CN117421610B - Data anomaly analysis method for electric energy meter running state early warning - Google Patents
Data anomaly analysis method for electric energy meter running state early warning Download PDFInfo
- Publication number
- CN117421610B CN117421610B CN202311743281.5A CN202311743281A CN117421610B CN 117421610 B CN117421610 B CN 117421610B CN 202311743281 A CN202311743281 A CN 202311743281A CN 117421610 B CN117421610 B CN 117421610B
- Authority
- CN
- China
- Prior art keywords
- data point
- abnormal
- data
- electricity
- obtaining
- 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.)
- Active
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 16
- 230000005611 electricity Effects 0.000 claims abstract description 143
- 230000002159 abnormal effect Effects 0.000 claims abstract description 141
- 230000005856 abnormality Effects 0.000 claims abstract description 83
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 238000000034 method Methods 0.000 claims description 26
- 238000000354 decomposition reaction Methods 0.000 claims description 15
- 238000012937 correction Methods 0.000 claims description 9
- 238000012512 characterization method Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 230000006399 behavior Effects 0.000 description 22
- 238000009499 grossing Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of electric power metering detection, in particular to a data anomaly analysis method for early warning of an operation state of an electric energy meter; obtaining an electricity abnormal index according to the electricity difference characteristics of the data points in the electricity data sequence; obtaining the power consumption abnormality degree according to the power consumption change characteristics of the data points and the power consumption difference characteristics of the data points and other power consumption data sequences; and obtaining abnormal data points and abnormal weight coefficients according to the power utilization abnormality indexes and the power utilization abnormality degrees. Obtaining a final abnormal data point according to the number characteristics of other abnormal data points in the neighborhood of the abnormal data point; and obtaining the self-adaptive weight of the final abnormal data point. Fitting is carried out according to an initial residual error item, a preset weight and a self-adaptive weight to obtain a trend item; and obtaining a final residual error item according to the initial season item and the trend item and carrying out anomaly detection, so that the detection accuracy of the operation anomaly of the electric energy meter is improved.
Description
Technical Field
The invention relates to the technical field of electric power metering detection, in particular to a data anomaly analysis method for early warning of an electric energy meter running state.
Background
The electric energy meter is an instrument for measuring and recording electric energy consumption, electric energy metering is an important step of energy calculation between an electric power enterprise and a user, and along with the diversified development of an electric power market, the electric energy metering work is required to ensure the reliability and accuracy of metering; therefore, the monitoring and early warning of the running state of the electric energy meter are of great importance, and timely reminding of abnormal conditions such as power interruption, overload, underload and power theft is carried out.
The existing monitoring method generally adopts a time sequence decomposition algorithm to process, the data sequence of the electric energy meter is decomposed to obtain decomposed residual components, and the residual components are reminded through the change characteristics of the residual components. According to the time sequence decomposition principle, the prior STL time sequence decomposition algorithm generally decomposes an original data information curve into three components of a season term, a trend term and a residual term; in the decomposition process, the fitting mode is local weighted regression smoothing, and the fitting mode generally gives greater weight to data points closer to the observed value and gives smaller weight to data points farther from the observed value; therefore, the residual component obtained by the decomposition method only contains abnormal data points or noise data points with larger mutation degree. However, for data points with lower abnormal characteristics, accurate capturing cannot be realized; such as power theft during operation of the electric energy meter; the reading of the electric energy meter does not stop or even jump according to the normal speed increase or the increase speed, and abnormal data change occurs; however, the weight of the data point is not adjusted according to the abnormal characteristics of the data point in the fitting process, so that the errors of the data and the weight in the fitting process are larger, the accuracy of the obtained residual error component is low, the abnormal condition of the power stealing behavior is difficult to be highlighted, and the operation early warning accuracy of the electric energy meter is influenced.
Disclosure of Invention
In order to solve the technical problem that the accuracy of residual components of the electric energy meter data obtained through the existing time sequence decomposition algorithm is low and influences the operation early warning accuracy of the electric energy meter, the invention aims to provide a data anomaly analysis method for the operation state early warning of the electric energy meter, and the adopted technical scheme is as follows:
acquiring a power consumption data sequence of the electric energy meter; obtaining an electricity utilization abnormality index of a data point according to the electricity utilization difference characteristic between the data point and an adjacent data point in the electricity utilization data sequence;
obtaining the power consumption abnormality degree according to the power consumption change characteristics of the data points and the adjacent data points and the power consumption difference characteristics of the data points and other power consumption data sequences at the same time; obtaining abnormal data points and corresponding abnormal weight coefficients according to the electricity consumption abnormality indexes and the electricity consumption abnormality degrees;
obtaining final abnormal data points according to the quantity characteristics of other abnormal data points in the preset neighborhood range of the abnormal data points; decomposing the electricity utilization data sequence through an STL decomposition algorithm to obtain an initial season term and an initial residual term; obtaining self-adaptive weights according to the abnormal weight coefficients of the final abnormal data points and preset weights;
fitting according to the initial residual error item, the preset weight and the self-adaptive weight to obtain a trend item; and obtaining a final residual error item according to the initial season item and the trend item, and performing anomaly detection.
Further, the step of obtaining the electricity utilization abnormality index of the data point according to the electricity utilization difference characteristic between the data point and the adjacent data point in the electricity utilization data sequence comprises the following steps:
calculating the absolute value of the difference value between the data point and the adjacent data point in the electricity utilization data sequence to obtain the adjacent electricity utilization difference; calculating the average value of the adjacent electricity consumption differences of the electricity consumption data sequence to obtain average difference of adjacent electricity consumption levels; and calculating the difference value between the average difference of the adjacent power utilization levels and the adjacent power utilization difference to obtain the power utilization abnormality index of the data point.
Further, the step of obtaining the abnormal degree of electricity consumption according to the electricity consumption change characteristics of the data point and the adjacent data points and the electricity consumption difference characteristics of the data point and other electricity consumption data sequences at the same time comprises the following steps:
constructing a two-dimensional rectangular coordinate system of the electricity utilization data sequence, calculating the slope of a line segment between the data point and an adjacent data point in the two-dimensional rectangular coordinate system and carrying out negative correlation mapping to obtain an electricity utilization change characteristic value of the data point; calculating the product of the electricity utilization change characteristic value and a preset first coefficient to obtain a first electricity utilization abnormality degree;
calculating the average value of the adjacent electricity utilization differences of all other electricity utilization data sequences at the same moment with the data point, and obtaining the average value of the adjacent electricity utilization difference moment; calculating and normalizing the absolute value of the difference between the adjacent power utilization difference of the data point and the average value of the adjacent power utilization difference time to obtain a power utilization difference time characterization value of the data point; calculating the product of a preset second coefficient and the electricity utilization difference moment characterization value to obtain a second degree of abnormality of electricity utilization; and calculating the sum value of the first power consumption abnormality degree and the second power consumption abnormality degree to obtain the power consumption abnormality degree of the data point.
Further, the step of obtaining the abnormal data point and the corresponding abnormal weight coefficient according to the electricity consumption abnormality index and the electricity consumption abnormality degree comprises the following steps:
when the electricity consumption abnormality index of the data point is not lower than a preset first threshold value and the electricity consumption abnormality degree is greater than a preset second threshold value, the data point is an abnormal data point, and the electricity consumption abnormality degree of the abnormal data point is the abnormality weight coefficient.
Further, the step of obtaining the final outlier data point according to the number characteristics of other outlier data points in the preset neighborhood range of the outlier data point includes:
for any abnormal data point in the electricity utilization data sequence, calculating the number ratio of other abnormal data points to data points in a preset neighborhood range of the any abnormal data point, and obtaining the neighborhood anomaly degree of the any abnormal data point; and when the neighborhood anomaly degree of the random anomaly data point exceeds a preset third threshold value, the random anomaly data point is the final anomaly data point.
Further, the step of obtaining the adaptive weight according to the abnormal weight coefficient of the final abnormal data point and the preset weight includes:
performing negative correlation mapping on the abnormal weight coefficient of the final abnormal data point to obtain a weight correction coefficient of the final abnormal data point; and calculating the product of the weight correction coefficient of the final abnormal data point and the preset weight to obtain the self-adaptive weight of the final abnormal data point.
Further, the step of obtaining a final residual term according to the initial season term and the trend term and performing anomaly detection comprises the following steps:
combining the initial season term and the trend term to obtain a corrected season term, and eliminating the corrected season term and the trend term in the electricity consumption data sequence to obtain the final residual term; and performing anomaly detection according to the value of the final residual error item corresponding to the data point.
The invention has the following beneficial effects:
in the embodiment of the invention, the obtained electricity consumption abnormality index can preliminarily indicate whether electricity stealing behavior exists in the electricity consumption process according to the electricity consumption difference characteristics; the obtained electricity consumption abnormality degree can be combined with other electricity consumption data sequences to further characterize whether electricity stealing behavior exists or not; and further, according to the power consumption abnormality index and the power consumption abnormality degree, abnormal data points and corresponding abnormal weight coefficients can be accurately acquired. The final abnormal data point can be obtained, so that misjudgment of the abnormal data point caused by normal electricity behavior can be eliminated, and the calculation accuracy of a final residual error item is further improved; the self-adaptive weight can be obtained to carry out self-adaptive adjustment on the weight according to the abnormal degree of the data point, so that the fitting process of the trend item is more accurate, and the abnormal characteristics of the power stealing behavior are highlighted; finally, the invention carries out abnormal detection on the running state of the electric energy meter according to the final residual error item, thereby improving the detection accuracy on abnormal electricity consumption conditions such as electricity stealing behavior and the like.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a data anomaly analysis method for early warning of an operation state of an electric energy meter according to an embodiment of the present invention.
Detailed Description
In order to further explain the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects thereof for an abnormal data analysis method for early warning of an operation state of an electric energy meter according to the invention, which is provided by the invention, with reference to the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the data anomaly analysis method for early warning of the running state of the electric energy meter.
Referring to fig. 1, a flowchart of a data anomaly analysis method for early warning of an operation state of an electric energy meter according to an embodiment of the present invention is shown, and the method includes the following steps:
step S1, acquiring a power consumption data sequence of an electric energy meter; and obtaining the electricity utilization abnormality index of the data point according to the electricity utilization difference characteristics between the data point and the adjacent data point in the electricity utilization data sequence.
In the embodiment of the invention, the implementation scene is to pre-warn the running state of the electric energy meter of the factory with continuous electricity consumption and small power variation amplitude. Firstly, acquiring a power consumption data sequence of the electric energy meter, wherein in the embodiment of the invention, the power consumption data sequence is acquired by taking 24 hours per day as a time period, under the condition that the power consumption behavior is normal, the change characteristics of the power consumption data sequence per day are similar, and an implementer can determine the length of the power consumption data sequence according to implementation scenes.
After the electricity data sequence of the electric energy meter in the electricity scene is acquired, the electricity state can be detected, the electricity data sequence is generally decomposed into a trend item, a season item and a residual item through an STL time sequence decomposition algorithm, and the STL time sequence decomposition algorithm is required to be explained, and the specific decomposition steps are not repeated; and determining the abnormal condition of the electric energy meter through the numerical variation characteristics in the residual error items. In the STL algorithm, the fitting is completed by acquiring smooth estimation of three components through a local weighted regression smoothing algorithm, the local weighted regression smoothing algorithm belongs to the prior art, and because weights of different data in the smoothing fitting process are determined based on distances from data points needing to be fitted and the situation that the weights need to be reduced when the data points are abnormal is not considered, errors easily occur in the fitting process, so that the running state monitoring of the electric energy meter is inaccurate. Therefore, the self-adaptive adjustment of the data fitting weight in the analysis process is needed, the abnormal data of the power stealing behavior is highlighted, and the accuracy of the running state detection is improved.
The electricity consumption data displayed by the electric energy meter is gradually increased according to the actual electricity consumption situation, if the electricity stealing behavior occurs, the electricity consumption data of the electric energy meter is slowly increased or even not increased, so that the electricity consumption abnormality index of the data points can be obtained according to the characteristic of the electricity consumption difference between the data points and the adjacent data points in the electricity consumption data sequence.
Preferably, in one embodiment of the present invention, obtaining the power consumption abnormality index includes: calculating the absolute value of the difference value between the data point and the adjacent data point in the electricity data sequence to obtain the adjacent electricity difference; in the embodiment of the invention, the adjacent electricity consumption difference is calculated by the data point and the adjacent last data point, and the first data point does not participate in calculation; the adjacent electricity consumption difference represents the change and increase condition of the electricity consumption, and the larger the adjacent electricity consumption difference characteristic is, the larger the electricity consumption in the period is. Calculating the average value of adjacent power consumption differences of the power consumption data sequence to obtain average difference of adjacent power consumption levels; the average difference of adjacent power consumption levels represents the overall power consumption degree in the power consumption data sequence, and the larger the average difference of the adjacent power consumption levels is, the larger the overall power consumption of the current day is. And calculating the difference value between the adjacent power utilization average difference and the adjacent power utilization difference to obtain a power utilization abnormality index of a data point, wherein for a factory with continuous power utilization and smaller power variation, when the power utilization abnormality index is larger, the possibility of power theft in the period is larger.
Step S2, obtaining the power consumption abnormality degree according to the power consumption change characteristics of the data points and the adjacent data points and the power consumption difference characteristics of the data points and other power consumption data sequences at the same time; and obtaining abnormal data points and corresponding abnormal weight coefficients according to the power consumption abnormality indexes and the power consumption abnormality degrees.
After the electricity consumption abnormality indexes of the data points at different moments are obtained, in order to accurately analyze the data points with electricity stealing behaviors, the accuracy of abnormality detection is improved, and further analysis on electricity stealing characteristics is needed, so that the electricity consumption abnormality degree can be obtained according to the electricity consumption change characteristics of the data points and adjacent data points and the electricity consumption difference characteristics of the data points and other electricity consumption data sequences at the same moment.
Preferably, in one embodiment of the present invention, obtaining the degree of power consumption abnormality includes: constructing a two-dimensional rectangular coordinate system of the electricity utilization data sequence, calculating the slope of a line segment between a data point and an adjacent data point in the two-dimensional rectangular coordinate system, and carrying out negative correlation mapping to obtain an electricity utilization change characteristic value of the data point; when the slope is larger, the electricity consumption is increased faster, the possibility of electricity stealing is smaller, and the characteristic value of electricity consumption change is smaller; conversely, when the power consumption change characteristic value is larger, the possibility of power theft behavior is higher. Calculating the product of the electricity variation characteristic value and a preset first coefficient to obtain a first electricity abnormality degree; in the embodiment of the invention, the preset first coefficient is 0.4, and an implementer can determine according to an implementation scene by himself, and the implementation is used for determining the weight ratio of the first abnormal degree of electricity consumption in the abnormal degree of electricity consumption.
Further, the average value of adjacent electricity consumption differences of all other electricity consumption data sequences at the same time as the data point is calculated, the average value of adjacent electricity consumption difference time is obtained, and for a factory scene with continuous electricity consumption and small power change, adjacent electricity consumption difference values corresponding to the same time every day are similar. Calculating the absolute value of the difference value between the adjacent power utilization difference of the data point and the average value of the adjacent power utilization difference time, and obtaining and normalizing the power utilization difference time characterization value of the data point; the larger the electricity utilization difference moment characterization value is, the larger the electricity utilization characteristic difference between the data point and other electricity utilization data sequences at the same moment is, which means that the electricity stealing behavior can occur in the time period of the data point. Calculating the product of a preset second coefficient and the electricity consumption difference moment characterization value to obtain a second electricity consumption abnormality degree; in the embodiment of the invention, the preset second coefficient is 0.6, and an implementer can determine the second abnormal degree of electricity consumption according to the implementation scene by himself, and the second abnormal degree of electricity consumption is used for determining the weight ratio of the second abnormal degree of electricity consumption in the abnormal degree of electricity consumption. And calculating the sum of the first power consumption abnormality degree and the second power consumption abnormality degree to obtain the power consumption abnormality degree of the data point, wherein the greater the power consumption abnormality degree of the data point is, the greater the possibility of power theft behavior is, and the more abnormal the power consumption behavior is. The formula for acquiring the electricity utilization abnormality degree of the data point specifically comprises the following steps:
in the method, in the process of the invention,indicating the degree of abnormality of electricity consumption of data points in the electricity consumption data sequence, +.>Representing a preset first coefficient,/->Representing a preset second coefficient,/->Representing data points and phasesSlope of line segment between adjacent data points, +.>Represents an exponential function based on natural constants, < ->Characteristic value of the power consumption change representing the data point, +.>Representing a normalization function->Representing adjacent power consumption difference of data points, +.>Representing the number of other power sequences, +.>Adjacent electricity consumption difference time average value of the first other electricity consumption data sequence representing the same time as the data point,/>A value representing the time of day of the difference in power consumption of the data point, < >>Indicating a first degree of abnormality in the power consumption,indicating a second degree of abnormality in power consumption.
After the power consumption abnormality index and the power consumption abnormality degree of the data point are obtained, the abnormal characteristics of the data point can be analyzed according to the power consumption abnormality index and the power consumption abnormality degree of the data point, and the greater the power consumption abnormality index and the power consumption abnormality degree of the data point, the greater the possibility of power theft behavior of the data point at the moment; therefore, abnormal data points and corresponding abnormal weight coefficients can be obtained according to the power consumption abnormality index and the power consumption abnormality degree, and the method specifically comprises the following steps: when the power consumption abnormality index of the data point is not lower than a preset first threshold value and the power consumption abnormality degree is greater than a preset second threshold value, the data point is an abnormal data point, and the power consumption abnormality degree of the abnormal data point is an abnormal weight coefficient; in the embodiment of the invention, the first threshold value is preset to be 0, the second threshold value is preset to be 0.7, and the implementation can be determined by the implementation according to the implementation scene.
Step S3, obtaining final abnormal data points according to the quantity characteristics of other abnormal data points in the preset neighborhood range of the abnormal data points; decomposing the power consumption data sequence through an STL decomposition algorithm to obtain an initial season term and an initial residual term; and obtaining self-adaptive weights according to the abnormal weight coefficients of the final abnormal data points and preset weights.
Because the illegal connection is to the power grid, the electric energy meter is bypassed to carry out frequent electric energy theft, and the illegal connection can lead to sudden increase or decrease of current and fluctuation of voltage in a circuit, so that the numerical value of the electric energy meter is frequently jumped; however, transient loads of high power devices or high currents may cause short jumps in the meter; both can lead to data points to become abnormal data points, and in order to avoid the jump caused by normal conditions to misunderstand the abnormal data points, further analysis is needed to be carried out on the abnormal data points, so that the final abnormal data points are obtained according to the quantity characteristics of other abnormal data points in the preset neighborhood range of the abnormal data points.
Preferably, in one embodiment of the present invention, obtaining the final outlier data point comprises: for any abnormal data point in the electricity utilization data sequence, calculating the number ratio of other abnormal data points to data points in a preset neighborhood range of the any abnormal data point, and obtaining the neighborhood abnormality degree of the any abnormal data point; when the neighborhood degree of any abnormal data point is larger, the more abnormal data points in the preset neighborhood range are, and the more frequent jump characteristics of the power stealing behavior are met; on the contrary, when the degree of abnormality of the neighborhood is smaller, which means that fewer abnormal data points are in a preset neighborhood range, the jump characteristic caused by starting and stopping high-power equipment is more met, in the embodiment of the invention, the preset neighborhood range is 15 data points around the abnormal data point as the center, if the abnormal data point is at the edge of the power consumption data sequence, the abnormal data point is calculated by using the neighborhood range closest to the center, and an implementer can determine the preset neighborhood range according to implementation scenes. When the neighborhood anomaly degree of any abnormal data point exceeds a preset third threshold value, the any abnormal data point is the final abnormal data point, and the preset third threshold value is 0.5 in the embodiment of the invention, so that an implementer can determine according to implementation scenes.
Further, the operation state of the electric energy meter can be obtained through an STL algorithm to carry out anomaly detection, firstly, the power consumption data sequence is decomposed through an STL decomposition algorithm to obtain an initial season item and an initial residual item, and then different weights are given to data points in the residual item to carry out trend item fitting through local weighted regression; when the STL time sequence decomposition algorithm uses local weighted regression smoothing to fit trend items, a weight is distributed to each data point according to the distance from a fitting point to be fitted, the data points in a smoothing window and the weights thereof are input into a regression model, a larger weight is distributed to the data points with a smaller distance, a smaller weight is distributed to the data points with a larger observation value, and for the final abnormal data points acquired according to the power theft behavior, the weight needs to be adjusted to highlight the abnormal characteristics of the power theft behavior, so that the self-adaptive weight is acquired according to the abnormal weight coefficient and the preset weight of the final abnormal data points.
Preferably, in one embodiment of the present invention, obtaining the adaptive weights includes: performing negative correlation mapping on the abnormal weight coefficient of the final abnormal data point to obtain a weight correction coefficient of the final abnormal data point; and calculating the product of the weight correction coefficient of the final abnormal data point and the preset weight to obtain the self-adaptive weight of the final abnormal data point, wherein when the abnormal weight coefficient of the final abnormal data point is larger, the corresponding self-adaptive weight is smaller, and the calculation principle that the more obvious the abnormal characteristic of the power stealing behavior is, the smaller the assigned weight is. For the data points in the normal condition, the weight in fitting is still a preset weight, and it is to be noted that the preset weight is the weight distributed according to the distance in fitting by the local weighted regression algorithm in the STL algorithm, which belongs to the prior art, and specific calculation steps are not repeated. The formula for obtaining the adaptive weights of the final outlier data points includes:
in the method, in the process of the invention,adaptive weights representing final outlier data points, +.>Representing the preset weights of the final outlier data points,abnormality weight coefficient representing final abnormality data point, +.>And weight correction coefficients representing the final outlier data points.
S4, fitting according to the initial residual error item, the preset weight and the self-adaptive weight to obtain a trend item; and obtaining a final residual error item according to the initial season item and the trend item, and performing anomaly detection.
After the preset weight of the normal data point and the self-adaptive weight of the final abnormal data point are obtained, fitting is carried out through a local weighted regression smoothing algorithm according to the preset weight of the normal data point and the self-adaptive weight of the final abnormal data point in the residual error item to obtain a trend item; further, the initial season term and the trend term are combined to correct the initial season term to obtain a corrected season term, so that the sum of the initial season terms is matched with the original time sequence data; removing the corrected season items and trend items in the electricity data sequence to obtain final residual items; the final residual error item comprises the rest part which cannot be interpreted by the trend item and the correction season item in the data change characteristic of the electricity utilization data sequence, namely an abnormal value contained in the operation process of the electric energy meter, and the abnormal value can accurately and prominently reflect the electricity stealing behavior of the electricity utilization process. And finally, carrying out anomaly detection according to the numerical value of a final residual error item corresponding to the data point, preferably, when the standard deviation of the final residual error item exceeds a preset anomaly threshold value, meaning that the data fluctuation degree of the final residual error item is larger, considering that the running state of the electric energy meter is abnormal, and carrying out early warning and reminding in time, wherein an implementer can determine the anomaly threshold value according to implementation scenes. Therefore, the data abnormality degree of the power theft behavior is highlighted and the accuracy of the running state early warning of the electric energy meter is improved by improving the acquisition of residual error items in the STL algorithm according to the power utilization characteristics of the power theft behavior.
In summary, the embodiment of the invention provides a data anomaly analysis method for early warning of an operation state of an electric energy meter; obtaining an electricity abnormal index according to the electricity difference characteristics of the data points in the electricity data sequence; obtaining the power consumption abnormality degree according to the power consumption change characteristics of the data points and the power consumption difference characteristics of the data points and other power consumption data sequences; and obtaining abnormal data points and abnormal weight coefficients according to the power utilization abnormality indexes and the power utilization abnormality degrees. Obtaining a final abnormal data point according to the number characteristics of other abnormal data points in the neighborhood of the abnormal data point; and obtaining the self-adaptive weight of the final abnormal data point. Fitting is carried out according to an initial residual error item, a preset weight and a self-adaptive weight to obtain a trend item; and obtaining a final residual error item according to the initial season item and the trend item and carrying out anomaly detection, so that the detection accuracy of the operation anomaly of the electric energy meter is improved.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (4)
1. The data anomaly analysis method for the early warning of the running state of the electric energy meter is characterized by comprising the following steps of:
acquiring a power consumption data sequence of the electric energy meter; obtaining an electricity utilization abnormality index of a data point according to the electricity utilization difference characteristic between the data point and an adjacent data point in the electricity utilization data sequence;
obtaining the power consumption abnormality degree according to the power consumption change characteristics of the data points and the adjacent data points and the power consumption difference characteristics of the data points and other power consumption data sequences at the same time; obtaining abnormal data points and corresponding abnormal weight coefficients according to the electricity consumption abnormality indexes and the electricity consumption abnormality degrees;
obtaining final abnormal data points according to the quantity characteristics of other abnormal data points in the preset neighborhood range of the abnormal data points; decomposing the electricity utilization data sequence through an STL decomposition algorithm to obtain an initial season term and an initial residual term; obtaining self-adaptive weights according to the abnormal weight coefficients of the final abnormal data points and preset weights;
fitting according to the initial residual error item, the preset weight and the self-adaptive weight to obtain a trend item; obtaining a final residual error item according to the initial season item and the trend item and performing anomaly detection;
the step of obtaining the abnormal data point and the corresponding abnormal weight coefficient according to the electricity consumption abnormality index and the electricity consumption abnormality degree comprises the following steps:
when the electricity consumption abnormality index of the data point is not lower than a preset first threshold value and the electricity consumption abnormality degree is greater than a preset second threshold value, the data point is an abnormal data point, and the electricity consumption abnormality degree of the abnormal data point is the abnormality weight coefficient;
the step of obtaining the final abnormal data point according to the number characteristics of other abnormal data points in the preset neighborhood range of the abnormal data point comprises the following steps:
for any abnormal data point in the electricity utilization data sequence, calculating the number ratio of other abnormal data points to data points in a preset neighborhood range of the any abnormal data point, and obtaining the neighborhood anomaly degree of the any abnormal data point; when the neighborhood anomaly degree of the any abnormal data point exceeds a preset third threshold value, the any abnormal data point is the final abnormal data point;
the step of obtaining the self-adaptive weight according to the abnormal weight coefficient of the final abnormal data point and the preset weight comprises the following steps:
performing negative correlation mapping on the abnormal weight coefficient of the final abnormal data point to obtain a weight correction coefficient of the final abnormal data point; calculating the product of the weight correction coefficient of the final abnormal data point and the preset weight to obtain the self-adaptive weight of the final abnormal data point;
specifically, the adaptive weights are:
;
in the method, in the process of the invention,adaptive weights representing final outlier data points, +.>Preset weights representing final outlier data points, +.>Abnormality weight coefficient representing final abnormality data point, +.>And the weight correction coefficient of the final abnormal data point is represented, and the preset weight is the weight of the final abnormal data point distributed according to the distance when the STL time sequence decomposition algorithm is used for fitting through a local weighted regression algorithm.
2. The method for data anomaly analysis for electric energy meter operation state warning according to claim 1, wherein the step of obtaining the electricity anomaly index of a data point in the electricity data sequence according to the electricity difference characteristic between the data point and an adjacent data point comprises:
calculating the absolute value of the difference value between the data point and the adjacent data point in the electricity utilization data sequence to obtain the adjacent electricity utilization difference; calculating the average value of the adjacent electricity consumption differences of the electricity consumption data sequence to obtain average difference of adjacent electricity consumption levels; and calculating the difference value between the average difference of the adjacent power utilization levels and the adjacent power utilization difference to obtain the power utilization abnormality index of the data point.
3. The method for data anomaly analysis for electric energy meter operation state early warning according to claim 2, wherein the step of obtaining the degree of electric anomaly according to the electric variation characteristics of the data point and the adjacent data points and the electric variation characteristics of the data point and other electric data sequences at the same time comprises the following steps:
constructing a two-dimensional rectangular coordinate system of the electricity utilization data sequence, calculating the slope of a line segment between the data point and an adjacent data point in the two-dimensional rectangular coordinate system and carrying out negative correlation mapping to obtain an electricity utilization change characteristic value of the data point; calculating the product of the electricity utilization change characteristic value and a preset first coefficient to obtain a first electricity utilization abnormality degree;
calculating the average value of the adjacent electricity utilization differences of all other electricity utilization data sequences at the same moment with the data point, and obtaining the average value of the adjacent electricity utilization difference moment; calculating and normalizing the absolute value of the difference between the adjacent power utilization difference of the data point and the average value of the adjacent power utilization difference time to obtain a power utilization difference time characterization value of the data point; calculating the product of a preset second coefficient and the electricity utilization difference moment characterization value to obtain a second degree of abnormality of electricity utilization; and calculating the sum value of the first power consumption abnormality degree and the second power consumption abnormality degree to obtain the power consumption abnormality degree of the data point.
4. The method for data anomaly analysis for electric energy meter operation state warning according to claim 1, wherein the step of obtaining a final residual term and anomaly detection according to the initial season term and the trend term comprises:
combining the initial season term and the trend term to obtain a corrected season term, and eliminating the corrected season term and the trend term in the electricity consumption data sequence to obtain the final residual term; and performing anomaly detection according to the value of the final residual error item corresponding to the data point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311743281.5A CN117421610B (en) | 2023-12-19 | 2023-12-19 | Data anomaly analysis method for electric energy meter running state early warning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311743281.5A CN117421610B (en) | 2023-12-19 | 2023-12-19 | Data anomaly analysis method for electric energy meter running state early warning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117421610A CN117421610A (en) | 2024-01-19 |
CN117421610B true CN117421610B (en) | 2024-03-15 |
Family
ID=89528805
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311743281.5A Active CN117421610B (en) | 2023-12-19 | 2023-12-19 | Data anomaly analysis method for electric energy meter running state early warning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117421610B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117786371B (en) * | 2024-02-27 | 2024-05-10 | 聊城市检验检测中心 | Temperature monitoring data optimization prediction analysis method and system |
CN117931094A (en) * | 2024-03-21 | 2024-04-26 | 山东奥斯瑞特检验检测有限公司 | Block chain-based reliable storage method for ambient air monitoring data |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112213687A (en) * | 2020-10-10 | 2021-01-12 | 国网四川省电力公司电力科学研究院 | Gateway electric energy meter data anomaly detection method and system based on pseudo anomaly point identification |
WO2021056724A1 (en) * | 2019-09-23 | 2021-04-01 | 平安科技(深圳)有限公司 | Anomaly detection method and apparatus, electronic device and storage medium |
CN112884000A (en) * | 2021-01-15 | 2021-06-01 | 国网河南省电力公司营销服务中心 | Power utilization inspection intelligent diagnosis method and diagnosis system based on data mining |
CN115310982A (en) * | 2022-04-01 | 2022-11-08 | 云南电网有限责任公司昭通供电局 | Electricity larceny prevention early warning data analysis method |
CN115935296A (en) * | 2023-03-09 | 2023-04-07 | 国网山东省电力公司营销服务中心(计量中心) | Electric energy data metering method and system |
CN116090916A (en) * | 2023-04-10 | 2023-05-09 | 淄博海草软件服务有限公司 | Early warning system for enterprise internal purchase fund accounting |
CN116627707A (en) * | 2023-07-20 | 2023-08-22 | 中孚安全技术有限公司 | Detection method and system for abnormal operation behavior of user |
CN116702081A (en) * | 2023-08-07 | 2023-09-05 | 西安格蒂电力有限公司 | Intelligent inspection method for power distribution equipment based on artificial intelligence |
CN116865269A (en) * | 2023-09-01 | 2023-10-10 | 山东泰开电力电子有限公司 | Wind turbine generator system high harmonic compensation method and system |
CN117220441A (en) * | 2023-09-20 | 2023-12-12 | 北京中电普华信息技术有限公司 | Information monitoring intelligent early warning method based on panoramic monitoring |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3623964A1 (en) * | 2018-09-14 | 2020-03-18 | Verint Americas Inc. | Framework for the automated determination of classes and anomaly detection methods for time series |
CN114365094A (en) * | 2019-09-23 | 2022-04-15 | 谷歌有限责任公司 | Timing anomaly detection using inverted indices |
-
2023
- 2023-12-19 CN CN202311743281.5A patent/CN117421610B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021056724A1 (en) * | 2019-09-23 | 2021-04-01 | 平安科技(深圳)有限公司 | Anomaly detection method and apparatus, electronic device and storage medium |
CN112213687A (en) * | 2020-10-10 | 2021-01-12 | 国网四川省电力公司电力科学研究院 | Gateway electric energy meter data anomaly detection method and system based on pseudo anomaly point identification |
CN112884000A (en) * | 2021-01-15 | 2021-06-01 | 国网河南省电力公司营销服务中心 | Power utilization inspection intelligent diagnosis method and diagnosis system based on data mining |
CN115310982A (en) * | 2022-04-01 | 2022-11-08 | 云南电网有限责任公司昭通供电局 | Electricity larceny prevention early warning data analysis method |
CN115935296A (en) * | 2023-03-09 | 2023-04-07 | 国网山东省电力公司营销服务中心(计量中心) | Electric energy data metering method and system |
CN116090916A (en) * | 2023-04-10 | 2023-05-09 | 淄博海草软件服务有限公司 | Early warning system for enterprise internal purchase fund accounting |
CN116627707A (en) * | 2023-07-20 | 2023-08-22 | 中孚安全技术有限公司 | Detection method and system for abnormal operation behavior of user |
CN116702081A (en) * | 2023-08-07 | 2023-09-05 | 西安格蒂电力有限公司 | Intelligent inspection method for power distribution equipment based on artificial intelligence |
CN116865269A (en) * | 2023-09-01 | 2023-10-10 | 山东泰开电力电子有限公司 | Wind turbine generator system high harmonic compensation method and system |
CN117220441A (en) * | 2023-09-20 | 2023-12-12 | 北京中电普华信息技术有限公司 | Information monitoring intelligent early warning method based on panoramic monitoring |
Non-Patent Citations (4)
Title |
---|
Anomaly Detection Model of Power Grid Data Based on STL Decomposition;Cuicui Zhang 等;2021 IEEE 5th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC);20211231;第1262-1265页 * |
基于ARMA和K-means聚类的用电量数据异常识别;梁捷;梁广明;;湖北电力;20190828(第04期);第65-70页 * |
物联网中异常数据优化检测仿真研究;张杰;景雯;任培花;;计算机仿真;20180415(第04期);第449-452页 * |
电能计量及反窃电方法;李季;;电子技术与软件工程;20130401(第07期);第43页 * |
Also Published As
Publication number | Publication date |
---|---|
CN117421610A (en) | 2024-01-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN117421610B (en) | Data anomaly analysis method for electric energy meter running state early warning | |
CN116879662B (en) | Transformer fault detection method based on data analysis | |
US10852357B2 (en) | System and method for UPS battery monitoring and data analysis | |
CN106505557B (en) | Remote measurement error identification method and device | |
CN108696368B (en) | Network element health state detection method and equipment | |
CN116703183B (en) | Novel power distribution network system power load assessment method | |
CN111191671B (en) | Electrical appliance waveform detection method, system, electronic equipment and storage medium | |
CN114491383A (en) | Abnormal data processing method and system for bridge monitoring | |
CN117150244B (en) | Intelligent power distribution cabinet state monitoring method and system based on electrical parameter analysis | |
JP7437584B2 (en) | Machine learning-based method and apparatus for power line disturbance classification | |
US20130158903A1 (en) | Method and System for Detecting Transients in Power Grids | |
CN115343579B (en) | Power grid fault analysis method and device and electronic equipment | |
CN110658414B (en) | Power electronic parametric fault detection method based on model | |
CN110196354B (en) | Method and device for detecting switching event of load | |
CN117668684B (en) | Power grid electric energy data anomaly detection method based on big data analysis | |
CN117892246B (en) | Data processing method for intelligent switch cabinet | |
CN118070195A (en) | Mining alternating current frequency converter abnormal data state monitoring system | |
CN117713505B (en) | High-side current detection method and system for switching power supply | |
CN117571107B (en) | Intelligent unattended wagon balance anomaly monitoring system | |
CN117110976B (en) | Ammeter anomaly detection method and device, electronic equipment and storage medium | |
CN117851414B (en) | Lightning arrester aging test data storage method and system | |
CN112307997B (en) | Power signal reconstruction method and system by using main mode decomposition | |
CN118091325A (en) | Intelligent cable detection method and system | |
CN116990578A (en) | Direct current detection method and system based on intelligent power | |
CN114336612A (en) | Power loss calculation method for power station and fault equipment thereof and related application equipment |
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 |