CN115841278B - Evaluation method, system, equipment and medium for operating error state of electric energy metering device - Google Patents

Abstract

The invention relates to a method, system, equipment and medium for evaluating the operation error state of an electric energy metering device, wherein the method includes the following steps: collecting historical operation monitoring data of several electric energy metering devices, each historical operation monitoring data includes Generate multiple different types of physical quantities; perform feature extraction on each historical operation monitoring data collected, including two steps of constructing features and feature selection, to obtain the final feature set; add errors to the final feature set of each historical operation monitoring data The state label forms a training sample, constructs a neural network model, trains the neural network model through the training sample to obtain an evaluation model of the operating error state of the electric energy metering device; obtains the current final feature set of the target electric energy metering device, and inputs it into the trained electric energy The error state evaluation model of the metering device operation obtains the error state evaluation results of the target electric energy metering device.

Description

Method, system, equipment and medium for evaluating operation error state of electric energy metering device

Technical Field

The invention relates to a method, system, equipment and medium for evaluating the operating error state of an electric energy metering device, and belongs to the technical field of power metering device state evaluation.

Background Art

Electric energy metering devices are important equipment for measuring electric energy. They are widely used in all aspects of the power system and can provide a perfect basis for power supply fee calculation, electricity measurement, etc. Due to various types of factors and risks, such as wiring errors, poor communication, parameter setting errors, etc., the normal use of electric energy metering devices will be affected. The above or other faults or defects will cause serious problems for the normal operation of the electric energy metering device, and ultimately affect the accuracy of the electric energy metering device. Status evaluation of electric energy metering devices is an important means of studying electric energy metering devices. Accurate status evaluation results can clearly reflect the status of the electric energy metering device during operation, and realize the transformation of the company's electric energy metering devices from traditional periodic control to future intelligent control.

The prior art patent with the publication number of "CN114065605A" discloses a system and method for detecting and evaluating the operation status of a smart electric energy meter. The method comprises the following steps: S1: obtaining multiple error state data of the smart electric energy meter, and recording them as the historical data of the smart electric energy meter; S2: performing quantization and normalization processing on the multiple error state data respectively; S3: performing normalization evaluation weighting on the normalized data, and performing state evaluation according to a preset threshold, and recording the evaluation result as performance degradation data; S4: performing data preprocessing on the performance degradation data and the historical data of the smart electric energy meter to obtain training data; S5: establishing a detection and evaluation model, and training the model with the training data to obtain the optimal detection and evaluation model; S6: substituting the error state data of the smart electric energy meter to be detected into the optimal detection and evaluation model to obtain the detection and evaluation result of the operation status of the smart electric energy meter to be detected. The prior art can scientifically evaluate the operation status of the smart electric energy meter and the prediction of performance degradation failure.

The evaluation index of the above-mentioned prior art, that is, the error status data of the electric energy meter, adopts a large amount of human intervention data including meter type selection error, basic error, operation error, operation time error, operation failure rate error, operation monitoring event error, operation monitoring abnormality error, error dispersion error, full inspection return rate error, operation quality sampling error, lead seal status error, installation environment error and user reputation error, which leads to the error status evaluation of the electric energy meter being strongly correlated with human intervention and not being objective and reliable enough.

Summary of the invention

In order to solve the above problems existing in the prior art, the present invention proposes a method, system, equipment and medium for evaluating the operating error state of an electric energy metering device.

The technical solution of the present invention is as follows:

In one aspect, the present invention provides a method for evaluating an operation error state of an electric energy metering device, comprising the following steps:

Collecting historical operation monitoring data of a number of electric energy metering devices, each historical operation monitoring data includes a plurality of different types of physical quantities generated during the operation of the electric energy metering device;

Perform feature extraction on each collected historical operation monitoring data, including two steps: feature construction and feature selection;

The step of constructing features is specifically as follows: for each historical operation monitoring data, the kurtosis sorting energy feature, the curve complexity feature and the phase cosine similarity feature corresponding to each physical quantity are calculated by various types of physical quantities;

The feature selection step specifically includes: for any historical operation monitoring data, screening the set of features obtained in the feature construction step to obtain a final feature set;

Add error state labels to the final feature set of each historical operation monitoring data to form training samples, build a neural network model, iteratively train the neural network model through the training samples, and obtain a trained electric energy metering device operation error state evaluation model;

The current operation monitoring data of the target electric energy metering device is obtained, and feature extraction is performed according to the current operation monitoring data to obtain the current final feature set of the target electric energy metering device, which is input into the trained electric energy metering device operation error state evaluation model to obtain the error state evaluation result of the target electric energy metering device.

As a preferred implementation, the multiple different types of physical quantities generated during the operation of the electric energy metering device specifically include:

Voltage data, current data, active power data, reactive power data, power factor data, three-phase imbalance data and load rate data.

As a preferred implementation, in the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and inter-phase cosine similarity characteristics corresponding to each physical quantity through various types of physical quantities, the calculation method of the kurtosis sorting energy characteristics of each physical quantity is:

Draw corresponding physical quantity curves according to various types of physical quantities, including current curve, voltage curve, active power curve, reactive power curve, power factor curve, three-phase unbalance curve, and load rate curve;

For each physical quantity curve, the kurtosis reflects the numerical statistics of the random variable distribution characteristics of each physical quantity curve. The expression of the kurtosis K is as follows:

;

Where N represents the signal length of the physical quantity curve; represents the i-th signal value in the physical quantity curve; μ represents the signal average value of the physical quantity curve; σ represents the signal standard deviation of the physical quantity curve;

For each signal in each physical quantity curve, perform EMD empirical mode decomposition, calculate the kurtosis of the decomposed signal and sort it in descending order according to the calculated kurtosis, and then calculate the kurtosis sorting energy E according to the following formula:

;

Thus, the current kurtosis sorting energy characteristics, voltage kurtosis sorting energy characteristics, active power kurtosis sorting energy characteristics, reactive power kurtosis sorting energy characteristics, power factor kurtosis sorting energy characteristics, three-phase imbalance kurtosis sorting energy characteristics, and load rate kurtosis sorting energy characteristics are obtained.

As a preferred implementation, in the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and inter-phase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the curve complexity characteristics of each physical quantity is:

Define fuzzy entropy as:

;

Among them, m is the dimension of phase space, r is the similarity tolerance, N is the dimension of time series, is the fuzzy membership function;

According to the defined fuzzy entropy, the curve complexity characteristics are calculated respectively , the second voltage curve complexity characteristics , Complexity characteristics of the second active power curve , Complexity characteristics of the second reactive power curve , the second power factor curve complexity order characteristics 2. Complexity characteristics of the third-phase imbalance curve , Complexity characteristics of the second load rate curve They are:

;

;

;

;

;

;

in, , , Respectively represent the A phase, B phase, and C phase currents; , , They are phase A, phase B, and phase C voltages respectively; , , , Respectively represent the active power of phase A, phase B, phase C and the total; , , , Respectively represent the reactive power of phase A, phase B, phase C and the total; , , , Respectively represent the current power factors of phase A, phase B, phase C and the total; It is the three-phase unbalance data;

R _{load factor} indicates the load factor.

As a preferred implementation, in the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and interphase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the interphase cosine similarity characteristics of each physical quantity is:

According to the cosine similarity formula, the cosine similarity characteristics of phase A and phase B current are calculated respectively. , B-phase and C-phase current cosine similarity characteristics , A phase and C phase current cosine similarity characteristics , A-phase and B-phase voltage cosine similarity features , B-phase and C-phase voltage cosine similarity features , A phase and C phase voltage cosine similarity characteristics , A phase and B phase active power cosine similarity characteristics , B phase and C phase active power cosine similarity characteristics , A phase and C phase active power cosine similarity characteristics , A-phase and B-phase reactive power cosine similarity characteristics , B-phase and C-phase reactive power cosine similarity characteristics , A phase and C phase reactive power cosine similarity characteristics , A phase and B phase power factor cosine similarity characteristics , B phase and C phase power factor cosine similarity characteristics , A phase and C phase power factor cosine similarity characteristics ;

Among them, the cosine similarity formula is:

;

Where X, Y represent X and Y vectors; , Represents the magnitude of the X and Y vectors respectively.

As a preferred implementation, the method of screening the set of features obtained in the feature construction step to obtain the final feature set is specifically:

The kurtosis sorting energy features of each physical quantity are used as the first feature set, the curve complexity features of each physical quantity are used as the second feature set, the inter-phase cosine similarity features of each physical quantity are used as the third feature set, and the first feature set, the second feature set and the third feature set are combined into a fourth feature set;

Use the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm to perform feature selection on the fourth feature set respectively to obtain n feature subsets;

Merge n feature subsets into the fifth feature set;

Use the n+1th feature selection algorithm to perform feature selection on the fifth feature set to obtain a sixth feature set, and use the sixth feature set as the final feature set;

Among them, the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm and the n+1th feature selection algorithm are specifically selected from variance selection method, correlation coefficient method, chi-square test method, relief algorithm, recursive feature elimination method, penalty-based feature selection method and tree model-based feature selection method.

As a preferred implementation, the neural network model adopts a nonlinear time series prediction NARX neural network model, and the Golden Eagle algorithm is used to optimize the neural network model. The specific steps are:

Initialize the number of golden eagle individuals in the golden eagle population. Each golden eagle contains different weight coefficient information and threshold parameter information of the NARX neural network model.

Calculate the fitness value of the golden eagle individual and initialize the group memory according to the fitness value; the fitness value calculation formula of the golden eagle individual is:

;

Among them, value represents the fitness value, N is the total number of golden eagle individuals, is the predicted value of the NARX neural network model obtained according to the weight coefficient information and threshold parameter information of the NARX neural network model contained in the i-th Golden Eagle, is the true value of the sample corresponding to the i-th golden eagle;

Initialize the Golden Eagle's attack tendency and cruising tendency ;

Update the attack tendency according to the following formula and cruising tendency :

;

in, and for The initial and final values of and for The initial and final values of ; t is the tth iteration, and T is the total number of iterations.

Prey is randomly selected from the attack vector calculated from the memory of the population:

;

in, is the attack vector of the i-th golden eagle, This is the best place for the Golden Eagle to reach at present. is the current position of the i-th golden eagle;

Calculate the cruise vector d:

;

in, is the normal vector, is the decision variable vector;

set up is any point on the hyperplane, then: ;

Bundle As the normal of the hyperplane, the hyperplane is expressed as:

;

in, is the attack vector, is the decision vector, The location of the selected prey; represents the attack vector at the tth iteration;

but:

;

in, is the kth element of the target point, k is the number of the fixed variable;

The destination point on the cruise hyperplane is expressed as:

The step vector of the Golden Eagle iteration is defined as:

;

in, and is a random vector in [0,1]. The position of the golden eagle in the iteration is calculated by adding the step vector in the iteration to the position in the iteration:

in, For the Golden Eagle Second position, For the Golden Eagle The second position, is the step size of the Golden Eagle's movement;

Calculate the fitness value according to the updated position of the new golden eagle individual, and update the optimal solution and optimal position;

Determine whether the maximum number of iterations has been reached. If the maximum number of iterations has been reached, the weight coefficient information and threshold parameter information contained in the golden eagle individual at the current optimal position are output as the weight coefficient and threshold parameter of the NARX neural network model; if the maximum number of iterations has not been reached, continue to iterate.

On the other hand, the present invention also provides an electric energy metering device operation error state evaluation system, comprising:

A collection module, used to collect historical operation monitoring data of a number of electric energy metering devices, each historical operation monitoring data includes a plurality of different types of physical quantities generated during the operation of the electric energy metering device;

The feature extraction module is used to extract features from each collected historical operation monitoring data, including two steps: feature construction and feature selection;

The step of constructing features is specifically as follows: for each historical operation monitoring data, the kurtosis sorting energy feature, the curve complexity feature and the phase cosine similarity feature corresponding to each physical quantity are calculated by various types of physical quantities;

The feature selection step specifically includes: for any historical operation monitoring data, screening the set of features obtained in the feature construction step to obtain a final feature set;

A model training module is used to add an error state label to the final feature set of each historical operation monitoring data to form a training sample, construct a neural network model, and iteratively train the neural network model through the training sample to obtain a trained electric energy metering device operation error state evaluation model;

The evaluation module is used to obtain the current operation monitoring data of the target electric energy metering device, perform feature extraction based on the current operation monitoring data to obtain the current final feature set of the target electric energy metering device, and input it into the trained electric energy metering device operation error state evaluation model to obtain the error state evaluation result of the target electric energy metering device.

On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for evaluating the operating error state of an electric energy metering device as described in any embodiment of the present invention is implemented.

On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for evaluating the operating error state of an electric energy metering device as described in any embodiment of the present invention.

The present invention has the following beneficial effects:

1. The present invention provides an electric energy metering device operating error state evaluation method, system, equipment and medium, which collects multiple different types of physical quantities generated during the operation of the electric energy metering device as original data to ensure the objectivity and reliability of monitoring data, then constructs and selects features for the monitoring data to reduce the dimension of the feature data, and then uses the selected feature data to train an electric energy metering device operating error state evaluation model, so that the electric energy metering device operating error state can be evaluated based on objective monitoring data.

2. The present invention provides an electric energy metering device operating error state evaluation method, system, equipment and medium, which calculate various types of physical quantities to obtain kurtosis sorting energy characteristics, curve complexity characteristics and phase cosine similarity characteristics as feature data, which can well reflect the distribution characteristics of the collected monitoring data.

3. The present invention provides a method, system, equipment and medium for evaluating the operating error state of an electric energy metering device, and proposes to optimize the NARX neural network model through the Golden Eagle algorithm, which can limit the overall parameter scale of the network, ultimately improve the generalization performance of the NARX neural network, and ultimately improve the accuracy of the error state evaluation.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method of Example 1 of the present invention;

FIG2 is a flowchart illustrating a feature selection process according to an embodiment of the present invention;

FIG3 is a diagram showing an example of a network structure of a NARX neural network model according to an embodiment of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and/or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or combinations thereof.

The term "and/or" means and includes any and all possible combinations of one or more of the associated listed items.

Embodiment 1:

Referring to FIG. 1 , this embodiment provides a method for evaluating an operation error state of an electric energy metering device, comprising the following steps:

S100. Collect historical operation monitoring data of several electric energy metering devices, each of which includes multiple different types of physical quantities generated during the operation of the electric energy metering device. Specifically, the electric energy metering device includes a voltage transformer, a current transformer and the corresponding secondary circuit, and an electric energy meter, and typical faults include electric energy meter faults such as loss of pressure, loss of current, and voltage drop tolerance. The operation monitoring data of the electric energy metering device mainly includes more than ten physical quantities such as voltage, current, and phase power.

As a preferred implementation of this embodiment, the multiple different types of physical quantities generated during the operation of the electric energy metering device collected in this embodiment specifically include:

Voltage data, current data, active power data, reactive power data, power factor data, three-phase unbalance data and load rate data, i.e. historical operation monitoring data for:

in, is the three-phase current data, is the three-phase voltage data, is the active power data, is the reactive power data, is the power factor data, is the three-phase unbalance data, For load factor data, further:

;

;

;

;

;

in, , , Respectively represent the A phase, B phase, and C phase currents; , , Respectively represent the voltage of phase A, phase B, and phase C; , , , Respectively represent the active power of phase A, phase B, phase C and the total; , , , Respectively represent the reactive power of phase A, phase B, phase C and the total; , , , Represents the power factors of phase A, phase B, phase C and the total respectively.

S200, extracting features from each collected historical operation monitoring data, including two steps of constructing features S210 and selecting features S220;

Among them, S210 is specifically:

For each historical operation monitoring data collected , through various types of physical quantities (i.e. the above voltage data, current data, active power data, reactive power data, power factor data, three-phase imbalance data and load rate data), the kurtosis sorting energy characteristics, curve complexity characteristics and phase cosine similarity characteristics of each physical quantity are calculated. The specific steps are as follows:

S211, construct the first feature set (kurtosis sorting energy feature): based on historical operation monitoring data Draw current curve, voltage curve, active power curve, reactive power curve, power factor curve, three-phase unbalance curve, load rate curve;

For each physical quantity curve, the kurtosis is used to reflect the numerical statistics of the random variable distribution characteristics of each physical quantity curve. The kurtosis is a numerical statistic that reflects the distribution characteristics of the random variable. It is the normalized fourth-order central moment. The expression of the kurtosis K is as follows:

;

Where N represents the signal length of the physical quantity curve; represents the i-th signal value in the physical quantity curve; μ represents the signal average value of the physical quantity curve; σ represents the signal standard deviation of the physical quantity curve;

When the electric energy metering device is in different working conditions, the kurtosis values of each IMF (intrinsic mode functions) signal after signal decomposition are quite different. For each signal in each physical quantity curve, EMD empirical mode decomposition is performed, the kurtosis of the decomposed signal is calculated and sorted in descending order according to the calculated kurtosis, and the kurtosis sorting energy E is calculated according to the following formula:

;

Thus, a first feature set is obtained, including current kurtosis sorting energy characteristics, voltage kurtosis sorting energy characteristics, active power kurtosis sorting energy characteristics, reactive power kurtosis sorting energy characteristics, power factor kurtosis sorting energy characteristics, three-phase imbalance kurtosis sorting energy characteristics, and load rate kurtosis sorting energy characteristics.

S212, constructing a second feature set (curve complexity feature): according to the definition of fuzzy entropy, respectively calculating the current curve complexity feature, voltage curve complexity feature, active power curve complexity feature, reactive power curve complexity feature, power factor curve complexity sequence feature, three-phase unbalance curve complexity feature, and load rate curve complexity feature;

Fuzzy entropy (FuzzyEn) also measures the probability of a new pattern. The larger the measurement value, the greater the probability of a new pattern, that is, the greater the sequence complexity. Fuzzy entropy is described as follows:

;

Among them, m is the dimension of phase space, r is the similarity tolerance, N is the dimension of time series, is the fuzzy membership function;

According to the defined fuzzy entropy, the curve complexity characteristics are calculated respectively , the second voltage curve complexity characteristics , Complexity characteristics of the second active power curve , Complexity characteristics of the second reactive power curve , the second power factor curve complexity order characteristics 2. Complexity characteristics of the third-phase imbalance curve , Complexity characteristics of the second load rate curve They are:

;

;

;

;

;

;

.

S213, construct the third feature set (phase-to-phase cosine similarity feature): according to the cosine similarity formula, calculate the cosine similarity features of phase A and phase B currents respectively. , B-phase and C-phase current cosine similarity characteristics , A phase and C phase current cosine similarity characteristics , A-phase and B-phase voltage cosine similarity features , B-phase and C-phase voltage cosine similarity features , A phase and C phase voltage cosine similarity characteristics , A phase and B phase active power cosine similarity characteristics , B phase and C phase active power cosine similarity characteristics , A phase and C phase active power cosine similarity characteristics , A-phase and B-phase reactive power cosine similarity characteristics , B-phase and C-phase reactive power cosine similarity characteristics , A phase and C phase reactive power cosine similarity characteristics , A phase and B phase power factor cosine similarity characteristics , B phase and C phase power factor cosine similarity characteristics , A phase and C phase power factor cosine similarity characteristics ;

Among them, the cosine similarity formula is:

;

Where X, Y represent X and Y vectors; , Represents the magnitude of X and Y vectors respectively;

Therefore, the characteristics are obtained as follows:

.

Step S220 is specifically as follows: for any historical operation monitoring data , the kurtosis sorting energy features, curve complexity features and inter-phase cosine similarity feature sets of the voltage data, current data, active power data, reactive power data, power factor data, three-phase unbalance data and load rate data constructed in step S201 are selected/screened to obtain a final feature set; specifically referring to FIG. 2 , step S220 specifically includes:

S221, taking the kurtosis sorting energy features of each physical quantity as the first feature set, the curve complexity features of each physical quantity as the second feature set, and the inter-phase cosine similarity features of each physical quantity as the third feature set, and merging the first feature set, the second feature set and the third feature set into a fourth feature set;

S222, using the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm to perform feature selection on the fourth feature set respectively, to obtain n feature subsets;

S223, aggregating the n feature subsets into an intersection to obtain a fifth feature set;

S224, performing feature selection on the fifth feature set using the n+1th feature selection algorithm to obtain a sixth feature set, and using the sixth feature set as the final feature set;

Among them, the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm and the n+1th feature selection algorithm are specifically selected from feature selection methods such as variance selection method, correlation coefficient method, chi-square test method, relief algorithm, recursive feature elimination method, penalty-based feature selection method and tree model-based feature selection method.

S300, adding an error state label to the final feature set of each historical operation monitoring data to form a training sample, constructing a nonlinear time series prediction (Nonlinear Autoregressive models with Exogenous Inputs, NARX) neural network model, iteratively training the NARX neural network model through the training sample, and obtaining a trained electric energy metering device operation error state evaluation model;

Specifically refer to Figure 3, which is a schematic diagram of the reference structure of the NARX neural network. The NARX neural network is a dynamic neural network with a clear structure based on the BP neural network. On the basis of the BP neural network, it saves part of the output vector data and introduces it into the input vector in the form of external feedback.

Referring to FIG3 , it is assumed that the current time is t, nx is the order of the neural network input, ny is the order of the neural network output, represents the weight between the i-th input vector and the j-th hidden layer neuron of the neural network, Represents the weight between the i-th hidden layer neuron and the j-th output layer neuron of the neural network, is the offset value of the nth neuron in the hidden layer of the neural network, is the offset value of the neuron in the output layer of the neural network, X(t) is the input vector, Y(t) is the input vector of the feedback output, Represents the transfer function of the hidden layer of the neural network, mostly using the Tan-sigmoid function, represents the neuron output value of the nth hidden layer, Represents the transfer function of the output neuron of a neural network.

The structural parameters of the NARX neural network model are adjusted by using a simple mean square error performance function, which is equivalent to making the NARX neural network model obtain the best fitting effect on the training set data, thereby introducing the problem of "overfitting". The "overfitting" problem is manifested in that the trained NARX neural network model has a weak generalization ability, that is, it only shows good performance in the training set, but the prediction effect is poor when applied to the actual test set. When the training set is fixed, the generalization ability of the neural network is closely related to the scale of its network parameters. For this reason, this embodiment introduces the Golden Eagle algorithm, which aims to limit the overall parameter scale of the network by optimizing the weight coefficients and threshold parameters in the NARX neural network, and ultimately improve the generalization performance of the NARX neural network.

The Golden Eagle Algorithm is the intelligence of the Golden Eagle to tune its speed at different stages of its spiral trajectory for hunting. They show more tendencies to wander and search for prey in the initial stages of hunting, and more tendencies to attack in the final stages. A Golden Eagle adjusts these two components to capture the best prey in the shortest time in a feasible area. The specific steps are as follows:

S311, initializing the number of golden eagle individuals in the golden eagle population, each golden eagle containing different weight coefficient information and threshold parameter information of the NARX neural network model;

S312, calculate the fitness value of the golden eagle individual and initialize the group memory according to the fitness value; wherein, the fitness value calculation formula of the golden eagle individual is:

;

Among them, value represents the fitness value, N is the total number of golden eagle individuals, is the predicted value of the NARX neural network model obtained according to the weight coefficient information and threshold parameter information of the NARX neural network model contained in the i-th Golden Eagle, is the true value of the sample corresponding to the i-th golden eagle;

S313, Initialize the attack tendency of the Golden Eagle and cruising tendency ;

S314, update the attack tendency according to the following formula and cruising tendency :

;

in, and for The initial and final values of and for The initial and final values of ; t is the tth iteration, and T is the total number of iterations;

S315. Randomly select prey from the attack vector calculated by the memory of the population:

;

in, is the attack vector of the i-th golden eagle, This is the best place for the Golden Eagle to reach at present. is the current position of the i-th golden eagle;

S316, calculate the cruise vector d:

;

in, is the normal vector, is the decision variable vector;

set up is any point on the hyperplane, then: ;

Bundle As the normal of the hyperplane, the hyperplane is expressed as:

;

in, is the attack vector, is the decision vector, The location of the selected prey; represents the attack vector at the tth iteration;

but:

;

in, is the kth element of the target point, k is the number of the fixed variable;

The destination point on the cruise hyperplane is expressed as:

The step vector of the Golden Eagle iteration is defined as:

;

in, and is a random vector in [0,1]. The position of the golden eagle in the iteration is calculated by adding the step vector in the iteration to the position in the iteration:

in, For the Golden Eagle Second position, For the Golden Eagle The second position, is the step size of the Golden Eagle's movement;

S317, calculating the fitness value according to the updated position of the new golden eagle individual, and updating the optimal solution and the optimal position;

S318. Determine whether the maximum number of iterations has been reached. If the maximum number of iterations has been reached, output the weight coefficient information and threshold parameter information contained in the golden eagle individual at the current optimal position as the weight coefficient and threshold parameter of the NARX neural network model; if the maximum number of iterations has not been reached, return to step S314 and continue iterating.

Based on the neural network optimization method provided in this embodiment, the iterative optimization process of the Golden Eagle algorithm is the update process of the weight parameters and threshold parameters in the NARX neural network training process, where the fitness corresponds to the NARX neural network error function (value), and the error is minimized by changing the position of the Golden Eagle. Each Golden Eagle contains the weight coefficient information and threshold parameter information in the NARX neural network. When the Golden Eagle reaches the optimal position, the weight coefficient information and threshold parameter information contained in the Golden Eagle individual are the updated optimal parameters.

S400, obtaining current operation monitoring data of the target electric energy metering device, performing feature extraction based on the current operation monitoring data to obtain the current final feature set of the target electric energy metering device, and inputting the feature set into a trained electric energy metering device operation error state evaluation model to obtain an error state evaluation result of the target electric energy metering device.

Embodiment 2:

This embodiment provides an electric energy metering device operation error state evaluation system, including:

A collection module, used to collect historical operation monitoring data of several electric energy metering devices, each of which includes multiple different types of physical quantities generated during the operation of the electric energy metering device; this module is used to implement the function of step S100 in the first embodiment, which will not be described in detail here;

A feature extraction module, used to extract features from each collected historical operation monitoring data, including two steps of constructing features and selecting features; this module is used to implement the function of step S200 in the first embodiment;

The step of constructing features is specifically as follows: for each historical operation monitoring data, the kurtosis sorting energy feature, the curve complexity feature and the phase cosine similarity feature corresponding to each physical quantity are calculated by various types of physical quantities;

The feature selection step specifically includes: for any historical operation monitoring data, screening the set of features obtained in the feature construction step to obtain a final feature set;

A model training module, used to add an error state label to the final feature set of each historical operation monitoring data to form a training sample, construct a neural network model, iteratively train the neural network model through the training sample, and obtain a trained electric energy metering device operation error state evaluation model; this module is used to implement the function of step S300 in the first embodiment;

An evaluation module is used to obtain the current operation monitoring data of the target electric energy metering device, perform feature extraction based on the current operation monitoring data to obtain the current final feature set of the target electric energy metering device, and input it into the trained electric energy metering device operation error state evaluation model to obtain the error state evaluation result of the target electric energy metering device; this module is used to implement the function of step S400 in Example 1.

Embodiment three:

This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for evaluating the operating error state of an electric energy metering device as described in any embodiment of the present invention is implemented.

Embodiment 4:

This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for evaluating the operating error state of an electric energy metering device as described in any embodiment of the present invention is implemented.

In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And/or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and/or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character "/" generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims (5)

1. A method for evaluating the operating error state of an electric energy metering device, characterized in that it comprises the following steps: Collecting historical operation monitoring data of a number of electric energy metering devices, each historical operation monitoring data includes a plurality of different types of physical quantities generated during the operation of the electric energy metering device; Perform feature extraction on each collected historical operation monitoring data, including two steps: feature construction and feature selection; The step of constructing features is specifically as follows: for each historical operation monitoring data, the kurtosis sorting energy feature, the curve complexity feature and the phase cosine similarity feature corresponding to each physical quantity are calculated by various types of physical quantities; The feature selection step specifically includes: for any historical operation monitoring data, screening the set of features obtained in the feature construction step to obtain a final feature set; Add error state labels to the final feature set of each historical operation monitoring data to form training samples, build a neural network model, iteratively train the neural network model through the training samples, and obtain a trained electric energy metering device operation error state evaluation model; Acquire current operation monitoring data of the target electric energy metering device, perform feature extraction based on the current operation monitoring data to obtain the current final feature set of the target electric energy metering device, and input it into the trained electric energy metering device operation error state evaluation model to obtain the error state evaluation result of the target electric energy metering device; The multiple different types of physical quantities generated during the operation of the electric energy metering device specifically include: Voltage data, current data, active power data, reactive power data, power factor data, three-phase unbalance data and load rate data; In the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and inter-phase cosine similarity characteristics of each physical quantity by various types of physical quantities, the calculation method of the kurtosis sorting energy characteristics of each physical quantity is: Draw corresponding physical quantity curves according to various types of physical quantities, including current curve, voltage curve, active power curve, reactive power curve, power factor curve, three-phase unbalance curve, and load rate curve; For each physical quantity curve, the kurtosis reflects the numerical statistics of the random variable distribution characteristics of each physical quantity curve. The expression of the kurtosis K is as follows: ; Where N represents the signal length of the physical quantity curve; represents the i-th signal value in the physical quantity curve; μ represents the signal average value of the physical quantity curve; σ represents the signal standard deviation of the physical quantity curve; For each signal in each physical quantity curve, perform EMD empirical mode decomposition, calculate the kurtosis of the decomposed signal and sort it in descending order according to the calculated kurtosis, and then calculate the kurtosis sorting energy E according to the following formula: ; Thus, the current kurtosis sorting energy characteristics, voltage kurtosis sorting energy characteristics, active power kurtosis sorting energy characteristics, reactive power kurtosis sorting energy characteristics, power factor kurtosis sorting energy characteristics, three-phase unbalance kurtosis sorting energy characteristics, and load rate kurtosis sorting energy characteristics are obtained; In the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and inter-phase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the curve complexity characteristics of each physical quantity is: Define fuzzy entropy as: ; Among them, m is the dimension of phase space, r is the similarity tolerance, N is the dimension of time series, is the fuzzy membership function; According to the defined fuzzy entropy, the curve complexity characteristics are calculated respectively , the second voltage curve complexity characteristics , Complexity characteristics of the second active power curve , Complexity characteristics of the second reactive power curve , the second power factor curve complexity order characteristics 2. Complexity characteristics of the third-phase imbalance curve , Complexity characteristics of the second load rate curve They are: ; ; ; ; ; ; ; in, , , Respectively represent the A phase, B phase, and C phase currents; , , They are phase A, phase B, and phase C voltages respectively; , , , Respectively represent the active power of phase A, phase B, phase C and the total; , , , Respectively represent the reactive power of phase A, phase B, phase C and the total; , , , Respectively represent the current power factors of phase A, phase B, phase C and the total; It is the three-phase unbalance data; R _{load factor} indicates the load factor; In the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and interphase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the interphase cosine similarity characteristics of each physical quantity is: According to the cosine similarity formula, the cosine similarity characteristics of phase A and phase B current are calculated respectively. , B-phase and C-phase current cosine similarity characteristics , A phase and C phase current cosine similarity characteristics , A-phase and B-phase voltage cosine similarity features , B-phase and C-phase voltage cosine similarity features , A phase and C phase voltage cosine similarity characteristics , A phase and B phase active power cosine similarity characteristics , B phase and C phase active power cosine similarity characteristics , A phase and C phase active power cosine similarity characteristics , A-phase and B-phase reactive power cosine similarity characteristics , B-phase and C-phase reactive power cosine similarity characteristics , A phase and C phase reactive power cosine similarity characteristics , A phase and B phase power factor cosine similarity characteristics , B phase and C phase power factor cosine similarity characteristics , A phase and C phase power factor cosine similarity characteristics ; Among them, the cosine similarity formula is: ; Where X, Y represent X and Y vectors; , Represents the magnitude of X and Y vectors respectively; The method of screening the set of features obtained in the feature construction step to obtain the final feature set is specifically: The kurtosis sorting energy features of each physical quantity are used as the first feature set, the curve complexity features of each physical quantity are used as the second feature set, the inter-phase cosine similarity features of each physical quantity are used as the third feature set, and the first feature set, the second feature set and the third feature set are combined into a fourth feature set; Use the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm to perform feature selection on the fourth feature set respectively to obtain n feature subsets; Merge n feature subsets into the fifth feature set; Use the n+1th feature selection algorithm to perform feature selection on the fifth feature set to obtain a sixth feature set, and use the sixth feature set as the final feature set; Among them, the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm and the n+1th feature selection algorithm are specifically selected from variance selection method, correlation coefficient method, chi-square test method, relief algorithm, recursive feature elimination method, penalty-based feature selection method and tree model-based feature selection method. 2. According to claim 1, a method for evaluating the operating error state of an electric energy metering device is characterized in that the neural network model adopts a nonlinear time series prediction NARX neural network model, and the Golden Eagle algorithm is used to optimize the neural network model, and the specific steps are: Initialize the number of golden eagle individuals in the golden eagle population, each of which contains different weight coefficient information and threshold parameter information of the NARX neural network model; Calculate the fitness value of the golden eagle individual and initialize the group memory according to the fitness value; the fitness value calculation formula of the golden eagle individual is: ; Among them, value represents the fitness value, N is the total number of golden eagle individuals, is the predicted value of the NARX neural network model obtained based on the weight coefficient information and threshold parameter information of the NARX neural network model contained in the i-th Golden Eagle, is the true value of the sample corresponding to the i-th golden eagle; Initialize the Golden Eagle's attack tendency and cruising tendency ; Update the attack tendency according to the following formula and cruising tendency : ; in, and for The initial and final values of and for The initial and final values of ; t is the tth iteration, and T is the total number of iterations; Prey is randomly selected from the attack vector calculated from the memory of the population: ; in, is the attack vector of the i-th golden eagle, This is the best place for the Golden Eagle to reach at present. is the current position of the i-th golden eagle; Calculate the cruise vector d: ; in, is the normal vector, is the decision variable vector; set up is any point on the hyperplane, then: ; Bundle As the normal of the hyperplane, the hyperplane is expressed as: ; in, is the attack vector, is the decision vector, The location of the selected prey; represents the attack vector at the tth iteration; but: ; in, is the kth element of the target point, k is the number of the fixed variable; The destination point on the cruise hyperplane is expressed as: The step vector of the Golden Eagle iteration is defined as: ; in, and is a random vector in [0,1]. The position of the golden eagle in the iteration is calculated by adding the step vector in the iteration to the position in the iteration: in, For the Golden Eagle Second position, For the Golden Eagle The second position, is the step size of the Golden Eagle's movement; Calculate the fitness value according to the updated position of the new golden eagle individual, and update the optimal solution and optimal position; Determine whether the maximum number of iterations has been reached. If the maximum number of iterations has been reached, output the weight coefficient information and threshold parameter information of the golden eagle individual at the current optimal position as the weight coefficient and threshold parameter of the NARX neural network model; if the maximum number of iterations has not been reached, continue iterating. 3. A system for evaluating the operating error state of an electric energy metering device, comprising: A collection module, used to collect historical operation monitoring data of a number of electric energy metering devices, each historical operation monitoring data includes a plurality of different types of physical quantities generated during the operation of the electric energy metering device; The feature extraction module is used to extract features from each collected historical operation monitoring data, including two steps: feature construction and feature selection; The step of constructing features is specifically as follows: for each historical operation monitoring data, the kurtosis sorting energy feature, the curve complexity feature and the phase cosine similarity feature corresponding to each physical quantity are calculated by various types of physical quantities; The feature selection step specifically includes: for any historical operation monitoring data, screening the set of features obtained in the feature construction step to obtain a final feature set; A model training module is used to add an error state label to the final feature set of each historical operation monitoring data to form a training sample, construct a neural network model, and iteratively train the neural network model through the training sample to obtain a trained electric energy metering device operation error state evaluation model; An evaluation module is used to obtain the current operation monitoring data of the target electric energy metering device, perform feature extraction based on the current operation monitoring data to obtain the current final feature set of the target electric energy metering device, and input it into the trained electric energy metering device operation error state evaluation model to obtain the error state evaluation result of the target electric energy metering device; The multiple different types of physical quantities generated during the operation of the electric energy metering device specifically include: Voltage data, current data, active power data, reactive power data, power factor data, three-phase unbalance data and load rate data; In the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and inter-phase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the kurtosis sorting energy characteristics of each physical quantity is: Draw corresponding physical quantity curves according to various types of physical quantities, including current curve, voltage curve, active power curve, reactive power curve, power factor curve, three-phase unbalance curve, and load rate curve; For each physical quantity curve, the kurtosis reflects the numerical statistics of the random variable distribution characteristics of each physical quantity curve. The expression of the kurtosis K is as follows: ; Where N represents the signal length of the physical quantity curve; represents the i-th signal value in the physical quantity curve; μ represents the signal average value of the physical quantity curve; σ represents the signal standard deviation of the physical quantity curve; For each signal in each physical quantity curve, perform EMD empirical mode decomposition, calculate the kurtosis of the decomposed signal and sort it in descending order according to the calculated kurtosis, and then calculate the kurtosis sorting energy E according to the following formula: ; Thus, the current kurtosis sorting energy characteristics, voltage kurtosis sorting energy characteristics, active power kurtosis sorting energy characteristics, reactive power kurtosis sorting energy characteristics, power factor kurtosis sorting energy characteristics, three-phase unbalance kurtosis sorting energy characteristics, and load rate kurtosis sorting energy characteristics are obtained; In the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and inter-phase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the curve complexity characteristics of each physical quantity is: Define fuzzy entropy as: ; Among them, m is the dimension of phase space, r is the similarity tolerance, N is the dimension of time series, is the fuzzy membership function; According to the defined fuzzy entropy, the curve complexity characteristics are calculated respectively , the second voltage curve complexity characteristics , Complexity characteristics of the second active power curve , Complexity characteristics of the second reactive power curve , the second power factor curve complexity order characteristics 2. Complexity characteristics of the third-phase imbalance curve , Complexity characteristics of the second load rate curve They are: ; ; ; ; ; ; ; in, , , Respectively represent the A phase, B phase, and C phase currents; , , They are phase A, phase B, and phase C voltages respectively; , , , Respectively represent the active power of phase A, phase B, phase C and the total; , , , Respectively represent the reactive power of phase A, phase B, phase C and the total; , , , Respectively represent the current power factors of phase A, phase B, phase C and the total; It is the three-phase unbalance data; R _{load factor} indicates the load factor; In the step of calculating the kurtosis sorting energy characteristics, curve complexity characteristics and interphase cosine similarity characteristics corresponding to each physical quantity by various types of physical quantities, the calculation method of the interphase cosine similarity characteristics of each physical quantity is: According to the cosine similarity formula, the cosine similarity characteristics of phase A and phase B current are calculated respectively. , B-phase and C-phase current cosine similarity characteristics , A phase and C phase current cosine similarity characteristics , A-phase and B-phase voltage cosine similarity features , B-phase and C-phase voltage cosine similarity features , A phase and C phase voltage cosine similarity characteristics , A phase and B phase active power cosine similarity characteristics , B phase and C phase active power cosine similarity characteristics , A phase and C phase active power cosine similarity characteristics , A-phase and B-phase reactive power cosine similarity characteristics , B-phase and C-phase reactive power cosine similarity characteristics , A phase and C phase reactive power cosine similarity characteristics , A phase and B phase power factor cosine similarity characteristics , B phase and C phase power factor cosine similarity characteristics , A phase and C phase power factor cosine similarity characteristics ; Among them, the cosine similarity formula is: ; Where X, Y represent X and Y vectors; , Represents the magnitude of X and Y vectors respectively; The method of screening the set of features obtained in the feature construction step to obtain the final feature set is specifically: The kurtosis sorting energy features of each physical quantity are used as the first feature set, the curve complexity features of each physical quantity are used as the second feature set, the inter-phase cosine similarity features of each physical quantity are used as the third feature set, and the first feature set, the second feature set and the third feature set are combined into a fourth feature set; Use the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm to perform feature selection on the fourth feature set respectively to obtain n feature subsets; Merge n feature subsets into the fifth feature set; Use the n+1th feature selection algorithm to perform feature selection on the fifth feature set to obtain a sixth feature set, and use the sixth feature set as the final feature set; Among them, the first feature selection algorithm, the second feature selection algorithm, ... the nth feature selection algorithm and the n+1th feature selection algorithm are specifically selected from variance selection method, correlation coefficient method, chi-square test method, relief algorithm, recursive feature elimination method, penalty-based feature selection method and tree model-based feature selection method. 4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for evaluating the operating error state of an electric energy metering device as described in any one of claims 1 to 2 is implemented. 5. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for evaluating the operating error state of an electric energy metering device according to any one of claims 1 to 2 is implemented.