Electric energy metering data real-time correction method based on edge calculation

Abstract

The invention relates to the field of data transmission correction, in particular to an electric energy metering data real-time correction method based on edge calculation. The method comprises the steps that after electric energy metering data are collected and preprocessed, an original vector is constructed; performing multi-scale nested function combination transformation on the original vector, and constructing a stable reference parameter vector after multi-scale expansion; based on the original vector and the stable reference parameter vector after multi-scale expansion, constructing a bidirectional drift estimation fitting model to obtain a forward drift trend estimation vector and a backward error estimation vector, and constructing an error behavior vector; and carrying out cross coupling fusion processing on the error behavior vector, carrying out mimicry feedback processing on the error behavior based on the obtained two-dimensional cross offset coupling matrix to obtain an error correction amount, and carrying out real-time correction on the original vector. The problems that a traditional data correction method is insufficient in precision, high in delay, weak in anti-interference capability and difficult to correct data errors in real time in an edge calculation scene are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data transmission correction, and in particular to an electric energy metering data real-time correction method based on edge computing. BACKGROUND

[0002] With the rapid development of smart grid and the comprehensive deployment of power internet of things, electric energy metering devices have gradually evolved from traditional mechanical meters and static electronic meters to smart meter systems with edge intelligent computing capabilities. Modern electric energy metering no longer meets the basic power collection function, but also requires the ability to perform real-time sensing and dynamic analysis on complex load states, power quality, power fluctuation behavior, etc. However, in the actual operating environment, the data collected by the electric energy meter is easily affected by various interference factors, such as power grid voltage fluctuation, load mutation, harmonic interference, electromagnetic interference, and the limitation of the sampling accuracy of the electric meter, resulting in significant deviation and non-linear drift of the collected electric energy metering data, which seriously affects its metering accuracy and the reliability of the back-end data decision system.

[0003] The error correction mechanism currently used in the electric energy metering system mainly relies on the central platform or the background system for batch offline processing. Typical methods include rule-based anomaly detection, static filtering model, clustering method analysis, statistical regression model, and offline training correction method based on machine learning. These technologies can identify and correct abnormal values or drift behavior in electric energy data to some extent, but generally have problems such as insufficient real-time performance, weak model generalization ability, poor robustness to non-stationary disturbances, high consumption of computing resources, and difficulty in deployment on edge devices or field terminals. At the same time, since most algorithm structures rely on data upload to the central node, limited by communication bandwidth, data synchronization delay, etc., it is difficult to meet the real-time requirements of millisecond-level correction of electric energy metering data on site.

[0004] The traditional data correction method also has the problem of insufficient precision, high delay, weak anti-interference ability, and difficulty in real-time correction of data errors in the edge computing scenario. SUMMARY

[0005] The present application provides an electric energy metering data real-time correction method based on edge computing to solve the problem of insufficient precision, high delay, weak anti-interference ability, and difficulty in real-time correction of data errors of traditional data correction methods in the edge computing scenario.

[0006] The electric energy metering data real-time correction method based on edge computing of the present application specifically includes the following technical solutions: An electric energy metering data real-time correction method based on edge computing includes the following steps: S1. Collecting and preprocessing power metering data to obtain preprocessed power metering data, and constructing an original vector; based on the original vector, introducing a time modulation-based multi-scale nested function combination transformation to obtain power metering parameter local trend quantity, and constructing a multi-scale expanded stable benchmark parameter vector; based on the original vector and the multi-scale expanded stable benchmark parameter vector, constructing a bidirectional drift estimation fitting model to obtain a forward drift trend estimation vector and a backward error estimation vector; S2. Based on the forward drift trend estimation vector and the backward error estimation vector, constructing an error behavior vector, and performing cross-coupling fusion processing on the error behavior vector to obtain a two-dimensional cross-offset coupling matrix; based on the two-dimensional cross-offset coupling matrix, performing quasi-state feedback processing on the error behavior to obtain an error correction quantity; based on the error correction quantity, reconstructing the original vector to obtain a corrected vector.

[0007] Preferably, the S1 specifically comprises: The multi-scale nested function combination transformation introduces a modulation term to perform nonlinear transformation on the preprocessed power metering data in the original vector at different frequency scales to obtain the power metering parameter local trend quantity.

[0008] Preferably, the S1 specifically comprises: The multi-scale nested function combination transformation introduces a modulation term to perform nonlinear transformation on the preprocessed power metering data in the original vector at different frequency scales to obtain the power metering parameter local trend quantity.

[0009] Preferably, the S1 specifically comprises: The bidirectional drift estimation fitting model includes a forward drift fitting structure and a backward trace set structure, which are realized by a forward prediction mapping function and a backward drift function, respectively.

[0010] Preferably, the S1 specifically comprises: The forward prediction mapping function introduces a dynamic time decay factor and a logarithmic compression function based on the multi-scale expanded stable benchmark parameter vector to nonlinearly estimate the evolution trend of future power metering data in the historical window to obtain the forward drift trend estimation vector.

[0011] Preferably, the S1 specifically comprises: The backward drift function calculates the difference between the original vector of the current time and the historical time, and introduces an error decay control term to obtain the backward error estimation vector.

[0012] Preferably, the S2 specifically comprises: Based on the error behavior vector, the error difference between different electric energy measurement parameters is calculated, a historical coupling sensitive weight is introduced, cross coupling fusion processing is carried out, and a two-dimensional cross offset coupling matrix is constructed.

[0013] Preferably, S2 specifically includes: In the implementation process of the isomorphism feedback processing, based on the two-dimensional cross offset coupling matrix, a periodic modulation molecular term and an inhibition exponential factor are introduced to calculate the error correction amount; based on the error correction amount, the original vector is reconstructed to generate a corrected vector; the corrected vector includes corrected electric energy measurement data.

[0014] The technical scheme of the application has the following beneficial effects: 1. The multi-scale nested function combination transformation based on time modulation can decompose the original non-stationary and multi-disturbance signal into stable trend components, effectively shield high-frequency random fluctuations and periodic disturbances, realize stable signal extraction, and improve error identification accuracy from the source.

[0015] 2. The bidirectional drift estimation fitting model combining forward prediction and backward rollback is constructed, the data evolution path is evaluated in forward trend modeling, and the difference between the current state and the historical behavior is quantified in backward rollback, thereby forming a time-symmetric error fitting mechanism that not only can judge the existence of error, but also can accurately identify the direction, amplitude and evolution rate of error, greatly enhancing the error behavior capture ability.

[0016] 3. By constructing a two-dimensional cross offset coupling matrix, the synchronous or reverse drift relationship between different electric energy measurement parameters such as voltage, current, power and frequency can be perceived, avoiding missing the overall error trend due to isolated processing, especially in three-phase unbalanced or nonlinear load scenarios, which has important advantages, can capture the mutual influence and feedback between electric energy measurement parameters, and effectively improve the overall stability.

[0017] 4. The residual self-feedback path is established by using bionic feedback path to obtain error correction amount, which not only reduces the calculation overhead, but also better adapts to edge devices with limited computing power, has good embedded deployment capability and running stability. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A flowchart of the electric energy measurement data real-time correction method based on edge computing according to the application. DETAILED DESCRIPTION

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Unless otherwise defined, 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 pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific scheme for a real-time correction method for electricity metering data based on edge computing provided by the present invention.

[0022] See attached document Figure 1 The diagram illustrates a flowchart of a real-time correction method for electricity metering data based on edge computing, provided by an embodiment of the present invention. The method includes the following steps: S1. Collect and preprocess electricity metering data to obtain preprocessed electricity metering data and construct an original vector; based on the original vector, introduce a multi-scale nested function combination transformation based on time modulation to obtain the local trend of electricity metering parameters and construct a stable reference parameter vector after multi-scale expansion; based on the original vector and the stable reference parameter vector after multi-scale expansion, construct a bidirectional drift estimation fitting model to obtain the forward drift trend estimation vector and the backward error estimation vector. Electricity metering data (i.e., electricity metering parameters) is collected by edge devices and preprocessed to obtain preprocessed electricity metering data. The electricity metering data includes voltage, current, active power, reactive power, frequency, etc. The preprocessing process includes data cleaning, time synchronization, noise reduction, standardization and normalization, etc. The methods used are all technical means well known to those skilled in the art and will not be described in detail here. Based on the preprocessed electricity metering data, construct the current time... The original vector ,in, It is in time Pre-processed voltage; It is in time The current after pretreatment; It is in time Preprocessed active power; It is in time Reactive power after preprocessing; It is in time Preprocessed frequency; Indicates transpose; Furthermore, to avoid the preprocessed energy metering parameters in the original vector exhibiting significant non-stationarity and high-frequency disturbance components due to the influence of power frequency fluctuations, grid harmonics, and sudden load changes in the time domain, a multi-scale nested function combination transformation based on time modulation is introduced at the edge device. The core idea of ​​this multi-scale nested function combination transformation is to transform the original vector into a set of stable reference parameter vectors with stable statistical expectations through a combination of nonlinear transformation and frequency modulation. The specific operation is as follows: For the first in the original vector Preprocessed power metering data (such as) Indicates voltage. (representing current, etc.), in The following transformation is performed at each frequency scale:

[0023] in, It is a multi-scale trend component, representing the trend over time. , No. Preprocessed power metering parameters, in the first Local trend of electricity metering parameters at a frequency scale; It is a parameter type index, indicating which type of energy metering data (such as voltage / current) is being processed. It is a frequency scale index, representing the frequency decomposition scale used to construct trends. The range of values ​​is , This represents the total number of frequency scales, which can be set according to specific circumstances and is not limited here; It is the local modulation frequency, representing the first... The sinusoidal modulation frequency corresponding to each frequency scale component is used to extract the periodic trend. This value is determined based on the specific application scenario, with a reference range of values. ; It is the first The time drift exponential factor at each frequency scale is used to control the time decay rate of the exponential function to suppress the influence of long-term historical data. It is determined based on expert experience, with a reference value range of [value missing]. ; It is the first The stationary variation function at each frequency scale is selected, and the compactly supported symmetric function based on the Daubechies wavelet family is used to perform local transformation on the signal to enhance the ability to resist high-frequency disturbances; the cube root operation can ensure the continuity of the nonlinear response and avoid inverse distortion under large amplitude changes. Indicates time Upper Preprocessed power metering data; Is with the first The time delay window index corresponding to each frequency scale represents the integer time delay at that frequency scale. The modulation term is used to dynamically change the frequency scale component over time and enhance the short-term periodic components. Then, the local trends of the energy metering parameters at all frequency scales are accumulated to obtain the multi-scale expanded trend of the energy metering parameters. And construct a stable baseline parameter vector after multi-scale expansion. ; Furthermore, in edge devices, a bidirectional drift estimation fitting model is constructed based on the stable benchmark parameter vector after multi-scale expansion. The bidirectional drift estimation fitting model includes a forward drift fitting structure and a backward backtracking set structure, which are implemented by the forward prediction mapping function and the backward drift function, respectively. The forward prediction mapping function's core function is to perform nonlinear estimation of the evolution trend of future electricity metering parameters within a short-term historical window, obtaining a forward drift trend estimation vector, as shown in the following formula:

[0024] in, It is the forward drift trend estimation vector, representing the time... Through the previous The stable reference parameter vector after multi-scale expansion at each time point is used to model and predict the evolution trend of electricity metering parameters. This indicates the length of the history window used, which is determined based on specific requirements. It is in time (Next) The stable baseline parameter vector after multi-scale expansion (at each time step) It is the square of the L2 norm, representing the time... The squared magnitude of the stable reference parameter vector after multi-scale expansion is used to measure the instantaneous amplitude intensity. It is a logarithmic compression function used to prevent high-amplitude terms from affecting the overall trend estimate, and it has the function of compressing high-frequency interference; It is the next The overfitting suppression factor at each time step was determined based on expert experience, with a reference range of [value missing]. ; It is the next The local modulation frequency for each time step is determined based on the specific application scenario, with a reference range of values. ; It is a dynamic time decay factor used to control the contribution weight of each historical moment in the forward drift trend estimation vector, so as to dynamically adjust the decay rate. The backward drift function estimates the backward error vector by calculating the difference between the original vector at the current time (based on preprocessed electricity metering data) and the original vector at historical times, as shown in the following formula:

[0025] in, It is in time The backward error estimation vector; This refers to the size of the backward history window, which can be set according to the specific scenario and is not limited here. It is a vector difference, representing the value of the vector difference from the next position. The change in electricity metering data at each time step , It is the next The original vectors at each time step; It is the next The original vectors at each time step; It is a non-linear excitation term used to scale the impact of historical mutations; It is an error decay control term, used to suppress the error explosion caused by large drift; A saturation behavior suppression mechanism is used to construct a drift suppression mechanism to suppress the interference of extreme historical values ​​on the fit and keep the output stable.

[0026] S2. Based on the forward drift trend estimation vector and the backward error estimation vector, construct the error behavior vector, and perform cross-coupling fusion processing on the error behavior vector to obtain a two-dimensional cross-offset coupling matrix; based on the two-dimensional cross-offset coupling matrix, perform mimicry feedback processing on the error behavior to obtain the error correction amount; based on the error correction amount, reconstruct the original vector to obtain the corrected vector.

[0027] The error behavior vector is defined by calculating the difference between the forward drift trend estimation vector and the backward error estimation vector. This variable, used to characterize the multidimensional temporal inconsistency of the current error, is the core mediating variable for subsequent residual construction. To describe the coupling behavior between multidimensional errors, the error behavior vectors are cross-coupled and fused to construct a two-dimensional cross-offset coupling matrix, which is used to perceive the offset coupling strength between different energy metering parameters. The two-dimensional cross-offset coupling matrix is ​​calculated in the following form:

[0028] in, It is in time The two-dimensional cross-offset coupling matrix constructed between all preprocessed energy metering parameters is used to characterize the first... Preprocessed power metering parameters and the first The offset coupling strength between the preprocessed energy metering parameters; It is the first Preprocessed power metering parameters in time The error behavior index is taken from the error behavior vector. ; It is the first Preprocessed power metering parameters in time The error behavior index is taken from the error behavior vector. ; It is the first Preprocessed power metering parameters and the first The historical coupling sensitivity weights among the preprocessed electricity metering parameters are determined based on expert experience, with a reference range of values. ; The difference index is used to map the error difference to a contrast suppression term to enhance the penalty for “inconsistent parameter drift direction” so that the coupled value is not dominated by a single outlier. Indicates the error difference; This represents the total number of types of electricity metering parameters; Based on a two-dimensional cross-offset coupling matrix, a mimicry feedback processing method is applied to the error behavior. This mimicry feedback processing does not employ traditional loss functions and backpropagation mechanisms. Instead, it is based on a biomimetic neural mimicry mechanism, a path modulation model, and nonlinear dynamic path feedback control theory. A residual self-feedback path is established using a biomimetic feedback path to obtain the error correction amount. The specific formula is as follows:

[0029] in, It is in time Error correction amount; These are the mimicry feedback path numbers, totaling [number]. A parallel feedback mimicry channel Determined based on the specific application scenario requirements; Indicates the first The periodic response frequency control parameters of the mimicry feedback path are determined based on expert experience, with a reference range of values. ; Indicates the first The path offset phase constant of the mimicry feedback path is determined by simulating time phase misalignment and based on expert experience, with a reference range of values. ; It is a frequency factor, representing the frequency basis of the mimicry feedback. It is determined according to the specific application scenario requirements and will not be elaborated here. It is the first The inhibition index factor of the mimicry feedback path, representing the path inhibition strength (exponential steepness), is determined based on expert experience, with a reference value range of [value missing]. ; It is a path fusion sub-block, representing time. No. The sub-regions (such as the main diagonal block, upper and lower triangular regions, corner interaction regions, and the entire matrix) of the two-dimensional cross-offset coupling matrix received by the mimicry feedback path originate from the two-dimensional cross-offset coupling matrix. The partitions are determined based on expert experience. It is the Frobenius norm; It is the natural logarithm term, used for the response magnitude of the nonlinearly compressed two-dimensional cross-offset coupling matrix; It is the periodic modulation numerator, representing periodic regulated fluctuations; It is an exponential channel activation function used to control the nonlinear slope of the activation response of the mimicry feedback path; Finally, based on the error correction amount, the original vector is reconstructed using a Bayesian inference fusion mechanism to generate a corrected vector, which includes the corrected electricity metering data, thus achieving real-time correction of the electricity metering data. The Bayesian inference fusion mechanism is a well-known technique in the art and will not be described in detail here.

[0030] In summary, a real-time correction method for electricity metering data based on edge computing has been developed.

[0031] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0032] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0033] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A real-time correction method for electricity metering data based on edge computing, characterized in that, Includes the following steps: S1. Collect and preprocess electricity metering data to obtain preprocessed electricity metering data and construct an original vector; based on the original vector, introduce a multi-scale nested function combination transformation based on time modulation to obtain the local trend of electricity metering parameters and construct a stable reference parameter vector after multi-scale expansion; based on the original vector and the stable reference parameter vector after multi-scale expansion, construct a bidirectional drift estimation fitting model to obtain the forward drift trend estimation vector and the backward error estimation vector. S2. Based on the forward drift trend estimation vector and the backward error estimation vector, an error behavior vector is constructed, and the error behavior vector is cross-coupled and fused to obtain a two-dimensional cross-offset coupling matrix; based on the two-dimensional cross-offset coupling matrix, the error behavior is subjected to mimicry feedback processing to obtain the error correction amount; The original vector is reconstructed based on the error correction amount to obtain the corrected vector.

2. The real-time correction method for electricity metering data based on edge computing according to claim 1, characterized in that, S1 specifically includes: The multi-scale nested function combination transformation introduces a modulation term to perform nonlinear transformation on the preprocessed power metering data in the original vector at different frequency scales, thereby obtaining the local trend quantity of the power metering parameters.

3. The real-time correction method for electricity metering data based on edge computing according to claim 2, characterized in that, S1 specifically includes: The local trends of the power metering parameters at all frequency scales are accumulated to obtain the multi-scale expanded power metering parameter trends, and a stable reference parameter vector after multi-scale expansion is constructed.

4. The real-time correction method for electricity metering data based on edge computing according to claim 3, characterized in that, S1 specifically includes: The bidirectional drift estimation fitting model includes a forward drift fitting structure and a backward backtracking set structure, which are implemented through a forward prediction mapping function and a backward drift function, respectively.

5. The real-time correction method for electricity metering data based on edge computing according to claim 4, characterized in that, S1 specifically includes: The forward prediction mapping function, based on the stable reference parameter vector after multi-scale expansion, introduces a dynamic time decay factor and a logarithmic compression function to nonlinearly estimate the evolution trend of future electricity metering data within a historical window, thereby obtaining a forward drift trend estimation vector.

6. The real-time correction method for electricity metering data based on edge computing according to claim 5, characterized in that, S1 specifically includes: The backward drift function calculates the difference between the original vectors of the current time and the historical time, and introduces an error attenuation control term to obtain the backward error estimation vector.

7. The real-time correction method for electricity metering data based on edge computing according to claim 1, characterized in that, S2 specifically includes: Based on the error behavior vector, the error difference between different power metering parameters is calculated, and historical coupling sensitive weights are introduced to perform cross-coupling fusion processing to construct a two-dimensional cross-offset coupling matrix.

8. A real-time correction method for electricity metering data based on edge computing according to claim 7, characterized in that, S2 specifically includes: In the implementation of the mimicry feedback processing, based on the two-dimensional cross-offset coupling matrix, a periodic modulation numerator and a suppression exponent factor are introduced to calculate the error correction amount; based on the error correction amount, the original vector is reconstructed to generate a corrected vector; the corrected vector includes the corrected power metering data.

Images