Electricity meter abnormality detection method

Abstract

The present application discloses a method for detecting anomalies in an electric meter, including: collecting first electric energy data, second electric energy data and environmental characteristic data of the electric meter, obtaining third electric energy data based on a preset algorithm and the first electric energy data; obtaining fourth electric energy data based on the environmental characteristic data; based on each second electric energy data, calculating the difference between the corresponding third electric energy data and the second electric energy data to obtain a first difference, calculating the difference between the fourth electric energy data and the second electric energy data to obtain a second difference, and obtaining a target difference based on the first difference and the second difference; generating a target difference sequence within a preset acquisition window; based on a preset anomaly detection mechanism, judging whether the electric meter has an anomaly according to the target difference sequence and deviation index corresponding to each second electric energy data. Thus, based on the third electric energy data and the fourth electric energy data, a dual-path calibration analysis is performed on the second electric energy data to improve the accuracy and comprehensiveness of the anomaly detection of the electric meter.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to an abnormality detection method for an electric meter. Background Art

[0002] The power industry uses three-phase meters to monitor and measure users' energy consumption. The performance of the meter may be affected by the passage of time, aging of components, and the impact of the on-site environment, resulting in inaccurate energy measurement or other problems. It is very necessary to detect meter anomalies in a timely manner to ensure the accuracy and reliability of energy measurement.

[0003] At present, the abnormality detection of electric meters is generally done by manual inspection. The staff regularly go to the site to check whether there are any problems with the electric meter, or go to the site to perform abnormality detection and troubleshoot when the electric meter is abnormal. Obviously, manual inspection not only consumes a lot of manpower and material resources, but also has low accuracy of abnormality detection and low inspection efficiency.

[0004] The Chinese invention patent application number 202311378385.0 discloses a method for detecting abnormalities in an electric meter. The method periodically collects the first power data and the second power data of the electric meter at a preset interval, and then calculates the calibration parameters with the same data type as the second power data based on the first power data and the preset mechanism model. The difference between the second power data and the calibration parameters is then calculated to obtain the target difference. According to the target difference and the target tolerance interval corresponding to the electric meter, it is determined whether the electric meter has an abnormality without the need for manual on-site inspections.

[0005] However, the second power data calculated based on the first power data collected in real time is used as the calibration parameter. If the currently collected first power data itself is abnormal, the power data fluctuation has no significant regularity due to the unknown cause of the meter abnormality, and the calibration parameter will change due to the abnormal fluctuation of the first power data. Comparing such non-true standard calibration parameters with the actually collected second power data as the basis for judging the meter abnormality and timely using the pre-trained fault tolerance model for judgment support may also result in the judgment deviating from reality. In addition, due to the differentiation of the second power data type, the maximum acceptable floating range must also be different, and the unified fault tolerance range of the meter lacks targeted management. Summary of the invention

[0006] The present application provides an electric meter anomaly detection method, performs dual-path calibration analysis on second electric energy data, and improves the accuracy and comprehensiveness of electric meter anomaly detection.

[0007] The present application provides a method for detecting anomalies in an electric meter, comprising:

[0008] S101, collecting first electric energy data, second electric energy data and environmental characteristic data of an electric meter, and obtaining third electric energy data corresponding to the second electric energy data based on a preset algorithm and the first electric energy data;

[0009] S102, converting the environmental characteristic data into a multidimensional feature vector, and inputting the vector into a pre-trained electric energy data prediction model to obtain fourth electric energy data corresponding to the second electric energy data, where the fourth electric energy data is used to represent standard electric energy data of the electric meter in a normal operating state under the current environmental characteristics;

[0010] S103, based on each second electric energy data, calculating the difference between the corresponding third electric energy data and the second electric energy data to obtain a first difference, calculating the difference between the fourth electric energy data and the second electric energy data to obtain a second difference, and obtaining a target difference according to the first difference and the second difference;

[0011] S104, within a preset acquisition window, periodically executing steps S101 to S103 for every preset acquisition unit to obtain a target difference value for each acquisition unit, and generating a target difference value sequence [x1, x2, ..., xn], where n is the number of acquisition units within the acquisition window, and xn is the target difference value of the nth acquisition unit;

[0012] S105, based on a preset abnormality detection mechanism, judging whether the electric meter has an abnormality according to the target difference sequence and the deviation index corresponding to each second electric energy data.

[0013] Preferably, the first electric energy data is electric energy data that can be directly collected without calculation, the second electric energy data is electric energy data that can be directly collected and can be calculated based on the first electric energy data, the number of the second electric energy data is denoted as k, each second electric energy data corresponds to a type of second electric energy data, and the third electric energy data and the fourth electric energy data correspond to it one by one respectively;

[0014] The preset algorithm is set as a relationship formula between the first electric energy data and the second electric energy data, that is, the first electric energy data is input into the corresponding preset algorithm to obtain the theoretical value of the second electric energy data at the current moment as the third electric energy data corresponding to the second electric energy data.

[0015] Preferably, the environmental characteristic data includes environmental parameters, seasonal attributes, and power load attributes. The environmental parameters include environmental temperature and environmental humidity. The power load attribute is set to the power load level of the time period to which the current time belongs, including high, medium, and low.

[0016] The environmental characteristic data is converted into a multidimensional feature vector according to a preset conversion rule, which is expressed as: [(T, H), a, b], wherein T is the ambient temperature, H is the ambient humidity, a is the coded value corresponding to the seasonal attribute, a∈[1,2,3,4], and b is the coded value corresponding to the power load attribute, b∈[1,2,3].

[0017] Preferably, a method for acquiring the deviation index corresponding to the second electric energy data includes:

[0018] B1. Obtain a large amount of second historical electric energy data of the i-th type of second electric energy data collected by the electric meter under normal operation, and refer to steps S101 to S103 to obtain the third historical electric energy data, the fourth historical electric energy data, and the target difference corresponding to each second historical electric energy data; wherein i=1,2,...,k;

[0019] B2. Calculate the deviation index of the second electric energy data of the i-th type according to the following formula:

[0020]

[0021] in, is the deviation index of the second electric energy data of the i-th type of electric meter, is the number of second historical electric energy data of the i-th type of second electric energy data, is the target difference corresponding to the j-th historical second electric energy data of the i-th type of second electric energy data, is the difference between the instantaneous electric energy stability value of the j-th historical second electric energy data and its corresponding third historical electric energy data, is the difference between the instantaneous electric energy stability value of the j-th historical second electric energy data and its corresponding fourth historical electric energy data, and are the adjustment coefficient values ​​of the impact of the target difference and the instantaneous electric energy stability value difference on the deviation index of the i-th second electric energy data.

[0022] Preferably, the and The value of satisfies and , the value determination methods include:

[0023] C1. Select the second historical electric energy data, the third historical electric energy data, and the fourth historical electric energy data corresponding to the i-th type of second electric energy data collected by the electric meter in normal operation and abnormal operation, and obtain several target difference values ​​and their corresponding abnormal results of the electric meter (normal operation and abnormal operation) as the verification data set;

[0024] C2. First, set it randomly and The value of and Then, we can get the deviation index ;

[0025] C3, dynamically adjust according to the validation data set and deviation indicators and The value of (it should be noted that always keep and ), until and The deviation index corresponding to the value of is consistent with the verification results of all verification data sets, and the final deviation index and its corresponding and The value of .

[0026] Preferably, the instantaneous electric energy stability value of the j-th historical second electric energy data is obtained in the following manner:

[0027] In a collection interval consisting of a preset number of collection units before and after the collection moment corresponding to the historical second electric energy data, the standard deviation of the historical second electric energy data corresponding to all collection units in the collection interval is calculated, and the standard deviation is used as the instantaneous electric energy stability value of the jth historical second electric energy data.

[0028] Preferably, the S105 specifically includes:

[0029] S201, based on k second electric energy data, according to the corresponding target difference sequences and corresponding deviation indicators, generate an electric energy abnormality distribution array [P1, P2, ..., Pk] of the electric meter, where Pk is the abnormal probability value of the kth second electric energy data,

[0030] , The acquisition method is as follows: for the kth second electric energy data, traverse its target difference sequence [x1, x2, ..., xn], and count the number of target differences greater than the corresponding deviation index, recorded as ;

[0031] S202, based on the electric energy abnormality distribution array [P1, P2, ..., Pk] of the electric meter, obtain the abnormality index of the electric meter according to the following first calculation formula:

[0032]

[0033] Wherein, Z is the abnormal index of the electric meter, k is the number of the second electric energy data, is the abnormal probability value of the i-th second electric energy data, The weight value of the abnormal probability value of the i-th second electric energy data for the abnormal judgment of the electric meter is set according to expert experience or historical data verification, satisfying ;

[0034] S203, based on the abnormality index of the electric meter and the preset abnormality threshold, if the abnormality index is greater than the preset abnormality threshold, the electric meter is determined to be abnormal, triggering the abnormality handling process.

[0035] Preferably, before S202, the method further includes:

[0036] S301, determining an abnormality detection path according to the distribution of the abnormal power distribution array, the abnormality detection path comprising: jumping to step S302 to perform difference correlation analysis, and jumping to step S202;

[0037] S302, based on each type of second electric energy data of the electric meter, collecting first differences and second differences corresponding to a large amount of second historical electric energy data collected by the electric meter under normal operating conditions, constructing a number of difference data points (second difference, first difference), taking the second difference as the horizontal coordinate and the first difference as the vertical coordinate, performing fitting, and obtaining a difference correlation fitting curve graph;

[0038] S303, for each second electric energy data, inputting the second difference corresponding to each target difference in the target difference sequence into the corresponding difference association fitting curve diagram, and generating a calibration difference corresponding to the first difference;

[0039] S304, based on the first difference corresponding to each target difference in the target difference sequence and the calibration difference corresponding to each target difference, a first difference sequence and a calibration difference sequence are generated, the calibration difference sequence is subtracted from the first difference sequence and an absolute value is taken to obtain a deviation difference sequence [d1, d2, ..., dn], and all deviation differences in the deviation difference sequence are averaged to obtain a target deviation value;

[0040] S305, based on the target deviation value corresponding to each second electric energy data, obtaining a replacement value of the abnormal probability value of each second electric energy data for the weight value of the electric meter abnormality judgment according to the second calculation formula , replace the first calculation formula , execute step S202 to obtain the abnormality index of the electric meter.

[0041] Preferably, the second calculation formula is:

[0042]

[0043] in, is the target deviation value corresponding to the i-th second electric energy data, is the sum of the target deviation values ​​corresponding to the k second electric energy data, is a replacement value of the abnormal probability value of the i-th second electric energy data for the weight value of the electric meter abnormality judgment.

[0044] Preferably, the S301 specifically includes:

[0045] S401, inputting the power abnormal distribution array into a pre-trained probability distribution abnormality recognition model, and outputting the distribution of the power abnormal distribution array, the distribution including: normal distribution and abnormal distribution;

[0046] S402, if the distribution is normal, the abnormal detection path is determined as: jump to step S202; if the distribution is abnormal, the abnormal detection path is determined as: jump to step S302.

[0047] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0048] Based on the second electric energy data, the third electric energy data is introduced through a preset algorithm, and the fourth electric energy data is introduced through the environmental characteristic data. The second electric energy data is calibrated in a dual path, and the internal state of the electric energy meter and the external environmental factors are comprehensively considered, thereby improving the accuracy of the electric energy meter abnormality detection. Even if the first electric energy data fluctuates or is abnormal, relatively accurate standard electric energy data can still be obtained through the calibration of the second path, thereby reducing the misjudgment caused by inaccurate calibration of a single path. The target difference is obtained by combining the first difference and the second difference for comprehensive judgment, thereby more comprehensively reflecting the actual operating state of the electric energy meter and improving the comprehensiveness of the evaluation. The deviation indicators of different types of second electric energy data are differentiated, and based on historical data, the difference between the target difference and the instantaneous electric energy stability value is considered, so that the deviation indicator can dynamically adapt to different types of electric meters and usage environments, thereby improving the flexibility and accuracy of abnormality detection.

[0049] By introducing the electric energy anomaly distribution array, an abnormal probability value can be generated for each second electric energy data, so as to more finely evaluate the abnormal state of the electric meter in different electric energy data dimensions, thereby improving the accuracy of anomaly detection; not only the abnormal situation of a single electric energy data is considered, but also the overall state of the electric meter is comprehensively judged by setting different weight values, avoiding the risk of misjudgment of a single data point.

[0050] Based on the distribution of the electric energy anomaly distribution array, a suitable anomaly detection method is selected, and the detection method can be flexibly adjusted according to the actual situation; by constructing a difference correlation fitting curve diagram, not only the anomaly of a single electric energy data is considered, but also the correlation between the first difference and the second difference is analyzed, so that the anomaly can be identified more accurately; the target deviation value of the first difference is calculated according to the second difference and the replacement value of the weight value is adjusted accordingly, which can dynamically reflect the importance of different second electric energy data in anomaly judgment, thereby improving the accuracy and flexibility of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1The figure is a flow chart of an electric meter abnormality detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0053] It should be noted that the terms “vertical”, “horizontal”, “up”, “down”, “left”, “right” and similar expressions used in this document are only for illustrative purposes and do not represent the only implementation method.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0055] Embodiment 1: Figure 1 It is a flow chart of an electric meter abnormality detection method according to an embodiment of the present invention.

[0056] like Figure 1 As shown, a method for detecting abnormality of an electric meter includes the following steps:

[0057] S101, collecting first electric energy data, second electric energy data and environmental characteristic data of an electric meter, and obtaining third electric energy data corresponding to the second electric energy data based on a preset algorithm and the first electric energy data.

[0058] Specifically, the first electric energy data is electric energy data that can be directly collected and obtained without calculation, the second electric energy data includes electric energy data that can be directly collected and obtained by calculation based on the first electric energy data, the number of second electric energy data is recorded as k, each second electric energy data corresponds to a second electric energy data type, and the third electric energy data corresponds one-to-one to it.

[0059] The preset algorithm is set as a relationship formula between the first electric energy data and the second electric energy data, that is, the first electric energy data is input into the corresponding preset algorithm to obtain the theoretical value of the second electric energy data at the current moment, as the third electric energy data corresponding to the second electric energy data. For example, the first electric energy data includes phase current and phase voltage, and the second electric energy data includes phase apparent power. The corresponding preset algorithm is: phase apparent power = phase current × phase voltage. The first electric energy data is input into the preset algorithm, and the obtained phase apparent power is used as the third electric energy data.

[0060] In actual power scenarios, due to the passage of time, aging of components, on-site environmental impact, etc., the first electric energy data collected at the current moment may fluctuate or be abnormal. Since the impact of the electric meter abnormality on the second electric energy data is not based on the fixed rules of the preset algorithm compared to the impact on the first electric energy data, the abnormality judgment index of the electric meter cannot be more accurately evaluated by simply comparing the second electric energy data with the first electric energy data and the third electric energy data generated by the preset algorithm. Therefore, in order to obtain the abnormality judgment index of the electric meter more comprehensively and accurately, this embodiment introduces the current environmental feature data of the electric meter.

[0061] S102, converting the environmental characteristic data into a multi-dimensional feature vector, and inputting it into a pre-trained electric energy data prediction model to obtain fourth electric energy data corresponding to the second electric energy data, wherein the fourth electric energy data is used to represent standard (ideal) electric energy data of the electric meter in the normal operating state under the current environmental characteristics, and serves as reference electric energy data for the second electric energy data.

[0062] Specifically, environmental characteristic data include environmental parameters, seasonal attributes, and power load attributes. Environmental parameters include ambient temperature and ambient humidity. The power load attributes are set to the power load level of the time period to which the current time belongs, including high, medium, and low. The time period is divided and set according to the power load conditions at different times based on the usage of various types of electricity meters. For example, a certain electricity meter is of commercial type. According to historical electricity consumption data, it is determined that the power load level from 9 am to 5 pm is higher.

[0063] Specifically, the environmental feature data is converted into a multidimensional feature vector according to the preset conversion rules: represented as [(T, H), a, b], where T is the ambient temperature (degrees Celsius), H is the ambient humidity (percentage), a is the coded value corresponding to the seasonal attribute, a∈[1,2,3,4], spring: a=1, summer: a=2, autumn: a=3, winter: a=4, b is the coded value corresponding to the power load attribute, b∈[1,2,3], low power load level: b=1, medium power load level: b=2, high power load level: b=3.

[0064] In some embodiments, the method for obtaining the pre-trained electric energy data prediction model specifically includes:

[0065] A1. Select an electric meter that is similar to the electric meter to be tested (which can be defined as similar in geographical location, electric meter specifications and electric meter type) as a training electric meter, collect a large amount of environmental feature data of the training electric meter at the time of normal operation in history, convert it into a multidimensional feature vector, and label each multidimensional feature vector to form a training set, and the label content is set to the second electric energy data collected at the corresponding time of the environmental feature data.

[0066] A2. Use the training set to train the pre-set deep learning model, and continuously optimize the model parameters to generate an electric energy data prediction model.

[0067] S103, based on each second electric energy data, calculating the difference between the corresponding third electric energy data and the second electric energy data to obtain a first difference, calculating the difference between the fourth electric energy data and the second electric energy data to obtain a second difference, and obtaining a target difference according to the first difference and the second difference.

[0068] Among them, since the electric meter includes k second electric energy data, which correspond to k third electric energy data and fourth electric energy data respectively, when calculating the first difference and the second difference, the second electric energy data are matched with the third electric energy data and the fourth electric energy data in pairs according to the k types of electric energy data, and the difference between the second electric energy data and the third electric energy data or the fourth electric energy data of the same type of electric energy data is calculated.

[0069] Specifically, obtaining a target difference value according to the first difference value and the second difference value includes:

[0070] The absolute values ​​of the first difference and the second difference are weighted and summed to obtain the target difference. The weight factors of the first difference and the second difference to the target difference are set according to historical experience or actual conditions, and the sum of the two weight factors can be 1.

[0071] For example, ,in, is the target difference, and are the first difference and the second difference respectively, and are the weight factors of the first difference and the second difference to the target difference, .

[0072] S104, within the preset acquisition window, periodically execute steps S101 to S103 for every preset acquisition unit to obtain the target difference of each acquisition unit, and generate a target difference sequence [x1, x2, ..., xn], where n is the number of acquisition units in the acquisition window, and xn is the target difference of the nth acquisition unit.

[0073] S105, based on a preset abnormality detection mechanism, judging whether the electric meter has an abnormality according to the target difference sequence and the deviation index corresponding to each second electric energy data.

[0074] In some embodiments, a method for obtaining a deviation indicator (i=1, 2, ..., k) of the i-th second electric energy data of the electric meter includes:

[0075] B1. Obtain a large amount of second historical electric energy data of the i-th type of second electric energy data collected by the electric meter under normal operating conditions, and refer to steps S101 to S103 to obtain the third historical electric energy data, the fourth historical electric energy data, and the target difference corresponding to each second historical electric energy data.

[0076] B2. Calculate the deviation index of the second electric energy data of the i-th type according to the following formula:

[0077]

[0078] in, is the deviation index of the second electric energy data of the i-th type of electric meter, is the number of second historical electric energy data of the i-th type of second electric energy data, is the target difference corresponding to the j-th historical second electric energy data of the i-th type of second electric energy data, is the difference between the instantaneous electric energy stability value of the j-th historical second electric energy data and its corresponding third historical electric energy data, is the difference between the instantaneous electric energy stability value of the j-th historical second electric energy data and its corresponding fourth historical electric energy data, and They are the influence adjustment coefficient values ​​of the target difference and the instantaneous electric energy stability value difference on the deviation index of the i-th second electric energy data, which are set according to the actual situation.

[0079] Among them, the instantaneous power stability value of the j-th historical second power data is set as: within the collection interval composed of a preset number of collection units before and after the corresponding collection moment of the historical second power data, the standard deviation of the historical second power data corresponding to all collection units in the collection interval is calculated, and the standard deviation is used as the instantaneous power stability value of the j-th historical second power data. Similarly, the instantaneous power stability values ​​of the third historical power data and the fourth historical power data are obtained in the same way, which is not described in detail in the present invention.

[0080] It should be noted that and The value of satisfies and The specific value should be set according to the actual situation. The specific methods may include:

[0081] C1. Select the second historical electric energy data, the third historical electric energy data, and the fourth historical electric energy data corresponding to the i-th type of second electric energy data collected by the electric meter in normal operation and abnormal operation, and obtain several target difference values ​​and their corresponding abnormal results of the electric meter (normal operation and abnormal operation) as the verification data set;

[0082] C2. First, set it randomly and The value of and , get the deviation index ;

[0083] C3, dynamically adjust according to the validation data set and deviation indicators and The value of and The deviation index corresponding to the value of is consistent with the verification results of all verification data sets, and the final deviation index and its corresponding and The value of .

[0084] Among them, until and The deviation index corresponding to the value of is consistent with the verification results of all verification data sets. Specifically, for each verification data, it is compared with the deviation index. If the verification data is less than the deviation index and the verification result is that the meter is normal, or the verification data is greater than the deviation index and the verification result is that the meter is abnormal, then it means that at this time and The value of does not need to be adjusted, and the deviation index corresponding to the i-th second electric energy data of the electric meter can be accurately output. That is, if the target difference currently collected by the electric meter is less than the deviation index, it means that the i-th second electric energy data of the electric meter is currently normal.

[0085] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0086] Based on the second electric energy data, the third electric energy data is introduced through a preset algorithm, and the fourth electric energy data is introduced through environmental feature data, and the second electric energy data is calibrated in a dual path, which comprehensively considers the internal state of the electric meter and the external environmental factors, thereby improving the accuracy of abnormal detection of the electric meter; even if the first electric energy data fluctuates or is abnormal, relatively accurate standard electric energy data can still be obtained through calibration of the second path, thereby reducing misjudgment caused by inaccurate calibration of a single path; the first difference and the second difference are combined to obtain the target difference for comprehensive judgment, thereby more comprehensively reflecting the actual operating state of the electric meter and improving the comprehensiveness of the evaluation;

[0087] The deviation indicators of different second electric energy data types are differentiated. Based on historical data, the difference between the target difference and the instantaneous electric energy stability value is considered, so that the deviation indicators can dynamically adapt to different meter types and usage environments, thereby improving the flexibility and accuracy of anomaly detection.

[0088] Example 2: In Example 1, whether the electric meter is abnormal is determined by comparing the target difference and the deviation index. The traditional abnormality detection method mainly relies on the independent analysis of a single second electric energy data, and fails to fully consider the correlation and importance differences between different second electric energy data. In actual power scenarios, different second electric energy data of the electric meter may be affected by different factors, and their abnormal performances may also be different. Moreover, for the abnormality detection mechanism in Example 1, no specific detection process is given.

[0089] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0090] In some embodiments, in step S105, the preset anomaly detection mechanism specifically includes:

[0091] S201, based on k second electric energy data, according to the respectively corresponding target difference sequences and corresponding deviation indicators, generate an electric energy abnormality distribution array [P1, P2, ..., Pk] of the electric energy meter, wherein Pk is the abnormal probability value of the kth second electric energy data.

[0092] Specifically, for the kth second electric energy data, traverse its target difference sequence [x1, x2, ..., xn], and count the number of target differences greater than the corresponding deviation index, which is recorded as , .

[0093] S202, based on the electric energy abnormality distribution array [P1, P2, ..., Pk] of the electric meter, obtain the abnormality index of the electric meter according to the following first calculation formula:

[0094]

[0095] Wherein, Z is the abnormal index of the electric meter, k is the number of the second electric energy data, is the abnormal probability value of the i-th second electric energy data, The weight value of the abnormal probability value of the i-th second electric energy data for the abnormal judgment of the electric meter is set according to expert experience or historical data verification, satisfying That is, the present invention will not be described in detail.

[0096] S203, based on the abnormality index of the electric meter and the preset abnormality threshold, if the abnormality index is greater than the preset abnormality threshold, the electric meter is determined to be abnormal, and the abnormality handling process is triggered, such as sending maintenance information, recording abnormal time, etc.

[0097] The preset abnormality threshold may be set based on historical data verification and expert experience to define the abnormality index of the electric meter. For example, the preset abnormality threshold may be set to 0.7.

[0098] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0099] By introducing the power anomaly distribution array, an abnormal probability value can be generated for each second power data, so as to more finely evaluate the abnormal state of the meter in different power data dimensions, and improve the accuracy of anomaly detection; not only the abnormal situation of a single power data is considered, but also the overall state of the meter is comprehensively judged by setting different weight values, avoiding the risk of misjudgment of a single data point;

[0100] By calculating the abnormality index of the electric meter and setting different weight values ​​to comprehensively judge the overall status of the electric meter, not only the abnormality of a single electric energy data is taken into account, but also the importance of different electric energy data in abnormality judgment is reflected through weight setting, thereby improving the comprehensiveness and accuracy of the evaluation.

[0101] Embodiment 3: In Embodiment 1 and Embodiment 2, the abnormality of the electric meter is determined by comparing the target difference and the deviation index, and the correlation and importance differences between different electric energy data are taken into account. However, in actual situations, the positive and negative directions of the data difference may have slightly different effects on the determination of the abnormality of different second electric energy data, and the original method cannot reflect this difference (taking the absolute value of the positive and negative differences to obtain the target difference and thus unifying the upper and lower ranges of the deviation index), which may lead to missed detection or false detection. If the positive and negative directions of the data difference are separately refined for training, the workload and training calculation complexity will inevitably increase. Therefore, how to make up for the errors caused by hidden differences becomes a key issue.

[0102] Therefore, the embodiments of the present application are optimized to a certain extent based on the above embodiments.

[0103] In some embodiments, before step S202, the method further includes:

[0104] S301, determining an abnormality detection path according to the distribution of the abnormal power distribution array, the abnormality detection path includes: jumping to step S302 to perform difference correlation analysis, and jumping to step S202.

[0105] Specifically, according to the distribution of the power anomaly distribution array, choose whether to perform difference correlation analysis. If the data distribution indicates that more sophisticated anomaly detection is required, choose to perform difference correlation analysis (jump to S302); otherwise, use the original anomaly detection mechanism (jump to S202).

[0106] S302, based on each type of second electric energy data of the electric meter, collect the first difference and the second difference corresponding to a large amount of second historical electric energy data collected by the electric meter under normal operating conditions, construct a number of difference data points (second difference, first difference), use the second difference as the horizontal coordinate and the first difference as the vertical coordinate, perform fitting, and obtain a difference-related fitting curve graph.

[0107] Specifically, the difference-related fitting curve diagram is used to simulate the approximate dynamic change relationship between the first difference and the second difference of the electric meter under normal operating conditions, and each type of second electric energy data corresponds to a difference-related fitting curve diagram.

[0108] S303, for each second electric energy data, input the second difference corresponding to each target difference in the target difference sequence into the corresponding difference association fitting curve diagram, generate a calibration difference corresponding to the first difference, which is used to reflect the first difference that the second difference should be associated with.

[0109] S304, based on the first difference corresponding to each target difference in the target difference sequence and its corresponding calibration difference, generate a first difference sequence and a calibration difference sequence, take the absolute value of the difference between the calibration difference sequence and the first difference sequence to obtain a deviation difference sequence [d1, d2, ..., dn], average all deviation differences in the deviation difference sequence to obtain a target deviation value.

[0110] S305, based on the target deviation value corresponding to each second electric energy data, obtain the replacement value of the abnormal probability value of each second electric energy data for the weight value of the electric meter abnormality judgment according to the following second calculation formula , replace the first calculation formula , execute step S202 to obtain the abnormality index of the electric meter.

[0111] Among them, the second calculation formula is: , is the target deviation value corresponding to the i-th second electric energy data, is the sum of the target deviation values ​​corresponding to the k second electric energy data, is a replacement value of the abnormal probability value of the i-th second electric energy data for the weight value of the electric meter abnormality judgment.

[0112] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0113] Based on the distribution of the electric energy anomaly distribution array, a suitable anomaly detection method is selected, and the detection method can be flexibly adjusted according to the actual situation; by constructing a difference correlation fitting curve diagram, not only the anomaly of a single electric energy data is considered, but also the correlation between the first difference and the second difference is analyzed, so that the anomaly can be identified more precisely; the target deviation value of the first difference is calculated according to the second difference and the replacement value of the weight value is adjusted accordingly, which can dynamically reflect the importance of different second electric energy data in anomaly judgment and improve the accuracy and flexibility of the evaluation; by introducing difference correlation analysis and target deviation value calculation, anomalies caused by data fluctuations or external environmental factors can be more accurately identified, reducing the risk of misjudgment caused by fluctuations in a single data point.

[0114] Embodiment 4: Further limit and improve the selection of the abnormality detection path in Embodiment 3.

[0115] In some embodiments, step S301 specifically includes:

[0116] S401, inputting the power anomaly distribution array into a pre-trained probability distribution anomaly recognition model, and outputting the distribution of the power anomaly distribution array, the distribution including: normal distribution and abnormal distribution.

[0117] S402, if the distribution is normal, the abnormal detection path is determined as: jump to step S202; if the distribution is abnormal, the abnormal detection path is determined as: jump to step S302.

[0118] Specifically, the method of obtaining the pre-trained probability distribution anomaly recognition model includes:

[0119] D1. Collect a large number of historical power anomaly distribution arrays of the electric meter (the historical power anomaly distribution array can be obtained by analyzing the historical power data of the electric meter) as training data, and label each training data according to the corresponding operating status of the electric meter. If the electric meter is normal, it is labeled as normal distribution, and if the electric meter is abnormal, it is labeled as abnormal distribution.

[0120] It should be noted that it is also possible to extract features from the historical power anomaly distribution array, such as the mean, standard deviation, maximum, minimum, skewness, kurtosis and other statistics of the abnormal probability value, to generate a feature vector for model analysis training, or to directly use the historical power anomaly array as a feature vector for model analysis training, which is not elaborated in the present invention.

[0121] D2. Use the labeled training data to train the pre-set neural network architecture, verify and optimize the model, and obtain a probability distribution anomaly recognition model.

[0122] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0123] By introducing the probability distribution anomaly recognition model, the intelligent judgment of the distribution of the electric energy anomaly distribution array is realized, so that the anomaly detection path is selected according to the model judgment result, that is, whether further refined detection is needed, that is, the method of selecting the weight value of the meter anomaly judgment for the abnormal probability value of different second electric energy data. By analyzing the distribution situation in advance, it is determined whether it is necessary to perform difference correlation analysis, which improves the accuracy of meter anomaly detection and effectively reduces unnecessary redundant work.

[0124] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting abnormality of an electric meter, characterized in that: include: S101, collecting first electric energy data, second electric energy data and environmental characteristic data of the electric energy meter, and obtaining third electric energy data corresponding to the second electric energy data based on a preset algorithm and the first electric energy data; the first electric energy data is electric energy data that can be directly collected and obtained without calculation, the second electric energy data is electric energy data that can be directly collected and can be calculated based on the first electric energy data, the number of second electric energy data is recorded as k, each second electric energy data corresponds to a type of second electric energy data, and the third electric energy data and the fourth electric energy data correspond to them one by one respectively; the preset algorithm is set as a relationship formula between the first electric energy data and the second electric energy data, that is, the first electric energy data is input into the corresponding preset algorithm to obtain the theoretical value of the second electric energy data at the current moment as the third electric energy data corresponding to the second electric energy data; S102, converting the environmental characteristic data into a multidimensional feature vector, and inputting the vector into a pre-trained electric energy data prediction model to obtain fourth electric energy data corresponding to the second electric energy data, where the fourth electric energy data is used to represent standard electric energy data of the electric meter in a normal operating state under the current environmental characteristics; S103, based on each second electric energy data, calculating the difference between the corresponding third electric energy data and the second electric energy data to obtain a first difference, calculating the difference between the fourth electric energy data and the second electric energy data to obtain a second difference, and obtaining a target difference according to the first difference and the second difference; S104, within a preset acquisition window, periodically executing steps S101 to S103 for every preset acquisition unit to obtain a target difference value for each acquisition unit, and generating a target difference value sequence [x1, x2, ..., xn], where n is the number of acquisition units within the acquisition window, and xn is the target difference value of the nth acquisition unit; S105, based on a preset abnormality detection mechanism, judging whether the electric meter has an abnormality according to the target difference sequence and the deviation index corresponding to each second electric energy data; The method for obtaining the deviation index corresponding to the second electric energy data includes: B1. Obtain a large amount of second historical electric energy data of the i-th type of second electric energy data collected by the electric meter under normal operation, and refer to steps S101 to S103 to obtain the third historical electric energy data, the fourth historical electric energy data, and the target difference corresponding to each second historical electric energy data; wherein i=1,2,...,k; B2. Calculate the deviation index of the second electric energy data of the i-th type according to the following formula: ； in, is the deviation index of the second electric energy data of the i-th type of electric meter, is the number of second historical electric energy data of the i-th type of second electric energy data, is the target difference corresponding to the j-th historical second electric energy data of the i-th type of second electric energy data, is the difference between the instantaneous electric energy stability value of the j-th historical second electric energy data and its corresponding third historical electric energy data, is the difference between the instantaneous electric energy stability value of the j-th historical second electric energy data and its corresponding fourth historical electric energy data, and are the adjustment coefficient values ​​of the impact of the target difference and the instantaneous electric energy stability value difference on the deviation index of the i-th second electric energy data.

2. The method for detecting abnormality of an electric meter according to claim 1, characterized in that: The environmental characteristic data includes environmental parameters, seasonal attributes, and power load attributes. The environmental parameters include environmental temperature and environmental humidity. The power load attribute is set to the power load level of the time period to which the current time belongs, including high, medium, and low; The environmental characteristic data is converted into a multidimensional feature vector, which is expressed as: [(T, H), a, b], where T is the ambient temperature, H is the ambient humidity, a is the coded value corresponding to the seasonal attribute, a∈[1,2,3,4], and b is the coded value corresponding to the power load attribute, b∈[1,2,3].

3. The method for detecting abnormality of an electric meter according to claim 1, wherein: Said and The value of satisfies and , the value determination methods include: C1. Select the second historical electric energy data, the third historical electric energy data, and the fourth historical electric energy data corresponding to the i-th type of second electric energy data collected by the electric meter in the normal operating state and the abnormal operating state, respectively, and obtain several target difference values ​​and their corresponding abnormal results of the electric meter as the verification data set, and the abnormal results of the electric meter include the normal operating state and the abnormal operating state; C2. First, set it randomly and The value of and Then, we can get the deviation index ; C3, dynamically adjust according to the validation data set and deviation indicators and The value of and The deviation index corresponding to the value of is consistent with the verification results of all verification data sets, and the final deviation index and its corresponding and The value of .

4. The method for detecting abnormality of an electric meter according to claim 1, wherein: The instantaneous electric energy stability value of the j-th historical second electric energy data is obtained in the following manner: In a collection interval consisting of a preset number of collection units before and after the collection moment corresponding to the historical second electric energy data, the standard deviation of the historical second electric energy data corresponding to all collection units in the collection interval is calculated, and the standard deviation is used as the instantaneous electric energy stability value of the jth historical second electric energy data.

5. The method for detecting abnormality of an electric meter according to claim 1, wherein: The S105 specifically includes: S201, based on k second electric energy data, according to the corresponding target difference sequences and corresponding deviation indicators, generate an electric energy abnormality distribution array [P1, P2, ..., Pk] of the electric meter, where Pk is the abnormal probability value of the kth second electric energy data, , The acquisition method is as follows: for the kth second electric energy data, traverse its target difference sequence [x1, x2,..., xn], and count the number of target differences greater than the corresponding deviation index, recorded as ; S202, based on the electric energy abnormality distribution array [P1, P2, ..., Pk] of the electric meter, obtain the abnormality index of the electric meter according to the following first calculation formula: ; Wherein, Z is the abnormal index of the electric meter, k is the number of the second electric energy data, is the abnormal probability value of the i-th second electric energy data, The weight value of the abnormal probability value of the i-th second electric energy data for the abnormal judgment of the electric meter is set according to expert experience or historical data verification, satisfying ; S203, based on the abnormality index of the electric meter and the preset abnormality threshold, if the abnormality index is greater than the preset abnormality threshold, the electric meter is determined to be abnormal, triggering the abnormality handling process.

6. The method for detecting abnormality of an electric meter according to claim 5, characterized in that: Before S202, the method further includes: S301, determining an abnormality detection path according to the distribution of the abnormal power distribution array, the abnormality detection path comprising: jumping to step S302 to perform difference correlation analysis, and jumping to step S202; S302, based on each type of second electric energy data of the electric meter, collecting first differences and second differences corresponding to a large amount of second historical electric energy data collected by the electric meter under normal operating conditions, taking the second differences as the horizontal coordinates and the first differences as the vertical coordinates, constructing a number of difference data points, performing fitting, and obtaining a difference correlation fitting curve graph; S303, for each second electric energy data, inputting the second difference corresponding to each target difference in the target difference sequence into the corresponding difference association fitting curve diagram, and generating a calibration difference corresponding to the first difference; S304, based on the first difference corresponding to each target difference in the target difference sequence and the calibration difference corresponding to each target difference, a first difference sequence and a calibration difference sequence are generated, the calibration difference sequence is subtracted from the first difference sequence and an absolute value is taken to obtain a deviation difference sequence [d1, d2, ..., dn], and all deviation differences in the deviation difference sequence are averaged to obtain a target deviation value; S305, based on the target deviation value corresponding to each second electric energy data, obtaining a replacement value of the abnormal probability value of each second electric energy data for the weight value of the electric meter abnormality judgment according to the second calculation formula , replace the first calculation formula , execute step S202 to obtain the abnormality index of the electric meter.

7. The method for detecting abnormality of an electric meter according to claim 6, characterized in that: The second calculation formula is: ; in, is the target deviation value corresponding to the i-th second electric energy data, is the sum of the target deviation values ​​corresponding to the k second electric energy data, is a replacement value of the abnormal probability value of the i-th second electric energy data for the weight value of the electric meter abnormality judgment.

8. The method for detecting abnormality of an electric meter according to claim 6, wherein: The S301 specifically includes: S401, inputting the power abnormal distribution array into a pre-trained probability distribution abnormality recognition model, and outputting the distribution of the power abnormal distribution array, the distribution including: normal distribution and abnormal distribution; S402, if the distribution is normal, the abnormal detection path is determined as: jump to step S202; if the distribution is abnormal, the abnormal detection path is determined as: jump to step S302.

Images