CN112213687A - Gateway electric energy meter data anomaly detection method and system based on pseudo anomaly point identification - Google Patents
Gateway electric energy meter data anomaly detection method and system based on pseudo anomaly point identification Download PDFInfo
- Publication number
- CN112213687A CN112213687A CN202011079730.7A CN202011079730A CN112213687A CN 112213687 A CN112213687 A CN 112213687A CN 202011079730 A CN202011079730 A CN 202011079730A CN 112213687 A CN112213687 A CN 112213687A
- Authority
- CN
- China
- Prior art keywords
- abnormal
- anomaly
- data
- electric energy
- random
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R35/00—Testing or calibrating of apparatus covered by the other groups of this subclass
- G01R35/04—Testing or calibrating of apparatus covered by the other groups of this subclass of instruments for measuring time integral of power or current
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a gateway electric energy meter data anomaly detection method and system based on pseudo anomaly point identification, which divide data anomaly modes into two types of random anomaly points and continuous anomaly points: for the random abnormal point mode, according to the characteristic that the power load data has regularity, the detection of the random abnormal points is realized by carrying out time series decomposition and autocorrelation analysis on a data curve, and abnormal detection interference data caused by periodic power utilization of a user is avoided; for the continuous abnormal point mode, according to the abnormal electricity utilization characteristics of the power users, two characteristics of line loss rate and continuous descending trend of electric quantity are extracted based on historical data, a user electricity stealing behavior detection model is trained, abnormal detection interference data caused by the abnormal electricity utilization behaviors of the users are avoided, and then a gateway electric energy meter abnormal detection model is built so as to realize automatic identification of pseudo abnormal data. The method can avoid the interference data caused by the dynamic behavior of the user, further accurately identify the abnormal value and improve the data quality.
Description
Technical Field
The invention relates to the technical field of power equipment detection, in particular to a gateway electric energy meter data abnormity detection method and system based on pseudo abnormity point identification.
Background
At present, the research on abnormal measurement data of the gateway electric energy meter mostly focuses on the aspects of abnormal reason analysis and on-line monitoring system design, and from the quality of the collected data, the research on the value of the data is less. The metering data of the gateway electric energy meter is typical one-dimensional time sequence data, abnormal value detection can be carried out through the traditional 3 sigma criterion, box line diagram and other methods, but the single traditional method is high in detection false detection rate and poor in applicability due to the fact that electricity data have the characteristics of obvious volatility and regularity.
Under ideal working conditions, the data of each monitoring index of the gateway electric energy meter should be stabilized within a reasonable range. However, under actual working conditions, due to accidental factors such as environmental factors and mechanical faults, the data of the gateway table often has some quality problems. Meanwhile, the dynamic behavior of the user can also generate data points with similar expression forms to the former, such as abnormal electricity utilization behavior and periodic electricity utilization behavior. The expression form of the two cases in the metering data is an abnormal point which does not accord with the rule of the historical data. However, the mechanism of the generation of the two abnormal points is different, and the intrinsic mining value is also different. The abnormal point caused by the dynamic behavior of the user can reflect the load demand or the abnormal electricity utilization condition of the load, has subsequent mining significance, and is defined as a pseudo abnormal point; on the contrary, the abnormal points generated by environmental factors, mechanical problems and the like have accidental factors, and the subsequent excavation value is small.
Disclosure of Invention
The invention aims to solve the technical problems that the detection false detection rate of the traditional method for researching the abnormal metering data of the gateway electric energy meter is high, the applicability is poor, and aims to provide a method and a system for detecting the abnormal data of the gateway electric energy meter based on the identification of a false abnormal point, so that the problems that the interference data caused by the dynamic behavior of a user can be avoided, the abnormal value can be accurately identified, and the data quality is improved are solved.
The invention is realized by the following technical scheme:
the gateway electric energy meter data anomaly detection method based on the pseudo anomaly point identification comprises the following steps: judging the type of an abnormal data set based on the abnormal data set collected by a gateway electric energy meter, wherein the type comprises the following steps: a random anomaly mode and a continuous anomaly mode; a: anomaly data set for random anomaly patterns: processing the load curve of the abnormal data set in the random abnormal mode through an STL time series decomposition algorithm to obtain a trend term, a period term and a residual term of the load curve, and the method comprises the following substeps: step A1: performing de-trending processing on the load curve, extracting a period item and a residual item of the load curve, performing Fourier transform on the de-trended load curve to a frequency domain, and obtaining all candidate periods K according to peak intervals; step A2: carrying out autocorrelation analysis on the de-trended load curve to obtain an autocorrelation function value P corresponding to the candidate period K, respectively eliminating potential abnormal values and carrying out iterative computation to obtain a maximum autocorrelation function value rho max, wherein a period ki corresponding to the maximum autocorrelation function value rho max is a real period containing the periodic power utilization behavior of the user, and therefore random abnormal points are screened out; wherein, K is { K1, K2 …, kn }, P is { ρ 1, ρ 2 …, ρ n }, i is ∈ {1, 2, …, n }, and n is a natural number; b: anomaly data set for continuous anomaly patterns: based on a historical data set, forming an expert sample by extracting the line loss rate and the continuous descending trend of the electric quantity, and sending the expert sample into a classifier for training to form a user electricity stealing behavior detection model; and sending the abnormal data set of the continuous abnormal mode into the user electricity stealing behavior detection model, thereby screening out continuous abnormal points.
The factors such as external environment, mechanical failure and the like and the dynamic electricity utilization behavior of the user can cause the abnormal metering data, but the generation mechanism is different, and the inherent mining value is also different. The abnormal point caused by the user electricity consumption behavior is similar to the abnormal point of the device fault in expression form, and is easy to be detected by mistake, so the abnormal point is defined as a false abnormal point. According to the invention, based on the metering data quality of the gateway electric energy meter, dynamic electricity utilization behaviors of users are mined, and data points which are shown as abnormal in the metering data and are caused by the dynamic electricity utilization behaviors are avoided. Data anomaly patterns are divided into two categories, namely random anomaly points and continuous anomaly points: for the random abnormal point mode, according to the characteristic that the power load data has regularity, the detection of the random abnormal points is realized by carrying out time series decomposition and autocorrelation analysis on a data curve, and abnormal detection interference data caused by periodic power utilization of a user is avoided; for the continuous abnormal point mode, two characteristics of line loss rate and continuous descending trend of electric quantity are extracted based on historical data according to the abnormal electricity utilization characteristics of the power users, a user electricity stealing behavior detection model is trained, and abnormal detection interference data caused by the abnormal electricity utilization behaviors of the users are avoided. And then a gateway electric energy meter abnormity detection model is built so as to realize automatic identification of false abnormal data and improve the economy and accuracy of metering.
Further, the algorithm used by the classifier comprises a decision tree algorithm and an SVM algorithm.
Further, the user electricity stealing behavior detection model is evaluated through an AUC index in an ROC curve.
Further, screening all abnormal values according to the abnormal data set, and judging the abnormal data set to be a random abnormal mode or a continuous abnormal mode according to the mapping relation between the abnormal values and the time stamps.
Further, all random outliers are screened out by a 3 sigma criterion method or a box plot method.
Further, the load curve is detrended by an STL time-series decomposition algorithm, which is as follows:
Yv＝f(Tv,Sv,Rv)v＝1,…,N；
in the formula, TvIs a trend item, representing long-term characteristics of the data; svIs a periodic item, representing a periodic characteristic of the data, RvIs a residual term, representing the uncertainty characteristics of the data; n is a natural number.
The invention also discloses a gateway electric energy meter data abnormity detection system based on the identification of the false abnormity point, which comprises the following steps: an acquisition module: the gateway electric energy meter is used for acquiring an abnormal data set from the gateway electric energy meter and judging whether the type of the abnormal data set is a random abnormal mode or a continuous abnormal mode; a random outlier screening module: the load curve processing method comprises the steps of processing a load curve of an abnormal data set in the random abnormal mode through an STL time series decomposition algorithm, performing detrending processing on the load curve, extracting a period item and a residual error item of the load curve, performing Fourier transform on the detrended load curve to a frequency domain, and obtaining all candidate periods K according to peak intervals; carrying out autocorrelation analysis on the de-trended load curve to obtain an autocorrelation function value P corresponding to the candidate period K, respectively eliminating potential abnormal values and carrying out iterative computation to obtain a maximum autocorrelation function value rho max, wherein a period ki corresponding to the maximum autocorrelation function value rho max is a real period containing the periodic power utilization behavior of the user, and therefore random abnormal points are screened out; wherein, K is { K1, K2 …, kn }, P is { ρ 1, ρ 2 …, ρ n }, i is ∈ {1, 2, …, n }, and n is a natural number; a continuous abnormal point screening module: the system is used for forming an expert sample by extracting the line loss rate and the continuous descending trend of the electric quantity based on a historical data set, and sending the expert sample into a classifier for training to form a user electricity stealing behavior detection model; sending the abnormal data set of the continuous abnormal mode into the user electricity stealing behavior detection model so as to screen out continuous abnormal points; an output module: and the random abnormal point and the continuous abnormal point are output and displayed.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the random abnormal point detection can avoid the misjudgment of abnormal values caused by the periodic power utilization behavior of the user, the continuous abnormal point detection can accurately identify the power stealing behavior of the user, and the abnormal point detection interference caused by the abnormal power utilization mode of the user is avoided. And then a gateway electric energy meter abnormity detection model is built, so that the metering accuracy and economy are improved, and the data quality is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention. In the drawings:
FIG. 1 is a schematic view of the process of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to examples and accompanying drawings, and the exemplary embodiments and descriptions thereof are only used for explaining the present invention and are not meant to limit the present invention.
Interpretation of terms:
1. the gateway electric energy meter is an electric energy meter installed in gateway electric energy metering devices such as a power generation enterprise internet access device, a cross-regional connecting line, a provincial network connecting line, a provincial intranet and the like, is used for checking trade settlement and internal economic indexes, and is a meter holding a gateway, commonly called a master meter;
2. the abnormal metering data of the gateway electric energy meter refers to the phenomenon that the metering data obviously deviates from the value of a data curve due to environmental factors, human factors, mechanical faults and the like in the actual working condition;
3. the pseudo abnormal point refers to a data point which is caused by the electricity consumption behavior of the power consumer and is expressed as abnormal in the metering data, and the data point comprises modes such as periodic electricity consumption behavior and abnormal electricity consumption behavior, and the expression form of the data point is similar to that of the abnormal point caused by factors such as device faults;
4. the construction of the abnormal data identification method is mainly to classify abnormal modes according to the data abnormal expression form, including the detection of random abnormal points and continuous abnormal points. The periodic power consumption behaviors of the users are mined through time sequence decomposition and autocorrelation analysis to realize the identification of random abnormal points, and two indexes of line loss rate and continuous descending trend of electric quantity are extracted and sent to a classifier to be trained to realize the detection of continuous abnormal points represented by the abnormal power consumption behaviors.
Example 1
The factors such as external environment, mechanical failure and the like and the dynamic electricity utilization behavior of the user can cause the abnormal metering data, but the generation mechanism is different, and the inherent mining value is also different. The abnormal point caused by the user electricity consumption behavior is similar to the abnormal point of the device fault in expression form, and is easy to be detected by mistake, so the abnormal point is defined as a false abnormal point. In this embodiment 1, starting from the quality of metering data of a gateway electric energy meter, a method for detecting data abnormality of the gateway electric energy meter based on pseudo-abnormal point identification is provided, so as to mine dynamic power consumption behaviors of a user and avoid data points which are represented as abnormalities in the metering data due to the dynamic power consumption behaviors.
Data anomaly patterns are divided into two categories, namely random anomaly points and continuous anomaly points: for the random abnormal point mode, according to the characteristic that the power load data has regularity, the detection of the random abnormal points is realized by carrying out time series decomposition and autocorrelation analysis on a data curve, and abnormal detection interference data caused by periodic power utilization of a user is avoided; for the continuous abnormal point mode, two characteristics of line loss rate and continuous descending trend of electric quantity are extracted based on historical data according to the abnormal electricity utilization characteristics of the power users, a user electricity stealing behavior detection model is trained, and abnormal detection interference data caused by the abnormal electricity utilization behaviors of the users are avoided. And then a gateway electric energy meter abnormity detection model is built so as to realize automatic identification of false abnormal data and improve the economy and accuracy of metering.
Embodiment 1 is a gateway electric energy meter data anomaly detection method based on pseudo anomaly point identification, which specifically includes the following steps:
the method comprises the following steps: based on an abnormal data set acquired by the gateway electric energy meter, the abnormal modes are firstly distinguished. And preliminarily screening all abnormal values by using a traditional 3 sigma criterion, a box diagram and other methods. And classifying the abnormal modes into random abnormal modes and continuous abnormal modes according to the mapping relation between the abnormal values and the time stamps.
Step two: performing detrending on a data set with random abnormal points through an STL time series decomposition algorithm, extracting a period item and a residual item of a data curve, performing Fourier transform on a detrended load curve to a frequency domain, and obtaining all candidate periods k according to peak intervals1,k2…,kn。
Step three: carrying out autocorrelation analysis on the data curve after trend removal to obtain a candidate period k1,k2…,knCorresponding autocorrelation function value rho1,ρ2…,ρn. Obtaining the maximum autocorrelation function value rho by respectively eliminating the potential abnormal values and performing iterative computationmaxCorresponding to period kiThe random abnormal points are screened out according to the real period containing the periodic power utilization behavior of the user.
Step four: and (4) inspecting the electricity stealing behavior of the user aiming at the continuous abnormal mode. Based on a historical data set, an expert sample is formed by extracting two monitoring indexes of a line loss rate and a continuous descending trend of electric quantity, a decision tree algorithm and an SVM algorithm are selected to train a classifier through supervised learning, and the test set accounts for 20%. And evaluating the quality of the detection model through AUC indexes in the ROC curve, and finding that the detection effect of the decision tree algorithm is superior to that of the SVM algorithm.
By the method for mining the dynamic behavior of the user, 1) abnormal value misjudgment caused by periodic power utilization behavior of the user can be avoided by detecting random abnormal points, 2) accurate identification of power stealing of the user can be realized by detecting continuous abnormal points, and abnormal point detection interference caused by abnormal power utilization patterns of the user can be avoided. And then a gateway electric energy meter abnormity detection model is built, so that the metering accuracy and economy are improved, and the data quality is improved.
Example 2
The embodiment 2 is further developed on the basis of the embodiment 1, and as shown in fig. 1, specifically includes:
STL time series decomposition based on power load regularity
The power load change has obvious regularity, the regularity contains dynamic behaviors of users, the fluctuations include fluctuations caused by natural season influence and fluctuations caused by working time laws and the like, the fluctuation length (season, month, week and the like) is fixed, and the specific change form depends on the characteristics of different types of loads. Therefore, a method of local-Trend decomposition product based on local Weighted Regression (LOESS) time series decomposition is adopted to decompose the electric load data measured by the gateway electric energy meter, and the load data is assumed to be composed of a Trend term, a period term and a residual term, and is shown in formula (1).
Yv＝f(Tv,Sv,Rv) v＝1,…,N (1)
In the formula, TvIs a trend item, representing long-term characteristics of the data; svIs a periodic item, representing a periodic characteristic of the data, RvAre residual terms that characterize the uncertainty of the data.
The STL algorithm consists of two loop mechanisms, an inner loop nested within an outer loop. The inner circulation mainly performs trend fitting and periodic component calculation, and T is set at the beginningv (0)0. The internal circulation is first worked to detrended, i.e. Yv-Tv(k) (ii) a Second, periodic subsequence smoothing, each length n using Loess(p)The subsequence of (A) is regressed to obtain a time sequence Cv (k+1)(ii) a Then the result sequence Cv (k+1)Low-pass filtering and one Loess smoothing treatment can extract the low-pass quantity L of the periodic subsequencev (k+1)(ii) a Finally, the smooth subsequence is subjected to detrending to obtain a periodic item Sv (k+1)And the original data period removing item is subjected to Loess smoothing to obtain a trend component Tv (k +1)The specific calculation formula is as follows:
the outer loop is mainly used for adjusting the robustness of the weight, and defines an expression (4):
h＝6×median(|Rv|) (4)
for a data point with a position v, the weight robustness is as follows:
βv＝B(|Rv|/h) (5)
wherein the B function is a bisquare function:
aiming at the characteristic that the power load curve has trend change, a multiplication model is adopted to decompose the power load curve, and the formula is as follows:
Yv＝Tv×Sv×Rv (7)
and obtaining a trend item, a period item and a residual error item of the load curve through an STL time series decomposition algorithm, and then performing detrending on the trend item, the period item and the residual error item, so as to better analyze the period item and mine an internal periodic rule.
Secondly, autocorrelation analysis of the de-trending load curve:
after the STL time series decomposition in the step 1), a plurality of candidate periods including a real period containing the periodic electricity utilization behavior of the user and a false period caused by the interference of an abnormal point can be obtained. The goal here is to screen out the candidate cycles for real cycles that have the characteristics of the curve itself. The self-correlation function (ACF) is adopted to calculate the correlation degree of the time series data before and after a certain delay tau and reflect the self-correlation of the data, and the calculation formula is as follows:
wherein Z is a load data set,
Ztand/n is the sample mean of the sequence.
Autocorrelation function ρkRelated only to time interval k, independent of time t, and | ρkLess than or equal to 1. Thus a fixed real period time interval k1Corresponding to rho1Time interval k of a dummy period2,k3…,knRespectively correspond to rho2,ρ3…,ρn. Because the generation of the false period is interference caused by random abnormal points, the periodic performance is poor, namely the rho value is small; whereas the value of p for the real period is larger. The method is converted into an iterative optimization problem by successively deleting potential abnormal values to obtain the maximum value rho of the autocorrelation functionmaxAnd then a real period is obtained, and the identification of abnormal points is realized.
Thirdly, selecting the characteristics of the electricity stealing behavior of the user:
based on field historical data experience, two indexes of line loss rate and electric quantity descending trend are selected as characteristics for detecting whether a user has electricity stealing behavior in an abnormal mode. The calculation formulas of the two are as follows:
wherein P is the amount of power supply, PiThe electricity consumption of the ith electric energy meter of the transformer area is measured;
tdelectric quantity trend at day d, PdThe used amount on day d.
Because electricity stealing is a long-term event, attention is paid to the line loss rate exceeding condition and the continuous descending trend of the electric quantity in a period of time, and the calculation formulas of the line loss rate exceeding condition and the electric quantity continuously descending trend are as follows:
wherein k is a line loss rate warning threshold value, T1The number of exceeding days when the line loss rate is within 10 continuously; t is2The total number of days with a decreasing trend over 10 consecutive days.
Fourth, evaluation index
Respectively training a user electricity stealing behavior detection model by using a decision tree algorithm and an SVM algorithm, evaluating the performance of the model by using a confusion matrix and an ROC curve, wherein the confusion matrix applied in the user electricity stealing behavior detection is shown in the following table:
| user' s | Detecting as electricity stealing users | Detect as a normal user |
| Actual electricity stealing subscriber | TP(true positive) | FN(false negative) |
| Actual normal user | FP(false positive) | TN(true negative) |
Here, a hit rate (TPR) and a False Positive Rate (FPR) are defined:
wherein, the value interval of TPR and FPR is [0,1], the closer the TPR is to 1, the closer the FPR is to 0, the better the detection effect is.
The ROC (receiver operating characteristic) curve has the function of describing the relative relationship between the increasing rates of the two indexes of the FPR and the TPR in the confusion matrix. In the ROC space coordinates, point (0,1) represents an ideal classifier, and the proximity of the ROC curve to point (0,1) represents the better the classification. The area under the curve (AUC) represents the quality of the classifier by a value, the value of AUC is the size of the area under the ROC curve, and the larger the value of AUC, the higher the precision of the classifier is.
Example 3
This embodiment 3 is a gateway electric energy meter data anomaly detection system based on pseudo anomaly point identification, including the following modules:
an acquisition module: the gateway electric energy meter is used for acquiring an abnormal data set from the gateway electric energy meter and judging whether the type of the abnormal data set is a random abnormal mode or a continuous abnormal mode;
a random outlier screening module: the load curve processing method comprises the steps of processing a load curve of an abnormal data set in a random abnormal mode through an STL time series decomposition algorithm, performing detrending processing on the load curve, extracting a period item and a residual item of the load curve, performing Fourier transform on the detrended load curve to a frequency domain, and obtaining all candidate periods K according to peak value intervals; carrying out autocorrelation analysis on the de-trended load curve to obtain an autocorrelation function value P corresponding to a candidate period K, respectively eliminating potential abnormal values to carry out iterative computation to obtain a maximum autocorrelation function value rho max, wherein a period ki corresponding to the maximum autocorrelation function value rho max is a real period containing the periodic power utilization behavior of the user, and therefore random abnormal points are screened out; wherein, K is { K1, K2 …, kn }, P is { ρ 1, ρ 2 …, ρ n }, i is ∈ {1, 2, …, n }, and n is a natural number;
a continuous abnormal point screening module: based on a historical data set, forming an expert sample by extracting the line loss rate and the continuous descending trend of the electric quantity, and sending the expert sample into a classifier for training to form a user electricity stealing behavior detection model; sending the abnormal data set of the continuous abnormal mode into a user electricity stealing behavior detection model so as to screen out continuous abnormal points;
an output module: for output display of random outliers and continuous outliers.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (7)
1. A gateway electric energy meter data anomaly detection method based on pseudo anomaly point identification is characterized by comprising the following steps:
judging the type of an abnormal data set based on the abnormal data set collected by a gateway electric energy meter, wherein the type comprises the following steps: a random anomaly mode and a continuous anomaly mode;
a: anomaly data set for random anomaly patterns:
processing the load curve of the abnormal data set in the random abnormal mode through an STL time series decomposition algorithm to obtain a trend term, a period term and a residual term of the load curve, and the method comprises the following substeps:
step A1: performing de-trending processing on the load curve, extracting a period item and a residual item of the load curve, performing Fourier transform on the de-trended load curve to a frequency domain, and obtaining all candidate periods K according to peak intervals;
step A2: carrying out autocorrelation analysis on the de-trended load curve to obtain autocorrelation function values rho corresponding to the candidate periods K, and respectively eliminating potential abnormal values to carry out iterative computation to obtain maximum autocorrelation function values rhomaxThe maximum autocorrelation function value ρmaxCorresponding period kiThe real period containing the periodic electricity utilization behavior of the user is obtained, and therefore random abnormal points are screened out;
wherein K ═ { K ═ K1,k2…,kn},ρ＝{ρ1,ρ2…,ρnThe i belongs to {1, 2, …, n }, and n is a natural number;
b: anomaly data set for continuous anomaly patterns:
based on a historical data set, forming an expert sample by extracting the line loss rate and the continuous descending trend of the electric quantity, and sending the expert sample into a classifier for training to form a user electricity stealing behavior detection model; and sending the abnormal data set of the continuous abnormal mode into the user electricity stealing behavior detection model, thereby screening out continuous abnormal points.
2. The gateway electric energy meter data anomaly detection method based on the pseudo-anomaly point identification is characterized in that algorithms used by the classifier comprise a decision tree algorithm and an SVM algorithm.
3. The gateway electric energy meter data anomaly detection method based on the pseudo anomaly point identification according to claim 2, wherein the user electricity stealing behavior detection model is evaluated through an AUC index in an ROC curve.
4. The gateway electric energy meter data abnormality detection method based on the pseudo-abnormal point identification as claimed in claim 1, wherein all abnormal values are screened out according to the abnormal data set, and the abnormal data set is judged to be a random abnormal mode or a continuous abnormal mode according to a mapping relation between the abnormal values and a time stamp.
5. The gateway electric energy meter data anomaly detection method based on the pseudo-anomaly point identification is characterized in that all random anomaly values are screened out through a 3 sigma criterion method or a box plot method.
6. The gateway electric energy meter data anomaly detection method based on pseudo anomaly point identification according to claim 1, characterized in that the load curve is detrended by an STL time series decomposition algorithm, which is as follows:
Yv＝f(Tv,Sv,Rv)v＝1,…,N；
in the formula, TvIs a trend item, representing long-term characteristics of the data; svIs a periodic item, representing a periodic characteristic of the data, RvIs a residual term, representing the uncertainty characteristics of the data; n is a natural number.
7. Pass electric energy meter data anomaly detection system based on pseudo anomaly point is discerned, its characterized in that includes:
an acquisition module: the gateway electric energy meter is used for acquiring an abnormal data set from the gateway electric energy meter and judging whether the type of the abnormal data set is a random abnormal mode or a continuous abnormal mode;
a random outlier screening module: the load curve processing method comprises the steps of processing a load curve of an abnormal data set in the random abnormal mode through an STL time series decomposition algorithm, performing detrending processing on the load curve, extracting a period item and a residual error item of the load curve, performing Fourier transform on the detrended load curve to a frequency domain, and obtaining all candidate periods K according to peak intervals; carrying out autocorrelation analysis on the de-trended load curve to obtain an autocorrelation function value P corresponding to the candidate period K, respectively eliminating potential abnormal values and carrying out iterative computation to obtain a maximum autocorrelation function value rho max, wherein a period ki corresponding to the maximum autocorrelation function value rho max is a real period containing the periodic power utilization behavior of the user, and therefore random abnormal points are screened out; wherein, K is { K1, K2 …, kn }, P is { ρ 1, ρ 2 …, ρ n }, i is ∈ {1, 2, …, n }, and n is a natural number;
a continuous abnormal point screening module: the system is used for forming an expert sample by extracting the line loss rate and the continuous descending trend of the electric quantity based on a historical data set, and sending the expert sample into a classifier for training to form a user electricity stealing behavior detection model; sending the abnormal data set of the continuous abnormal mode into the user electricity stealing behavior detection model so as to screen out continuous abnormal points;
an output module: and the random abnormal point and the continuous abnormal point are output and displayed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011079730.7A CN112213687B (en) | 2020-10-10 | 2020-10-10 | Gateway electric energy meter data anomaly detection method and system based on pseudo-anomaly point identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011079730.7A CN112213687B (en) | 2020-10-10 | 2020-10-10 | Gateway electric energy meter data anomaly detection method and system based on pseudo-anomaly point identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112213687A true CN112213687A (en) | 2021-01-12 |
| CN112213687B CN112213687B (en) | 2023-05-16 |
Family
ID=74053126
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011079730.7A Active CN112213687B (en) | 2020-10-10 | 2020-10-10 | Gateway electric energy meter data anomaly detection method and system based on pseudo-anomaly point identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112213687B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113704321A (en) * | 2021-08-11 | 2021-11-26 | 国电南瑞科技股份有限公司 | Power grid abnormal data identification method, device and system |
| CN114039886A (en) * | 2021-10-19 | 2022-02-11 | 国网山东省电力公司日照供电公司 | Data correctness identification method and device based on four-differential and autocorrelation analysis |
| CN117421610A (en) * | 2023-12-19 | 2024-01-19 | 山东德源电力科技股份有限公司 | Data anomaly analysis method for electric energy meter running state early warning |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014155385A (en) * | 2013-02-12 | 2014-08-25 | Chugoku Electric Power Co Inc:The | Electric conduction confirmation system using smart meter for electricity |
| CN104280714A (en) * | 2014-10-17 | 2015-01-14 | 国家电网公司 | Running abnormality detecting method of metering device |
| CN104484371A (en) * | 2014-12-05 | 2015-04-01 | 广州供电局有限公司 | Method and system for monitoring and analyzing electric power marketing exceptional data in online manner |
| CN106645934A (en) * | 2016-12-12 | 2017-05-10 | 国网北京市电力公司 | Power utilization behavior electricity theft prevention diagnosis method and device based on dynamic grid outliers |
| CN107423746A (en) * | 2017-04-06 | 2017-12-01 | 广东电网有限责任公司云浮供电局 | Dedicated for electric line communication electricity data analysis method and system based on cluster analysis |
| CN107622342A (en) * | 2017-08-31 | 2018-01-23 | 国网辽宁省电力有限公司电力科学研究院 | A kind of distributing net platform region data analysis system based on MVC frameworks |
| CN107944464A (en) * | 2017-10-12 | 2018-04-20 | 华南理工大学 | A kind of office building by when energy consumption abnormal data online recognition and complementing method |
| CN108416693A (en) * | 2018-02-01 | 2018-08-17 | 东华大学 | A kind of multiplexing electric abnormality Activity recognition method based on fuzzy neural network |
| CN110531305A (en) * | 2019-08-12 | 2019-12-03 | 广西电网有限责任公司 | A kind of electricity exception monitoring system and its platform and method |
| CN110634080A (en) * | 2018-06-25 | 2019-12-31 | 中兴通讯股份有限公司 | Abnormal electricity utilization detection method, device, equipment and computer readable storage medium |
| CN110675018A (en) * | 2019-08-13 | 2020-01-10 | 南方电网科学研究院有限责任公司 | Line loss anomaly analysis and closed-loop management method based on expert rules |
| CN110763886A (en) * | 2019-12-05 | 2020-02-07 | 国网江苏省电力有限公司南通供电分公司 | Single-phase user electricity stealing judgment and positioning method |
| US20200116766A1 (en) * | 2015-12-11 | 2020-04-16 | Schneider Electric USA, Inc. | Systems and methods for monitoring a power system |
| CN111190045A (en) * | 2019-12-27 | 2020-05-22 | 国网北京市电力公司 | Voltage abnormity prediction method and device and electronic equipment |
| US20200174437A1 (en) * | 2018-11-29 | 2020-06-04 | Industrial Technology Research Institute | Energy consumption management system and energy consumption management method |
| CN111525697A (en) * | 2020-05-09 | 2020-08-11 | 西安交通大学 | Medium and low voltage power distribution network electricity larceny prevention method and system based on current monitoring and line topology analysis |
-
2020
- 2020-10-10 CN CN202011079730.7A patent/CN112213687B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014155385A (en) * | 2013-02-12 | 2014-08-25 | Chugoku Electric Power Co Inc:The | Electric conduction confirmation system using smart meter for electricity |
| CN104280714A (en) * | 2014-10-17 | 2015-01-14 | 国家电网公司 | Running abnormality detecting method of metering device |
| CN104484371A (en) * | 2014-12-05 | 2015-04-01 | 广州供电局有限公司 | Method and system for monitoring and analyzing electric power marketing exceptional data in online manner |
| US20200116766A1 (en) * | 2015-12-11 | 2020-04-16 | Schneider Electric USA, Inc. | Systems and methods for monitoring a power system |
| CN106645934A (en) * | 2016-12-12 | 2017-05-10 | 国网北京市电力公司 | Power utilization behavior electricity theft prevention diagnosis method and device based on dynamic grid outliers |
| CN107423746A (en) * | 2017-04-06 | 2017-12-01 | 广东电网有限责任公司云浮供电局 | Dedicated for electric line communication electricity data analysis method and system based on cluster analysis |
| CN107622342A (en) * | 2017-08-31 | 2018-01-23 | 国网辽宁省电力有限公司电力科学研究院 | A kind of distributing net platform region data analysis system based on MVC frameworks |
| CN107944464A (en) * | 2017-10-12 | 2018-04-20 | 华南理工大学 | A kind of office building by when energy consumption abnormal data online recognition and complementing method |
| CN108416693A (en) * | 2018-02-01 | 2018-08-17 | 东华大学 | A kind of multiplexing electric abnormality Activity recognition method based on fuzzy neural network |
| CN110634080A (en) * | 2018-06-25 | 2019-12-31 | 中兴通讯股份有限公司 | Abnormal electricity utilization detection method, device, equipment and computer readable storage medium |
| US20200174437A1 (en) * | 2018-11-29 | 2020-06-04 | Industrial Technology Research Institute | Energy consumption management system and energy consumption management method |
| CN110531305A (en) * | 2019-08-12 | 2019-12-03 | 广西电网有限责任公司 | A kind of electricity exception monitoring system and its platform and method |
| CN110675018A (en) * | 2019-08-13 | 2020-01-10 | 南方电网科学研究院有限责任公司 | Line loss anomaly analysis and closed-loop management method based on expert rules |
| CN110763886A (en) * | 2019-12-05 | 2020-02-07 | 国网江苏省电力有限公司南通供电分公司 | Single-phase user electricity stealing judgment and positioning method |
| CN111190045A (en) * | 2019-12-27 | 2020-05-22 | 国网北京市电力公司 | Voltage abnormity prediction method and device and electronic equipment |
| CN111525697A (en) * | 2020-05-09 | 2020-08-11 | 西安交通大学 | Medium and low voltage power distribution network electricity larceny prevention method and system based on current monitoring and line topology analysis |
Non-Patent Citations (4)
| Title |
|---|
| JIA WANG: "Detection of Abnormal Data of Gateway Energy Meter Based on User Dynamic Behavior Mining" * |
| 吕炳: "基于云计算的电能计量数据异常分析方法" * |
| 周金萍: "异常用电用户检测方法研究" * |
| 孙一浩: "基于伪异常点辨识的关口电能表计量数据异常研究" * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113704321A (en) * | 2021-08-11 | 2021-11-26 | 国电南瑞科技股份有限公司 | Power grid abnormal data identification method, device and system |
| CN114039886A (en) * | 2021-10-19 | 2022-02-11 | 国网山东省电力公司日照供电公司 | Data correctness identification method and device based on four-differential and autocorrelation analysis |
| CN117421610A (en) * | 2023-12-19 | 2024-01-19 | 山东德源电力科技股份有限公司 | Data anomaly analysis method for electric energy meter running state early warning |
| CN117421610B (en) * | 2023-12-19 | 2024-03-15 | 山东德源电力科技股份有限公司 | Data anomaly analysis method for electric energy meter running state early warning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112213687B (en) | 2023-05-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223196B (en) | Anti-electricity-stealing analysis method based on typical industry feature library and anti-electricity-stealing sample library | |
| CN112213687A (en) | Gateway electric energy meter data anomaly detection method and system based on pseudo anomaly point identification | |
| CN106709662B (en) | Power equipment operation condition division method | |
| CN105071983B (en) | Abnormal load detection method for cloud calculation on-line business | |
| CN111612651B (en) | Abnormal electric quantity data detection method based on long-term and short-term memory network | |
| CN103996077B (en) | Electrical equipment fault prediction method based on multi-dimensional time sequence | |
| CN110336534A (en) | A kind of method for diagnosing faults based on photovoltaic array electric parameter time series feature extraction | |
| CN113518011B (en) | Abnormality detection method and apparatus, electronic device, and computer-readable storage medium | |
| Liu et al. | Wide-area phase-angle measurements for islanding detection—An adaptive nonlinear approach | |
| CN105512474A (en) | Transformer state monitoring data anomaly detection method | |
| CN106596754B (en) | The appraisal procedure and device of oil chromatography sensor availability | |
| CN111382789B (en) | Power load identification method and system based on machine learning | |
| CN102112933A (en) | Error detection method and system | |
| CN111027408A (en) | Load identification method based on support vector machine and V-I curve characteristics | |
| CN103103570B (en) | Based on the aluminium cell condition diagnostic method of pivot similarity measure | |
| CN109389325B (en) | Method for evaluating state of electronic transformer of transformer substation based on wavelet neural network | |
| CN113556629B (en) | Intelligent ammeter error remote estimation method and device | |
| CN116380445B (en) | Equipment state diagnosis method and related device based on vibration waveform | |
| Zhu et al. | Networked time series shapelet learning for power system transient stability assessment | |
| CN111371188A (en) | Real-time early warning system for abnormal power consumption data | |
| Pan et al. | Uncertainty analysis based on non-parametric statistical modelling method for photovoltaic array output and its application in fault diagnosis | |
| Iturrino-García et al. | An innovative single shot power quality disturbance detector algorithm | |
| CN115062272A (en) | Water quality monitoring data abnormity identification and early warning method | |
| CN110852906B (en) | Method and system for identifying electricity stealing suspicion based on high-dimensional random matrix | |
| CN115081673A (en) | Oil and gas pipeline abnormity prediction method and device, electronic equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230831 Address after: Floor 1, No. 50, section 4, Renmin South Road, Wuhou District, Chengdu, Sichuan 610000 Patentee after: Marketing service center of State Grid Sichuan Electric Power Co. Address before: 610000 1-7 25 Qinghua Road, Qingyang District, Chengdu, Sichuan, China 25 Patentee before: STATE GRID SICHUAN ELECTRIC POWER Research Institute |