormal monitoring data is screened out. Then, one-dimensional clustering and fault anomaly confidence analysis are used to dynamically adjust the monitoring frequency to improve monitoring accuracy.
This enables the monitoring frequency to be increased when the risk of faults is high and the monitoring frequency to be reduced when there are no faults, thus avoiding the generation of redundant data and ensuring the monitoring accuracy of the electricity meter.
Smart Images
Figure CN122084973B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to an energy meter with intelligent monitoring function for user electricity consumption. Background Technology
[0002] An electricity meter is a power metering device used to measure a user's electricity consumption, primarily recording the electrical energy consumed within a certain time period. It is typically installed at the user's incoming power line and is a fundamental piece of equipment for power supply companies to conduct electricity billing, electricity management, and operational analysis. Modern smart electricity meters also have the function of monitoring electrical parameters such as voltage, current, and power, using this data to determine if there are any faults or abnormalities in the user's circuit.
[0003] Existing electricity meters typically collect electricity data using a fixed monitoring frequency. However, this method suffers from insufficient data resolution during periods of drastic load fluctuations, failing to promptly detect faults and anomalies. Conversely, when the load is stable, redundant data is generated, consuming the meter's signal transmission path and terminal storage space. Therefore, dynamically adjusting the meter's monitoring frequency to improve the accuracy of electricity consumption monitoring is a pressing issue that needs to be addressed. Summary of the Invention
[0004] In view of this, embodiments of the present invention provide an energy meter with a smart monitoring function for user electricity consumption, in order to solve the problem of how to dynamically adjust the monitoring frequency of the energy meter to improve the accuracy of the energy meter's monitoring of user electricity consumption.
[0005] This invention provides an energy meter with intelligent monitoring function for user electricity consumption, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps:
[0006] Obtain the recent historical monitoring data sequence of each dimension collected by the current user's electricity meter at the current moment. The recent historical monitoring data sequence refers to the sequence composed of the monitoring data at the current moment and the monitoring data within the previous preset time period. Perform data change and outlier analysis on each monitoring data in the recent historical monitoring data sequence of each dimension to filter out suspected abnormal monitoring data in the recent historical monitoring data sequence of each dimension.
[0007] For any dimension, one-dimensional clustering is performed based on the time interval between suspected abnormal monitoring data in the recent historical monitoring data sequence under the given dimension to obtain multiple clusters. Based on the distribution characteristics of suspected abnormal monitoring data in each cluster, the differences between adjacent suspected abnormal monitoring data, and the overlap time span of the time domain range of suspected abnormal monitoring data in different dimensions, the fault anomaly confidence level of suspected abnormal monitoring data in the recent historical monitoring data sequence under the given dimension is obtained.
[0008] Based on the reliability of fault anomalies in the recent historical monitoring data sequence under each dimension, the risk level of recent monitoring anomalies at the current moment is obtained; based on the risk level of recent monitoring anomalies at the current moment, the adaptive monitoring frequency for future periods is obtained for data monitoring of the current user's electricity meter.
[0009] Preferably, the step of performing data mutation and outlier analysis on each monitoring data in the recent historical monitoring data sequence under each dimension to filter out suspected abnormal monitoring data in the recent historical monitoring data sequence under each dimension includes:
[0010] For any recent historical monitoring data sequence under any dimension, any monitoring data in the recent historical monitoring data sequence under any dimension is taken as the data to be analyzed. The absolute value of the difference between the data to be analyzed and its previous adjacent monitoring data is calculated and recorded as the previous difference value. The absolute value of the difference between the data to be analyzed and its subsequent adjacent monitoring data is calculated and recorded as the subsequent difference value. The sum of the previous difference value and the subsequent difference value is normalized to obtain the degree of adjacent difference of the data to be analyzed.
[0011] Calculate the mean value of the recent historical monitoring data sequence under any dimension, normalize the absolute value of the difference between the data to be analyzed and the mean value of the monitoring data to obtain the outlier degree of the data to be analyzed; and record the sum of the adjacent difference degree and the outlier degree of the data to be analyzed as the suspected anomaly degree of the data to be analyzed.
[0012] Based on the degree of suspected anomaly of each monitoring data in the recent historical monitoring data sequence under any dimension, suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension are selected.
[0013] Preferably, the step of filtering suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension based on the degree of suspected anomaly of each monitoring data in the recent historical monitoring data sequence under any dimension includes:
[0014] If the suspected anomaly level of any monitoring data in the recent historical monitoring data sequence under any dimension is greater than or equal to the preset suspected anomaly level threshold, then the suspected anomaly monitoring data is recorded as suspected anomaly monitoring data.
[0015] Preferably, the step of obtaining the fault anomaly confidence level of the suspected anomaly monitoring data in the recent historical monitoring data sequence under any dimension based on the distribution characteristics of the suspected anomaly monitoring data in each cluster, the differences between adjacent suspected anomaly monitoring data, and the overlap time span of the time domain range of the clusters where the suspected anomaly monitoring data are located in different dimensions, includes:
[0016] For any suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension, the non-noise distribution feature value of any suspected abnormal monitoring data is obtained based on the number of suspected abnormal monitoring data in the cluster to which the suspected abnormal monitoring data belongs and the time interval distance between the suspected abnormal monitoring data.
[0017] Obtain the time domain range corresponding to the cluster where any suspected anomaly monitoring data is located. Based on the differences between adjacent suspected anomaly monitoring data of any suspected anomaly monitoring data and the overlap time span between the time domain range and each time domain range in other dimensions, obtain the non-user electricity consumption behavior distribution feature value of any suspected anomaly monitoring data.
[0018] The fault anomaly confidence level of any suspected anomaly monitoring data is obtained by combining the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of any suspected anomaly monitoring data.
[0019] Preferably, obtaining the non-noise distribution characteristic value of any suspected anomaly monitoring data based on the number of suspected anomaly monitoring data in the cluster to which any suspected anomaly monitoring data belongs and the temporal interval distance between the suspected anomaly monitoring data includes:
[0020] The number of suspected anomaly monitoring data in the cluster to which any suspected anomaly monitoring data belongs is counted. The time interval between two adjacent suspected anomaly monitoring data in the cluster to which any suspected anomaly monitoring data belongs is calculated to obtain the mean of the time interval. The product between the number and the reciprocal of the mean of the time interval is normalized to obtain the non-noise distribution characteristic value of any suspected anomaly monitoring data.
[0021] Preferably, the step of obtaining the non-user electricity consumption behavior distribution feature value of any suspected anomaly monitoring data based on the differences between adjacent suspected anomaly monitoring data and the overlap time span between the time domain range and each time domain range in other dimensions includes:
[0022] Calculate the absolute value of the difference between any suspected anomaly monitoring data and its preceding adjacent suspected anomaly monitoring data, and record it as the preceding adjacent difference value. Calculate the absolute value of the difference between any suspected anomaly monitoring data and its following adjacent suspected anomaly monitoring data, and record it as the following adjacent difference value. Normalize the sum of the preceding adjacent difference value and the following adjacent difference value to obtain the degree of local suspected anomaly change of any suspected anomaly monitoring data.
[0023] Dimensions other than any of the above dimensions are denoted as other dimensions. For any other dimension, clusters of suspected anomaly monitoring data under any other dimension are obtained. The time domain range corresponding to each cluster is obtained and denoted as the comparison time domain range. The maximum overlap time span is obtained based on the overlap time span between the time domain range and each comparison time domain range.
[0024] Obtain the maximum overlap time span under each other dimension, obtain the intersection of the maximum overlap time span under all other dimensions, and denote it as the overall overlap time span. Normalize the overall overlap time span to obtain the similarity of the suspected anomaly location of any suspected anomaly monitoring data.
[0025] The product of the degree of local suspected anomaly change and the similarity of suspected anomaly location for any suspected anomaly monitoring data is denoted as the non-user electricity consumption behavior distribution characteristic value of any suspected anomaly monitoring data.
[0026] Preferably, the step of obtaining the fault anomaly confidence level of any suspected anomaly monitoring data by comprehensively analyzing the non-noise distribution characteristic values and non-user electricity consumption behavior distribution characteristic values of any suspected anomaly monitoring data includes:
[0027] The mean between the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of any suspected anomaly monitoring data is calculated to obtain the fault anomaly confidence level of any suspected anomaly monitoring data.
[0028] Preferably, the step of obtaining the recent monitoring anomaly risk level at the current moment based on the failure anomaly confidence level of suspected anomaly monitoring data in the recent historical monitoring data sequence under each dimension includes:
[0029] For any dimension, the failure anomaly confidence level of all suspected anomaly monitoring data in the recent historical monitoring data sequence under that dimension is accumulated to obtain the failure anomaly confidence level accumulation value. The product of the number of suspected anomaly monitoring data under that dimension and the failure anomaly confidence level accumulation value is recorded as the anomaly risk level under that dimension.
[0030] The cumulative value of the abnormal risk level under all dimensions is normalized to obtain the recent monitoring abnormal risk level at the current moment.
[0031] Preferably, obtaining the adaptive monitoring frequency for future periods based on the recent monitoring anomaly risk level at the current moment includes:
[0032] The sum of the recent monitoring anomaly risk level at the current moment and the constant 0.5 is recorded as the frequency adjustment coefficient. The product of the initial monitoring frequency and the frequency adjustment coefficient is rounded down to obtain the adaptive monitoring frequency for the future period.
[0033] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:
[0034] In this invention, suspected abnormal monitoring data in the recent historical monitoring data sequence under each dimension is obtained by filtering. Non-noise distribution characteristic values and non-user electricity consumption behavior distribution characteristic values are then obtained for each suspected abnormal monitoring data to determine the fault anomaly confidence level. This confidence level is used to exclude suspected abnormal monitoring data corresponding to noise and suspected abnormal monitoring data caused by user electricity consumption behavior from all suspected abnormal monitoring data, thereby obtaining more accurate suspected abnormal monitoring data under fault anomalies. Finally, the fault anomaly confidence levels of the suspected abnormal monitoring data under each dimension are combined to construct the recent monitoring anomaly risk level at the current moment. The monitoring frequency of the electricity meter is adaptively adjusted according to the recent monitoring anomaly risk level, so that the monitoring frequency of the electricity meter is higher when there is a high risk of circuit fault anomaly, and lower when there is no risk of fault anomaly, thus preventing redundant data while ensuring the accuracy of electricity meter monitoring. Attached Figure Description
[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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.
[0036] Figure 1 This is a flowchart of a method for implementing an energy meter with intelligent monitoring function for user electricity consumption, provided in Embodiment 1 of the present invention. Detailed Implementation
[0037] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.
[0038] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.
[0039] To illustrate the technical solution of the present invention, specific embodiments are described below.
[0040] See Figure 1 This is a flowchart illustrating a method for implementing an energy meter with intelligent user electricity consumption monitoring function, as provided in Embodiment 1 of the present invention. Figure 1 As shown, the method may include:
[0041] Step S101: Obtain the recent historical monitoring data sequence of each dimension collected by the current user's electricity meter at the current moment. The recent historical monitoring data sequence refers to the sequence composed of the monitoring data at the current moment and the monitoring data within the preset time period before it. Perform data change and outlier analysis on each monitoring data in the recent historical monitoring data sequence of each dimension to filter out suspected abnormal monitoring data in the recent historical monitoring data sequence of each dimension.
[0042] For electricity meters with smart monitoring functions for multi-dimensional electricity consumption data, when setting the monitoring frequency of the electricity meter, a fixed monitoring frequency that is too high will lead to excessive redundant data, while a fixed monitoring frequency that is too low will lead to untimely monitoring of faults and anomalies. Therefore, in this embodiment of the invention, in order to improve the monitoring data efficiency of the electricity meter, an adaptive monitoring frequency is dynamically obtained based on the distribution characteristics of the recent historical data of the user's electricity meter for monitoring multi-dimensional electricity consumption data in the future period.
[0043] Specifically, taking a single user as an example, designated as the current user, the electricity meter monitors and collects multi-dimensional monitoring data for this user, mainly including but not limited to voltage, current, and power. To facilitate subsequent calculation and analysis, the initial monitoring frequency for all dimensions is set to be the same, and all are 10Hz. All collected monitoring data is transmitted to the data processing terminal, with the monitoring data for each dimension stored separately in a sequence, meaning one dimension corresponds to one monitoring data sequence.
[0044] Using the current moment as the real-time monitoring point, the recent historical monitoring data sequence of the current user's electricity meter is obtained in each dimension at the current moment. The recent historical monitoring data sequence refers to the sequence composed of the current moment and the monitoring data within the previous preset time period. Preferably, the recent historical monitoring data sequence is composed of the monitoring data within 30 minutes from the current moment, with one recent historical monitoring data sequence for each dimension.
[0045] Furthermore, in this embodiment of the invention, by performing fault and anomaly risk analysis on the recent historical monitoring data sequences under each dimension, the monitoring frequency is dynamically adjusted according to the risk situation. When the fault and anomaly risk is low, the monitoring frequency is reduced to avoid data redundancy; when the fault and anomaly risk is high, the monitoring frequency is increased to avoid missing information from effective monitoring data. However, before performing the fault and anomaly risk analysis at the current moment, data drastic changes and outlier analysis are first performed on each monitoring data in the recent historical monitoring data sequences under each dimension to filter out suspected abnormal monitoring data in the recent historical monitoring data sequences under each dimension. This data is then used for subsequent fault and anomaly risk analysis at the current moment based on the temporal and spatial distribution characteristics of the suspected abnormal monitoring data.
[0046] The method for selecting suspected abnormal monitoring data from recent historical monitoring data sequences under each dimension is as follows: For the recent historical monitoring data sequence under the i-th dimension, the q-th monitoring data in the recent historical monitoring data sequence under the i-th dimension is taken as the data to be analyzed. Since abnormally changing monitoring data usually includes drastic changes, such as drastic changes in current or voltage, indicating a strong possibility of anomaly at that monitoring data point, it also includes monitoring data with excessively skewed values, such as current or voltage that is too low or too high, or too far from normal current or voltage values, which also indicates a strong possibility of anomaly at the corresponding monitoring data point. Therefore, the absolute value of the difference between the data to be analyzed and its preceding adjacent monitoring data is calculated and recorded as the preceding difference value, and the absolute value of the difference between the data to be analyzed and its subsequent adjacent monitoring data is calculated and recorded as the following difference value. The sum of the preceding difference value and the following difference value is normalized to obtain the degree of adjacent difference of the data to be analyzed, which is used to characterize the degree of difference between the data to be analyzed and its adjacent monitoring data. The higher the degree of difference, the greater the possibility of drastic change in the data to be analyzed.
[0047] Simultaneously, the mean value of the recent historical monitoring data sequence under the i-th dimension is calculated, and the absolute value of the difference between the data to be analyzed and the mean value of the monitoring data is normalized to obtain the outlier degree of the data to be analyzed, which is used to characterize whether the value of the data to be analyzed deviates from the normal monitoring data; finally, the sum of the adjacent difference degree and the outlier degree of the data to be analyzed is recorded as the suspected abnormality degree of the data to be analyzed.
[0048] In one implementation, the formula for calculating the suspected anomaly level of the q-th monitoring data is:
[0049]
[0050] in, This indicates the degree of suspected anomaly in the q-th monitoring data. This represents the maximum and minimum value normalization function. This represents the q-th monitoring data. This represents the monitoring data adjacent to the left of the q-th monitoring data. This represents the monitoring data adjacent to the right of the q-th monitoring data, where | represents the absolute value sign. This represents the mean value of the recent historical monitoring data sequence under the i-th dimension.
[0051] It should be noted that, This indicates the degree of difference between the q-th monitoring data and its adjacent monitoring data. The greater the degree of difference, the stronger the drastic change at the q-th monitoring data and the stronger the suspected anomaly of the q-th monitoring data. The mean of all monitoring data in the recent historical data series represents the normal data level. The absolute value of the difference between the q-th monitoring data and the mean is the largest possible difference. The larger the absolute value of this difference, the stronger the outlier the q-th monitoring data is compared with the normal data, and the stronger the suspected abnormality of the q-th monitoring data.
[0052] Similarly, the degree of suspected anomaly of each monitoring data in the recent historical monitoring data sequence under the i-th dimension is obtained. Then, based on the degree of suspected anomaly of each monitoring data in the recent historical monitoring data sequence under the i-th dimension, suspected abnormal monitoring data in the recent historical monitoring data sequence under the i-th dimension is selected. Since the value range of the degree of suspected anomaly is 0-2, the degree of suspected anomaly can be used to divide the suspected abnormal monitoring data and normal data in the recent historical monitoring data sequence under the i-th dimension into two ends of the value range. Therefore, the midpoint of the value range of the degree of suspected anomaly, 1, is used as the threshold of the degree of suspected anomaly. If the degree of suspected anomaly of the q-th monitoring data in the recent historical monitoring data sequence under the i-th dimension is greater than or equal to 1, then the q-th monitoring data is recorded as suspected abnormal monitoring data.
[0053] Based on the method for obtaining suspected abnormal monitoring data in the recent historical monitoring data sequence under the i-th dimension, obtain suspected abnormal monitoring data in the recent historical monitoring data sequence under each dimension.
[0054] Step S102: For any dimension, perform one-dimensional clustering based on the time interval between suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension to obtain multiple clusters. Based on the distribution characteristics of suspected abnormal monitoring data in each cluster, the differences between adjacent suspected abnormal monitoring data, and the overlap time span of the time domain range of the clusters where suspected abnormal monitoring data are located in different dimensions, obtain the fault anomaly credibility of suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension.
[0055] Since suspected anomaly monitoring data cannot be completely determined to be real anomaly data, and monitoring data with strong variability and outliers also includes noise data that is not real anomaly and monitoring data on changes in power data caused by user electricity consumption behavior, based on the suspected anomaly monitoring data in the recent historical monitoring data sequence under each dimension, we further analyze the reliability of each suspected anomaly monitoring data as real fault data based on the temporal and spatial distribution characteristics of each suspected anomaly monitoring data and the similar anomaly time domain under multiple dimensions.
[0056] Taking the p-th suspected anomaly monitoring data in the recent historical monitoring data sequence under the i-th dimension as an example, the method for obtaining the fault anomaly confidence of the p-th suspected anomaly monitoring data is as follows:
[0057] Noise data typically occurs independently and is relatively discretely distributed within recent historical monitoring data sequences. In contrast, genuine abnormal power data often involves sustained abnormal changes over a period of time, such as continuous voltage fluctuations due to voltage instability or concentrated voltage and current jumps caused by poor line contact. However, directly analyzing the distribution of recent historical monitoring data sequences can lead to inaccurate assessments due to interference from anomalies at other time points. For example, there might be suspected fluctuations at time a and time b, but because a and b are far apart, the distance between these two anomalies is factored into the overall data distribution calculation, resulting in inaccurate assessments. Therefore, based on the time intervals between suspected abnormal monitoring data in the recent historical monitoring data sequences under the i-th dimension, K-means clustering is used to perform one-dimensional clustering on all suspected abnormal monitoring data, resulting in multiple clusters. These clusters are used to group suspected abnormal monitoring data that are relatively located in the same time domain into the same cluster. The K value is determined using the elbow method.
[0058] Then, based on the number of suspected anomaly monitoring data in the cluster where the p-th suspected anomaly monitoring data is located and the time interval distance between the suspected anomaly monitoring data, the non-noise distribution characteristic value of the p-th suspected anomaly monitoring data is obtained: count the number of suspected anomaly monitoring data in the cluster where the p-th suspected anomaly monitoring data is located, calculate the time interval between two adjacent suspected anomaly monitoring data in the cluster where the p-th suspected anomaly monitoring data is located, obtain the mean of the time interval, and normalize the product between the number and the reciprocal of the mean of the time interval to obtain the non-noise distribution characteristic value of the p-th suspected anomaly monitoring data.
[0059] In one embodiment, the formula for calculating the non-noise distribution characteristic value of the p-th suspected anomaly monitoring data is:
[0060]
[0061] in, This represents the non-noise distribution characteristic value of the p-th suspected anomaly monitoring data. This represents the maximum and minimum value normalization function. This represents the number of suspected anomaly monitoring data in the cluster to which the p-th suspected anomaly monitoring data belongs. This represents the time interval between two adjacent suspected anomaly monitoring data points in the cluster containing the p-th suspected anomaly monitoring data point. Indicates the number of time intervals.
[0062] It should be noted that the more suspected anomaly monitoring data there are in the cluster where the p-th suspected anomaly monitoring data belongs, the higher the probability that the p-th suspected anomaly monitoring data is non-noise data, and the less it conforms to the characteristic that noise data usually appears independently. This value is used to characterize the density of the distribution of suspected anomaly monitoring data within the cluster containing the p-th suspected anomaly monitoring data. The smaller this value, the denser and more concentrated the distribution of suspected anomaly monitoring data within the cluster containing the p-th suspected anomaly monitoring data is, and the less likely it is to be a cluster of noisy data. Correspondingly, the p-th suspected anomaly monitoring data is less likely to be noisy data. Therefore, an inverse proportional form is used. The smaller the mean time interval between two adjacent suspected anomaly monitoring data in the cluster where the p-th suspected anomaly monitoring data is located, the larger the non-noise distribution characteristic value of the p-th suspected anomaly monitoring data is, and the more likely it is to be a real fault anomaly data.
[0063] However, simply excluding noise features from suspected abnormal monitoring data may leave behind situations where user electricity consumption behavior-related energy data is identified as suspected abnormal monitoring data. Therefore, it is necessary to further obtain the non-user electricity consumption behavior distribution feature value of the p-th suspected abnormal monitoring data based on the overlap time span between the time domain range corresponding to the cluster of the p-th suspected abnormal monitoring data and the time domain range corresponding to the clusters under other dimensions.
[0064] Specifically, changes in electricity consumption caused by user behavior usually exhibit a gradual process, such as the starting and stopping of high-power appliances and load changes, which typically result in high-amplitude changes in protection circuits, presenting a gradual process. However, the clustered fluctuations of fault anomalies are more abrupt. Therefore, the absolute value of the difference between the p-th suspected anomaly monitoring data and its preceding adjacent suspected anomaly monitoring data is calculated and denoted as the preceding adjacent difference value. The absolute value of the difference between the p-th suspected anomaly monitoring data and its subsequent adjacent suspected anomaly monitoring data is also calculated and denoted as the subsequent adjacent difference value. The sum of the preceding adjacent difference value and the subsequent adjacent difference value is normalized to obtain the local suspected anomaly change degree of the p-th suspected anomaly monitoring data. The smaller the difference in data values between adjacent suspected anomaly monitoring data, the stronger the gradual process of the p-th suspected anomaly monitoring data, and the weaker the corresponding fault anomaly degree.
[0065] Simultaneously, the similarity of the cluster containing the p-th suspected anomaly monitoring data to the corresponding time points in other dimensions is analyzed. If the similarity of the cluster positions in other dimensions is higher, it indicates that there are more co-occurring anomalies in other dimensions, and the data is more likely to be abnormal data caused by a fault. If the similarity of the cluster positions in other dimensions is lower, it indicates that there are fewer co-occurring anomalies in other dimensions, and the data is more likely to be caused by changes in power consumption due to user behavior, such as voltage changes caused by load changes or current changes caused by high-power appliances. Therefore, the time domain range corresponding to the cluster containing the p-th suspected anomaly monitoring data is obtained. The time domain range refers to the time period involved in the suspected anomaly monitoring data in the cluster. Then, the dimensions other than the i-th dimension are denoted as other dimensions. For the j-th other dimension, the clusters of suspected anomaly monitoring data under the i-th dimension are obtained according to the clustering method of suspected anomaly monitoring data under the j-th other dimension. Thus, the time domain range corresponding to each cluster is obtained, which is denoted as the comparison time domain range. Based on the overlap time span between the time domain range corresponding to the cluster of the p-th suspected anomaly monitoring data and each comparison time domain range, the maximum overlap time span is obtained. The larger the maximum overlap time span, the more co-occurring anomalies there are in the time domain range corresponding to the cluster of the p-th suspected anomaly monitoring data under the j-th other dimension.
[0066] Similarly, following the method for obtaining the maximum overlapping time span under the j-th other dimension, the maximum overlapping time span under each other dimension is obtained, and the intersection of the maximum overlapping time spans under all other dimensions is obtained, denoted as the overall overlapping time span. The larger the intersection, the larger the overlapping time span under other dimensions, and the higher the position similarity of the cluster where the p-th suspected anomaly monitoring data belongs under other dimensions. The p-th suspected anomaly monitoring data is more likely to be anomaly monitoring data caused by a fault. Then, the overall overlapping time span is normalized to obtain the suspected anomaly position similarity of the p-th suspected anomaly monitoring data.
[0067] Finally, the product of the degree of local suspected anomaly change and the similarity of suspected anomaly location for the p-th suspected anomaly monitoring data is denoted as the non-user electricity consumption behavior distribution characteristic value of the p-th suspected anomaly monitoring data. The formula for calculating the non-user electricity consumption behavior distribution characteristic value of the p-th suspected anomaly monitoring data is as follows:
[0068]
[0069] in, This represents the distribution characteristic value of non-user electricity consumption behavior for the p-th suspected abnormal monitoring data. This represents the maximum and minimum value normalization function. This represents the p-th suspected anomaly monitoring data. This represents the preceding adjacent suspected anomaly monitoring data of the p-th suspected anomaly monitoring data. This represents the next suspected anomaly monitoring data after the p-th suspected anomaly monitoring data, where | represents the absolute value sign. It represents the overall overlap time span, which is the intersection of the maximum overlap time span of the time domain corresponding to the cluster where the p-th suspected anomaly monitoring data belongs in all other dimensions.
[0070] It should be noted that, This is used to characterize the degree of local suspected anomaly change of the p-th suspected anomaly monitoring data. The smaller the difference between the p-th suspected anomaly monitoring data and its adjacent suspected anomaly monitoring data, the stronger the gradual change process of the p-th suspected anomaly monitoring data is, and the more it conforms to the characteristics of user electricity consumption behavior. The smaller the corresponding non-user electricity consumption behavior distribution characteristic value is. The similarity of the suspected anomaly location of the p-th suspected anomaly monitoring data is used to characterize the cluster where the p-th suspected anomaly monitoring data belongs. The greater the overlap of the time span of the time domain corresponding to the cluster where the p-th suspected anomaly monitoring data belongs in other dimensions, the more likely the p-th suspected anomaly monitoring data belongs to the anomaly caused by a real fault. It is less consistent with the change of power data caused by user power consumption behavior, and the larger the corresponding non-user power consumption behavior distribution characteristic value is.
[0071] After obtaining the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of the p-th suspected anomaly monitoring data, the fault anomaly confidence level of the p-th suspected anomaly monitoring data is obtained by combining the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of the p-th suspected anomaly monitoring data: the mean value between the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of the p-th suspected anomaly monitoring data is calculated to obtain the fault anomaly confidence level of the p-th suspected anomaly monitoring data. The formula for calculating the fault anomaly confidence level of the p-th suspected anomaly monitoring data is as follows:
[0072]
[0073] in, This indicates the reliability of the fault anomaly in the p-th suspected anomaly monitoring data. This represents the non-noise distribution characteristic value of the p-th suspected anomaly monitoring data. This represents the distribution characteristic value of non-user electricity consumption behavior for the p-th suspected abnormal monitoring data.
[0074] It should be noted that by using non-noise distribution characteristic values and non-user electricity consumption behavior distribution characteristic values, suspected anomaly monitoring data corresponding to noise and suspected anomaly monitoring data caused by user electricity consumption behavior are excluded from all suspected anomaly monitoring data. This results in more accurate suspected anomaly monitoring data under fault conditions. The larger the non-noise distribution characteristic value and non-user electricity consumption behavior distribution characteristic value of the p-th suspected anomaly monitoring data, the less likely the p-th suspected anomaly monitoring data is to be noise data or data under user electricity consumption behavior, the higher the probability that it belongs to a fault anomaly, and the higher the corresponding fault anomaly confidence. The range of values used to control the confidence level of fault anomalies is [0, 1].
[0075] Following the method for obtaining the fault anomaly confidence level of the p-th suspected anomaly monitoring data in the recent historical monitoring data sequence under the i-th dimension, the fault anomaly confidence level of each suspected anomaly monitoring data in the recent historical monitoring data sequence under the i-th dimension is obtained. This leads to the obtaining of the fault anomaly confidence level of each suspected anomaly monitoring data in the recent historical monitoring data sequence under each dimension.
[0076] Step S103: Based on the reliability of the fault anomaly of the suspected abnormal monitoring data in the recent historical monitoring data sequence under each dimension, obtain the degree of recent monitoring anomaly risk at the current moment; based on the degree of recent monitoring anomaly risk at the current moment, obtain the adaptive monitoring frequency in the future period for data monitoring of the current user's electricity meter.
[0077] By combining the reliability of fault anomalies in the recent historical monitoring data sequences under each dimension with the number of suspected abnormal monitoring data under each dimension, the greater the number of suspected abnormal monitoring data and the higher the reliability of fault anomalies for each suspected abnormal monitoring data, the higher the risk of recent monitoring anomalies in the electricity meter at the current moment, and thus the more necessary it is to adjust the monitoring frequency. Therefore, in this embodiment of the invention, the risk level of recent monitoring anomalies at the current moment is obtained based on the reliability of fault anomalies in the recent historical monitoring data sequences under each dimension. The method for obtaining this information is as follows:
[0078] For any dimension, the fault anomaly confidence level of all suspected abnormal monitoring data in the recent historical monitoring data sequence under that dimension is accumulated to obtain the fault anomaly confidence level accumulation value. The product of the number of suspected abnormal monitoring data under that dimension and the fault anomaly confidence level accumulation value is recorded as the anomaly risk level under that dimension. The accumulation value of the anomaly risk level under all dimensions is normalized to obtain the recent monitoring anomaly risk level at the current moment.
[0079] In one implementation, the formula for calculating the recent monitoring anomaly risk level at the current moment is:
[0080]
[0081] in, This indicates the level of recent monitoring anomaly risk at the current moment. This represents the maximum and minimum value normalization function. Indicates the number of dimensions. This represents the number of suspected anomaly monitoring data in the i-th dimension. This represents the j-th suspected anomaly monitoring data under the i-th dimension.
[0082] Furthermore, after determining the recent monitoring anomaly risk level at the current moment, the monitoring frequency of the current user's electricity meter can be adaptively adjusted based on the recent monitoring anomaly risk level at the current moment to obtain the adaptive monitoring frequency for the future period. The future period refers to 30 minutes after the current moment. The method for obtaining the adaptive monitoring frequency for the future period is as follows: the sum of the recent monitoring anomaly risk level at the current moment and a constant of 0.5 is recorded as the frequency adjustment coefficient. The product of the initial monitoring frequency and the frequency adjustment coefficient is rounded down to obtain the adaptive monitoring frequency for the future period.
[0083] In one embodiment, the formula for calculating the adaptive monitoring frequency in the future time period is:
[0084]
[0085] in, Indicates the adaptive monitoring frequency for the future period. This represents the initial monitoring frequency, and 0.5 represents a constant. This indicates the level of recent monitoring anomaly risk at the current moment. This indicates the floor function.
[0086] It should be noted that, This represents the frequency adjustment coefficient. Since the value of E ranges from 0 to 1, and the median value of 0.5 can classify suspected anomaly monitoring data into high and low anomaly risks, if the recent monitoring anomaly risk level at the current moment is high, the monitoring frequency needs to be increased; if the recent monitoring anomaly risk level at the current moment is low, the monitoring frequency needs to be decreased. Therefore, using 0.5 as the frequency adjustment dividing line, when the value of E is greater than the median value of 0.5, it indicates that the recent monitoring anomaly risk level at the current moment is high, and the corresponding frequency adjustment coefficient increases, resulting in a higher adaptive monitoring frequency. When the value of E is less than the median value of 0.5, it indicates that the recent monitoring anomaly risk level at the current moment is low, and the corresponding frequency adjustment coefficient decreases, resulting in a lower adaptive monitoring frequency. When the value of E is equal to the median value of 0.5, it indicates that the recent monitoring anomaly risk level at the current moment is at an intermediate level, and the initial monitoring frequency meets the monitoring needs under medium risk, so no adjustment is needed. Therefore, the corresponding frequency adjustment coefficient is 1, maintaining the adaptive monitoring frequency as the initial monitoring frequency.
[0087] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 the present invention, and should all be included within the protection scope of the present invention.
Claims
1. An energy meter with intelligent monitoring function for user electricity consumption, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Obtain the recent historical monitoring data sequence of each dimension collected by the current user's electricity meter at the current moment. The recent historical monitoring data sequence refers to the sequence composed of the monitoring data at the current moment and the monitoring data within the previous preset time period. Perform data change and outlier analysis on each monitoring data in the recent historical monitoring data sequence of each dimension to filter out suspected abnormal monitoring data in the recent historical monitoring data sequence of each dimension. For any dimension, one-dimensional clustering is performed based on the time interval between suspected abnormal monitoring data in the recent historical monitoring data sequence under the given dimension to obtain multiple clusters. Based on the distribution characteristics of suspected abnormal monitoring data in each cluster, the differences between adjacent suspected abnormal monitoring data, and the overlap time span of the time domain range of suspected abnormal monitoring data in different dimensions, the fault anomaly confidence level of suspected abnormal monitoring data in the recent historical monitoring data sequence under the given dimension is obtained. Based on the reliability of the fault anomalies in the recent historical monitoring data sequence under each dimension, the risk level of recent monitoring anomalies at the current moment is obtained. Based on the recent monitoring anomaly risk level at the current moment, an adaptive monitoring frequency for future periods is obtained for data monitoring of the current user's electricity meter; The step of obtaining the fault anomaly confidence level of suspected anomaly monitoring data in the recent historical monitoring data sequence under any dimension based on the distribution characteristics of suspected anomaly monitoring data in each cluster, the differences between adjacent suspected anomaly monitoring data, and the overlap time span of the time domain range of clusters containing suspected anomaly monitoring data in different dimensions includes: For any suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension, the non-noise distribution feature value of any suspected abnormal monitoring data is obtained based on the number of suspected abnormal monitoring data in the cluster to which the suspected abnormal monitoring data belongs and the time interval distance between the suspected abnormal monitoring data. Obtain the time domain range corresponding to the cluster where any suspected anomaly monitoring data is located. Based on the differences between adjacent suspected anomaly monitoring data of any suspected anomaly monitoring data and the overlap time span between the time domain range and each time domain range in other dimensions, obtain the non-user electricity consumption behavior distribution feature value of any suspected anomaly monitoring data. The fault anomaly confidence level of any suspected anomaly monitoring data is obtained by combining the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of any suspected anomaly monitoring data. The step of obtaining the non-noise distribution characteristic value of any suspected anomaly monitoring data based on the number of suspected anomaly monitoring data in the cluster to which any suspected anomaly monitoring data belongs and the temporal interval distance between the suspected anomaly monitoring data includes: The number of suspected anomaly monitoring data in the cluster where any suspected anomaly monitoring data is located is counted. The time interval between two adjacent suspected anomaly monitoring data in the cluster where any suspected anomaly monitoring data is located is calculated to obtain the mean of the time interval. The product between the number and the reciprocal of the mean of the time interval is normalized to obtain the non-noise distribution characteristic value of any suspected anomaly monitoring data. The step of obtaining the non-user electricity consumption behavior distribution feature value of any suspected anomaly monitoring data based on the differences between adjacent suspected anomaly monitoring data and the overlap time span between the time domain range and each time domain range in other dimensions includes: Calculate the absolute value of the difference between any suspected anomaly monitoring data and its preceding adjacent suspected anomaly monitoring data, and record it as the preceding adjacent difference value. Calculate the absolute value of the difference between any suspected anomaly monitoring data and its following adjacent suspected anomaly monitoring data, and record it as the following adjacent difference value. Normalize the sum of the preceding adjacent difference value and the following adjacent difference value to obtain the degree of local suspected anomaly change of any suspected anomaly monitoring data. Dimensions other than any of the above dimensions are denoted as other dimensions. For any other dimension, clusters of suspected anomaly monitoring data under any other dimension are obtained. The time domain range corresponding to each cluster is obtained and denoted as the comparison time domain range. The maximum overlap time span is obtained based on the overlap time span between the time domain range and each comparison time domain range. Obtain the maximum overlap time span under each other dimension, obtain the intersection of the maximum overlap time span under all other dimensions, and denote it as the overall overlap time span. Normalize the overall overlap time span to obtain the similarity of the suspected anomaly location of any suspected anomaly monitoring data. The product of the degree of local suspected anomaly change and the similarity of suspected anomaly location for any suspected anomaly monitoring data is denoted as the non-user electricity consumption behavior distribution characteristic value of any suspected anomaly monitoring data.
2. An energy meter with intelligent monitoring function for user electricity consumption as described in claim 1, characterized in that, The process involves analyzing the data variability and outlier characteristics of each monitoring data point in the recent historical monitoring data sequence for each dimension, in order to filter out suspected abnormal monitoring data in the recent historical monitoring data sequence for each dimension, including: For any recent historical monitoring data sequence under any dimension, any monitoring data in the recent historical monitoring data sequence under any dimension is taken as the data to be analyzed. The absolute value of the difference between the data to be analyzed and its previous adjacent monitoring data is calculated and recorded as the previous difference value. The absolute value of the difference between the data to be analyzed and its subsequent adjacent monitoring data is calculated and recorded as the subsequent difference value. The sum of the previous difference value and the subsequent difference value is normalized to obtain the degree of adjacent difference of the data to be analyzed. Calculate the mean value of the recent historical monitoring data sequence under any dimension, normalize the absolute value of the difference between the data to be analyzed and the mean value of the monitoring data to obtain the outlier degree of the data to be analyzed; and record the sum of the adjacent difference degree and the outlier degree of the data to be analyzed as the suspected anomaly degree of the data to be analyzed. Based on the degree of suspected anomaly of each monitoring data in the recent historical monitoring data sequence under any dimension, suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension are selected.
3. An energy meter with intelligent monitoring function for user electricity consumption as described in claim 2, characterized in that, The step of filtering suspected abnormal monitoring data in the recent historical monitoring data sequence under any dimension based on the degree of suspected anomaly of each monitoring data in the recent historical monitoring data sequence includes: If the suspected anomaly level of any monitoring data in the recent historical monitoring data sequence under any dimension is greater than or equal to the preset suspected anomaly level threshold, then the suspected anomaly monitoring data is recorded as suspected anomaly monitoring data.
4. An energy meter with intelligent monitoring function for user electricity consumption as described in claim 1, characterized in that, The reliability of the fault anomaly of any suspected anomaly monitoring data is obtained by combining the non-noise distribution characteristic values and non-user electricity consumption behavior distribution characteristic values of the combined data, including: The mean between the non-noise distribution characteristic value and the non-user electricity consumption behavior distribution characteristic value of any suspected anomaly monitoring data is calculated to obtain the fault anomaly confidence level of any suspected anomaly monitoring data.
5. An energy meter with intelligent monitoring function for user electricity consumption as described in claim 1, characterized in that, The step of obtaining the recent monitoring anomaly risk level at the current moment based on the failure anomaly confidence level of suspected anomaly monitoring data in the recent historical monitoring data sequence under each dimension includes: For any dimension, the failure anomaly confidence level of all suspected anomaly monitoring data in the recent historical monitoring data sequence under that dimension is accumulated to obtain the failure anomaly confidence level accumulation value. The product of the number of suspected anomaly monitoring data under that dimension and the failure anomaly confidence level accumulation value is recorded as the anomaly risk level under that dimension. The cumulative value of the abnormal risk level under all dimensions is normalized to obtain the recent monitoring abnormal risk level at the current moment.
6. An energy meter with intelligent monitoring function for user electricity consumption as described in claim 1, characterized in that, The step of obtaining the adaptive monitoring frequency for future periods based on the recent monitoring anomaly risk level at the current moment includes: The sum of the recent monitoring anomaly risk level at the current moment and the constant 0.5 is recorded as the frequency adjustment coefficient. The product of the initial monitoring frequency and the frequency adjustment coefficient is rounded down to obtain the adaptive monitoring frequency for the future period.