Electric energy meter anomaly detection method based on sliding window model

Through the combination of the sliding window model and the step-down module, real-time abnormality detection of the power meter is achieved, solving the problem that traditional methods cannot detect in real-time, and improving the accuracy of power metering and the reliability of equipment.

CN118914960AInactive Publication Date: 2024-11-08WUXI HENGTONG ELECTRIC CO LTD

Patent Information

Application Number
CN202411152573.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional power meter abnormality detection relies on manual regular inspections and cannot provide real-time abnormality detection, resulting in inaccurate power metering and equipment failure, affecting the normal operation of power methods.

Method used

Using the power meter abnormality detection method based on the sliding window model, collect power supply data, build a time series, extract status information using the sliding window method, calculate the temperature change rate and edge weight, combine the step-down module to judge and control outlier value, and detect in real time Abnormal behavior of the electricity meter.

Benefits of technology

Real-time abnormality detection and fault positioning of the power meter are realized, the ability to predict the state of the power meter is enhanced, and the accuracy of power metering and equipment safety are ensured.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118914960A_ABST
    Figure CN118914960A_ABST
Patent Text Reader

Abstract

The invention provides an electric energy meter anomaly detection method based on a sliding window model. The method comprises the steps of 1, collecting input and output power data of an electric energy meter; 2, constructing a time sequence according to the change of the collected power supply data; step 3, extracting state information of a current moment and a previous moment in the time sequence by using a sliding window method; step 4, intercepting the sub-sequence through a sliding window, and searching key points to reconstruct the sub-sequence to obtain a change rule of the power data of the electric energy meter in one operation cycle; abnormal behaviors of the electric energy meter are identified by collecting various input and output data of the electric energy meter and utilizing a time sequence analysis and sliding window method; through an abnormal value judgment and voltage reduction module monitoring mode, the abnormal condition of the electric energy meter can be effectively detected, and corresponding protection measures are taken. In addition, through calculation of the temperature change rate and the edge weight, the prediction capability of the sliding window on the state of the electric energy meter is further enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to an abnormality detection method for an electric energy meter based on a sliding window model, and in particular to the field of abnormality detection for an electric energy meter. Background Art

[0002] In modern electricity methods, energy meters, as an important metering device, are widely used in home, commercial and industrial environments to measure and record electricity consumption.

[0003] With the development of smart grids and the complexity of power methods, ensuring the accuracy and reliability of electricity meters has become particularly important; however, due to various reasons (such as equipment aging, environmental changes or malicious attacks), electricity meters may exhibit abnormal behavior, resulting in inaccurate measurement or equipment failure, which in turn affects the normal operation of power methods; traditional electricity meter anomaly detection mainly relies on manual regular inspections and cannot provide real-time anomaly detection. Summary of the invention

[0004] Purpose of the invention: To propose an abnormality detection method for electric energy meters based on a sliding window model to solve the above problems existing in the prior art;

[0005] Technical solution: A method for detecting abnormality of electric energy meters based on a sliding window model, comprising:

[0006] Step 1: Collect the input and output power data of the energy meter, including but not limited to temperature, voltage, current, energy meter reading, power supply parameters and user power consumption data;

[0007] Step 2: construct a time series based on the changes in the collected power supply data;

[0008] Step 3: Use the sliding window method to extract the status information of the current time and the previous time in the time series; set the window width to the average peak time of the pseudo-periodic data, and the step length to 1; after the time series with a length of T1 is divided by the sliding window method, T1-window width+1 subsequences are obtained;

[0009] Step 4: intercept subsequences through sliding windows, find key points to reconstruct subsequences, obtain the change pattern of power data within an operation cycle of the electric energy meter, and perform anomaly detection on the change pattern.

[0010] In a further embodiment, the temperature feature is extracted from the power supply data, and the cumulative sum of the differences of consecutive temperature readings of the power supply data is calculated according to the temperature change rate, and then divided by the window size to obtain the average temperature change rate;

[0011]

[0012] Where ΔT represents the average temperature change rate; T t represents the temperature at time t; T t-1 represents the temperature at time t-1; p represents the time span for calculating the average rate of change within the power supply data.

[0013] In a further embodiment, multiple window data are integrated. When window 1 contains data for a complete operating cycle, and the first half of the data in window 2 contains data from the previous day, the data in window 1 and window 2 are combined into complete cycle operating data, retaining the phase position of the power data on that day and the power data information.

[0014] In a further embodiment, the reconstructed value of the time series data at time t is The absolute value of the difference from the actual value x is defined as the error;

[0015]

[0016] Reconstruct the data of the subsequence obtained in the sliding window, and take the latest data point to calculate the current reconstruction error to obtain the reconstruction error curve;

[0017] The abnormal point is located by comparing the error value with the preset threshold; the specific point where the abnormality occurs is obtained.

[0018] In a further embodiment, in the time series, the weight of the edge is determined according to the current size of the node; Fault detection and location abnormal weight: if the weight of an edge changes, it means that a fault has occurred at the node or connection.

[0019] By tracking the weight changes of the edges, we can locate the problematic part of the electric energy meter. The expression is as follows:

[0020] Wij=log(1+avg(It,ij))

[0021] Where Wij represents the weight of the edge between node i and node j; avg(It,ij) represents the average value of the current between node i and node j within time t.

[0022] In a further embodiment, the abnormal value judgment in the power supply data is performed using the moving average value and the standard deviation; the specific expression is as follows:

[0023] MA=avg(Xt-n to Xt-1)

[0024] SD=stddev(Xt-n to Xt-1)

[0025] Outlier judgment: |Xt-MA|>3*SD

[0026] In the formula, MA represents the moving average, SD represents the standard deviation; Xt represents the data value of the current point; n represents the size of the moving window, which is used to identify points with significantly different data within the window period.

[0027] In a further embodiment, the voltage reduction module controls the electric energy meter based on the determination of abnormal values;

[0028] The step-down module includes field effect transistor S1 and field effect transistor S2, inductor L and capacitor C. By detecting the peak current of inductor L, it is judged whether the working current of the energy meter exceeds the protection current of the components.

[0029] When the FET S1 is turned on and the FET S2 is turned off, Vi supplies power to the load and charges the inductor L at the same time;

[0030] When FET S1 is disconnected and FET S2 is turned on, L discharges;

[0031] The voltage difference between the input voltage Vi and the switch node voltage VS of the energy meter is collected, and the collected current Is1 of the energy meter is obtained through conversion by the operational amplifier. The collected current Is1 is isolated by the current mirror to obtain the collected current Is1=Is2, and the output is converted by the resistor R to obtain the output voltage Vs1.

[0032] By controlling the transistor Q1 to adjust the conduction number K of the transistor Q2 and the transistor Q3, the internal resistance of the transmission path is adjusted to obtain different reference voltages Vr;

[0033] The output voltage Vs and the reference voltage Vr are input into the current threshold comparator for comparison; when overcurrent occurs, Vs is greater than Vr, causing the current threshold comparator to flip, and after the logic controller controls transistor S1 to be disconnected and transistor S2 to be turned on, and the Vi input is cut off, the inductor L continues to flow, and the output voltage will gradually decrease.

[0034] Beneficial effects: The present invention proposes a method for detecting abnormalities in electric energy meters based on a sliding window model. The method collects various input and output data of the electric energy meter and uses time series analysis and sliding window methods to identify abnormal behaviors of the electric energy meter. Through abnormal value judgment and step-down module monitoring, the abnormal conditions of the electric energy meter can be effectively detected and corresponding protective measures can be taken. In addition, through the calculation of the temperature change rate and edge weight, the sliding window's prediction ability for the electric energy meter state is further enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the process of the present invention.

[0036] Figure 2 This is a circuit diagram of the step-down module of the present invention. DETAILED DESCRIPTION

[0037] The applicant believes that with the development of smart grids and the complexity of power methods, it becomes particularly important to ensure the accuracy and reliability of electricity meters; however, due to various reasons (such as equipment aging, environmental changes or malicious attacks), electricity meters may exhibit abnormal behavior, resulting in inaccurate measurement or equipment failure, which in turn affects the normal operation of power methods; traditional electricity meter abnormality detection mainly relies on manual regular inspections and cannot provide real-time abnormality detection.

[0038] In order to solve the problems existing in the prior art, the present invention realizes detection of the operating status of the electric energy meter and ensures the safety of the electric energy meter through an electric energy meter abnormality detection method based on a sliding window model.

[0039] The present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0040] In this application, we propose a method for detecting anomalies in electric energy meters based on a sliding window model, which includes the following steps:

[0041] Step 1: Collect the input and output power data of the electric energy meter, the power data including but not limited to temperature, voltage, current, electric energy meter reading, power supply parameters and user power consumption data; extract temperature features from the power data, and calculate the cumulative sum of the difference of consecutive temperature readings of the power data according to the temperature change rate, and then divide it by the window size to obtain the average temperature change rate;

[0042]

[0043] Where ΔT represents the average temperature change rate; T t represents the temperature at time t; T t-1 represents the temperature at time t-1; p represents the time span for calculating the average rate of change within the power supply data.

[0044] Step 2: construct a time series based on the changes in the collected power supply data;

[0045] Step 3: Use the sliding window method to extract the status information of the current moment and the previous moment in the time series; set the window width to the average peak time of the pseudo-periodic data, with a step size of 1; divide the time series with a length of T1 by the sliding window method to obtain T1-window width + 1 subsequences; integrate multiple window data, when window 1 contains data of a complete operating cycle, and the first half of the data in window 2 contains data from the previous day, the data in window 1 and window 2 are combined into complete cycle operating data, retaining the phase position of the power data on that day and the power data information.

[0046] Step 4: intercept subsequences through sliding windows, find key points to reconstruct subsequences, obtain the change pattern of power data within an operation cycle of the electric energy meter, and perform anomaly detection on the change pattern.

[0047] The reconstructed value of the time series data at time t The absolute value of the difference from the actual value x is defined as the error;

[0048]

[0049] Reconstruct the data of the subsequence obtained in the sliding window, and take the latest data point to calculate the current reconstruction error to obtain the reconstruction error curve;

[0050] The abnormal point is located by comparing the error value with the preset threshold; the specific point where the abnormality occurs is obtained.

[0051] In the time series, the weight of the edge is determined according to the current size of the node; Fault detection and location abnormal weight: If the weight of an edge changes (such as a sharp increase or decrease), it means that a fault has occurred at the node or connection.

[0052] By tracking the weight changes of the edges, we can locate the problematic part of the electric energy meter. The expression is as follows:

[0053] Wij=log(1+avg(It,ij))

[0054] Where Wij represents the weight of the edge between node i and node j; avg(It,ij) represents the average value of the current between node i and node j within time t.

[0055] To judge the outliers in the power supply data, the moving average and standard deviation are used for outlier judgment; the specific expression is as follows:

[0056] MA=avg(Xt-n to Xt-1)

[0057] SD=stddev(Xt-n to Xt-1)

[0058] Outlier judgment: |Xt-MA|>3*SD

[0059] In the formula, MA represents the moving average, SD represents the standard deviation; Xt represents the data value of the current point; n represents the size of the moving window, which is used to identify points with significantly different data within the window period.

[0060] The step-down module controls the energy meter based on the abnormal value judgment;

[0061] The step-down module includes field effect transistor S1 and field effect transistor S2, inductor L and capacitor C. By detecting the peak current of inductor L, it is judged whether the working current of the energy meter exceeds the protection current of the components.

[0062] When the FET S1 is turned on and the FET S2 is turned off, Vi supplies power to the load and charges the inductor L at the same time;

[0063] When FET S1 is disconnected and FET S2 is turned on, L discharges;

[0064] The voltage difference between the input voltage Vi and the switch node voltage VS of the energy meter is collected, and the collected current Is1 of the energy meter is obtained through conversion by the operational amplifier. The collected current Is1 is isolated by the current mirror to obtain the collected current Is1=Is2, and the output is converted by the resistor R to obtain the output voltage Vs1.

[0065] By controlling the transistor Q1 to adjust the conduction number K of the transistor Q2 and the transistor Q3, the internal resistance of the transmission path is adjusted to obtain different reference voltages Vr;

[0066] The output voltage Vs and the reference voltage Vr are input into the current threshold comparator for comparison; when overcurrent occurs, Vs is greater than Vr, causing the current threshold comparator to flip, and the logic controller controls the transistor S1 to be disconnected, the transistor S2 to be turned on, and after the Vi input is cut off, the inductor L continues to flow, and the output voltage will gradually decrease to protect the energy meter and the output of the energy meter.

[0067] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes may be made to it in form and detail without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A method for detecting abnormality of electric energy meters based on a sliding window model, characterized in that: The steps include: Step 1: Collect the input and output power data of the energy meter, including but not limited to temperature, voltage, current, energy meter reading, power supply parameters and user power consumption data; Step 2: construct a time series based on the changes in the collected power supply data; Step 3: Use the sliding window method to extract the status information of the current time and the previous time in the time series; set the window width to the average peak time of the pseudo-periodic data, and the step length to 1; after the time series with a length of T1 is divided by the sliding window method, T1-window width+1 subsequences are obtained; Step 4: intercept subsequences through sliding windows, find key points to reconstruct subsequences, obtain the change pattern of power data within an operation cycle of the electric energy meter, and perform anomaly detection on the change pattern.

2. The method for detecting abnormality of electric energy meters based on a sliding window model according to claim 1, characterized in that: Extract temperature features from the power data, and calculate the cumulative sum of the differences of consecutive temperature readings of the power data according to the temperature change rate, and then divide it by the window size to obtain the average temperature change rate; Where ΔT represents the average temperature change rate; T t represents the temperature at time t; T t-1 represents the temperature at time t-1; p represents the time span for calculating the average rate of change within the power supply data.

3. The method for detecting abnormality of electric energy meters based on a sliding window model according to claim 1, characterized in that: Integrate data from multiple windows. When window 1 contains data from a complete operating cycle, and the first half of the data in window 2 contains data from the previous day, the data from window 1 and window 2 are combined into complete cycle operating data, retaining the phase position of the power data on that day and the power data information.

4. The method for detecting abnormality of electric energy meters based on a sliding window model according to claim 1, characterized in that: The reconstructed value of the time series data at time t The absolute value of the difference from the actual value x is defined as the error; Reconstruct the data of the subsequence obtained in the sliding window, and take the latest data point to calculate the current reconstruction error to obtain the reconstruction error curve; The abnormal point is located by comparing the error value with the preset threshold; the specific point where the abnormality occurs is obtained.

5. The method for detecting abnormality of electric energy meters based on a sliding window model according to claim 1, characterized in that: In the time series, the weight of the edge is determined according to the current size of the node; Fault Detection and Localization Abnormal Weight: If the weight of an edge changes, it means that a fault has occurred at the node or connection. By tracking the weight changes of the edges, we can locate the problematic part of the electric energy meter. The expression is as follows: Wij=log(1+avg(It,ij)) Where Wij represents the weight of the edge between node i and node j; avg(It,ij) represents the average value of the current between node i and node j within time t.

6. The method for detecting abnormality of electric energy meters based on a sliding window model according to claim 1, characterized in that: To judge the outliers in the power supply data, the moving average and standard deviation are used for outlier judgment; the specific expression is as follows: MA=avg(Xt-ntoXt-1) SD=stddev(Xt-ntoXt-1) Outlier judgment: |Xt-MA|>3*SD In the formula, MA represents the moving average, SD represents the standard deviation; Xt represents the data value of the current point; n represents the size of the moving window, which is used to identify points with significantly different data within the window period.

7. The method for detecting abnormality of electric energy meters based on a sliding window model according to claim 1, characterized in that: The step-down module controls the energy meter based on the abnormal value judgment; The step-down module includes field effect transistor S1 and field effect transistor S2, inductor L and capacitor C. By detecting the peak current of inductor L, it is judged whether the working current of the energy meter exceeds the protection current of the components. When the FET S1 is turned on and the FET S2 is turned off, Vi supplies power to the load and charges the inductor L at the same time; When FET S1 is disconnected and FET S2 is turned on, L discharges; The voltage difference between the input voltage Vi and the switch node voltage VS of the energy meter is collected, and the collected current Is1 of the energy meter is obtained through conversion by the operational amplifier. The collected current Is1 is isolated by the current mirror to obtain the collected current Is1=Is2, and the output is converted by the resistor R to obtain the output voltage Vs1. By controlling the transistor Q1 to adjust the conduction number K of the transistor Q2 and the transistor Q3, the internal resistance of the transmission path is adjusted to obtain different reference voltages Vr; The output voltage Vs and the reference voltage Vr are input into the current threshold comparator for comparison; when overcurrent occurs, Vs is greater than Vr, causing the current threshold comparator to flip, and after the logic controller controls transistor S1 to be disconnected and transistor S2 to be turned on, and the Vi input is cut off, the inductor L continues to flow, and the output voltage will gradually decrease.

Citation Information

Patent Citations

  • Gateway electric energy meter data anomaly detection method and system based on pseudo abnormal point identification

    CN112213687A

  • Sliding window-based high-frequency acquisition abnormity resistant electric energy meter accurate positioning method

    CN113484818A

  • Methods, devices and computer storage media for anomaly detection

    CN114072773A

  • Method and system for identifying false data in power grid

    CN114861774A

  • Multivariable time sequence anomaly detection method and system based on graph neural network

    CN114881115A

Cited By

  • Electric energy meter operation control method and system

    CN119202617A

  • A method and system for controlling electric energy meter operation

    CN119202617B

  • Distributed storage control method and system for computer big data

    CN120353407A

  • Vehicle-mounted starting power supply abnormal state detection method based on data driving

    CN120468712A

  • Intelligent measuring switch with electric energy error monitoring function

    CN121069303A