CN114814707A - A method, device, terminal and readable medium for analyzing stress error of smart meter - Google Patents

Abstract

The invention discloses a stress error analysis method for an intelligent ammeter, and discloses equipment and a storage medium with the method. The error analysis method analyzes the relevance of various stresses and errors by using a BP neural network algorithm optimized by LM. An additional error is introduced during error modeling, so that error model correction of the intelligent electric energy meter is realized. The method of the invention does not need excessive error calculation and error fitting in the past error analysis. By adopting the PCA idea, the phenomena of over-fitting and under-fitting are not easy to occur, the error analysis efficiency is improved, and higher reliability can be ensured. Therefore, the method of the invention can lead the error predicted value of the intelligent electric energy meter to have a higher confidence interval under certain conditions and lay a foundation for the error correction work of the intelligent electric energy meter.

Description

A method, device, terminal and readable medium for analyzing stress error of smart meter

technical field

The invention relates to the field of intelligent power transmission in power grids, in particular to a method, device and readable medium for analyzing stress error of a smart meter.

Background technique

In the smart grid, smart meters undertake the tasks of energy data collection, measurement and transmission, and develop two-way multi-rate metering functions, user-side control functions, two-way data communication functions with multiple data transmission modes, and anti-theft functions, etc. Intelligent services greatly promote information analysis, integration and optimization, and information presentation. The improvement of the function of smart energy meters is in line with the trend of smart grid and new energy development, and its development also promotes the integration of the physical layer and the information layer of the modern power system.

The difference of the structural modules of the smart energy meter, the internal complexity of each functional module, and the diversity of the topological relationship between the modules make the error situation of the smart energy meter more complicated during operation, which makes the existing error analysis methods such as polynomial regression, neural network. It is difficult to describe the error situation of smart energy meters objectively and accurately by using method, least squares method, etc. The existing error inspection methods have the following two problems: on the one hand, the inspection only verifies whether the error under the reference condition meets the requirements of its accuracy level, and then estimates the confidence interval of the measured value, so as to evaluate the qualified electric energy meter. Whether it is or not results in a waste of useful information; on the other hand, the experiment is carried out under the environmental conditions of the laboratory, which fails to fully consider the diverse use environments and the existence of various stresses in the actual working conditions.

SUMMARY OF THE INVENTION

The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a stress error analysis method for a smart meter, which can reveal a variety of stress factors and relationships from the data level by using the BP (Back-propagation error back-propagation) neural network optimized by LM (Levenberg-Marquardt). Intrinsic connection of energy meter errors.

The present invention also provides a device and a storage medium having the above-mentioned method for analyzing the stress error of a smart meter.

The method for analyzing the stress error of a smart meter according to the embodiment of the first aspect of the present invention is characterized in that it includes the following steps:

Obtain stress data and error data of smart meters;

determining typical stress data in the stress data of the smart meter;

Sending the typical stress data and its error data into a neural network for training to obtain a trained neural network;

Using the trained neural network, the error data relationship under different stress conditions is established.

The method for analyzing the stress error of a smart meter according to the embodiment of the present invention has at least the following beneficial effects: using the data collected by the smart meter, and then sending it into the BP neural network optimized by LM for training, so as to obtain the trained neural network, using It is used to predict the change relationship of the error under different stress conditions.

According to some embodiments of the present invention, the step of determining the typical stress in the stress data of the smart meter includes:

standardize the data;

Using the normalized data, calculate the contribution rate of each stress;

Several stresses with a large contribution rate are selected as typical stresses.

According to some embodiments of the present invention, the typical stress includes: temperature, humidity, air pressure, and regional voltage.

According to some embodiments of the present invention, the contribution rate of the typical stress exceeds 95%.

According to some embodiments of the present invention, after the step of determining the typical stress in the stress data of the smart meter, the step further includes:

The typical stress data and its error data are processed to remove abnormal data to obtain processed data.

According to some embodiments of the present invention, the steps of processing the typical stress data and its error data, removing abnormal data, and obtaining processed data include:

Arrange the error data in the smart meter according to the acquisition sequence to obtain x(t), and calculate the average μ and standard deviation σ;

Eliminate abnormal data in the error data;

Sequential anomaly correction using mean interpolation;

The processed data is obtained by transforming the stress sample series using a normalization method.

According to some embodiments of the present invention, the neural network used in the smart meter stress error analysis method is a BP neural network optimized based on LM.

According to some embodiments of the present invention, the step of sending the processed data into a neural network for training to obtain a trained neural network includes:

Set the stress parameter to the input vector;

Set the smart meter error as the output vector;

Set the actual error to the expected output vector.

According to some embodiments of the present invention, the method further comprises:

Use the trained neural network to predict the estimation error;

The estimated error is added to the reference error of the smart meter as an error offset of the smart meter.

The device for analyzing the stress error of a smart meter according to the embodiment of the second aspect of the present invention is characterized in that, it includes:

The data collection module can obtain the stress data and the error data of the smart meter;

a typical stress analysis module, capable of determining typical stress data in the stress data of the smart meter;

A model training module, capable of sending the typical stress data and its error data into a neural network for training to obtain a trained neural network;

The relationship analysis module can use the trained neural network to establish the error data relationship under different stress conditions.

A terminal according to an embodiment of the third aspect of the present invention includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to achieve The above-mentioned smart meter stress error analysis method.

The computer-readable medium according to the embodiment of the fourth aspect of the present invention is characterized in that, the computer medium stores computer software, and the software can implement the above-mentioned method for analyzing the stress error of a smart meter when running.

The stress error analysis method of the smart meter of the present invention uses the BP neural network algorithm optimized by LM to analyze the correlation between various stresses and errors. An additional error is introduced in the error modeling, so as to realize the error model correction of the smart energy meter. The method of the present invention does not need the error calculation and error fitting of the previous error analysis. Using the PCA idea, it is not easy to appear over-fitting and under-fitting, that is, to improve the efficiency of error analysis, and to ensure high reliability. Therefore, the method of the present invention can make the error prediction value of the smart electric energy meter have a higher confidence interval under certain conditions, and lay a foundation for the error correction work of the smart electric energy meter in the future.

Additional aspects and advantages of the present invention will be set forth, in part, from the following description, and in part will be apparent from the following description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of embodiments taken in conjunction with the accompanying drawings, wherein:

Fig. 1 is the metering principle diagram of the smart meter used in the embodiment of the present invention;

2 is a schematic diagram of steps of a method for analyzing stress error of a smart meter according to Embodiment 1 of the present invention;

3 is a graph showing the influence of temperature on the error of the electric meter in the second embodiment of the present invention;

FIG. 4 is a graph showing the influence of humidity on the error of the electric meter in the second embodiment of the present invention;

5 is a graph showing the influence of air pressure on the error of the electric meter in the second embodiment of the present invention;

Fig. 6 is the influence of the voltage in different regions on the error of the electric meter in the second embodiment of the present invention; wherein, Fig. 6-a represents the influence of the voltage in the Fujian region on the error of the electric meter, Fig. 6-b represents the influence of the voltage in the Heilongjiang region on the error of the electric meter, Fig. 6 -c represents the effect of voltage on the meter error in Tibet, and Figure 6-d represents the effect of voltage on the meter error in Xinjiang.

Detailed ways

The following describes in detail the embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary, only used to explain the present invention, and should not be construed as a limitation of the present invention.

In the description of the present invention, it should be understood that the azimuth description, such as the azimuth or position relationship indicated by up, down, front, rear, left, right, etc., is based on the azimuth or position relationship shown in the drawings, only In order to facilitate the description of the present invention and simplify the description, it is not indicated or implied that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore should not be construed as limiting the present invention.

In the description of the present invention, the meaning of several is one or more, the meaning of multiple is two or more, greater than, less than, exceeding, etc. are understood as not including this number, above, below, within, etc. are understood as including this number. If it is described that the first and the second are only for the purpose of distinguishing technical features, it cannot be understood as indicating or implying relative importance, or indicating the number of the indicated technical features or the order of the indicated technical features. relation.

In the description of the present invention, unless otherwise clearly defined, words such as setting, installation, connection should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

As shown in Figure 1, Figure 1 shows the metering principle of a smart meter. In a smart meter, the change of error is very complicated under the influence of various factors. At present, it is difficult to design a meter that can determine the meter by theoretical calculation. error method. However, most of the current energy meter error analysis methods fail to consider the influence of the actual working environment and working conditions on the error, so there are limitations.

In the analysis method of the present invention, various data affecting the error are abstracted into various stresses, and the relationship between the error and the stress is analyzed by the artificial intelligence method, which can improve the accuracy of the error analysis and compensate the error of the electric energy meter later. and lay the groundwork for corrections. details as follows.

Embodiment 1.

Referring to FIG. 2 , an embodiment of the present application provides a method for analyzing stress error of a smart meter, which includes the following steps:

Step S100, acquiring stress data and error data of the smart meter.

When acquiring data from smart meters, it is necessary to continuously collect data for a long period of time to obtain as much data as possible, which can increase the number of training sets of the model and thus increase the accuracy of the results.

Preferably, the electricity meter itself can record electricity consumption data, stress data and error data. When sampling at a certain time, take out all the stress condition data X _i , and the stress data at time i can be expressed as:

_Xi = [X ₁ , X ₂ , . . . , X _m ] (1)

where m is the number of stresses. The error data at time i is Y _i .

After collecting for a long enough time and obtaining n groups of data, these n groups of data can be expressed as a stress data matrix, as follows:

The corresponding error vector in n sets of data can be expressed as:

Y=[Y ₁ , Y ₂ , ..., Y _n ] ^T (3)

Step S200, determining the typical stress in the smart meter stress data.

The amount of stress data collected in step S100 is usually large, and some of the data itself has very little impact. If these are taken into account, it will not only affect the speed of subsequent training of the model, but also cause errors and dispersion due to these low-impact data. attention. It is only necessary to select the larger proportion of the stress for study.

In order to find several stresses that are most closely related to the error of the electric energy meter, an improved principal component analysis (PCA) idea is used to find a set of orthogonal vectors in the sample space, and use this set of orthogonal vectors to express the overall situation, so as to realize the use of A few principal components are used to describe the original high-dimensional data and preserve the information of the original data to the greatest extent.

Among the n valid data samples obtained from the experiment, each sample has m sampling data, the sample data matrix can be obtained as X _n×m , and each row vector of which is an observation data sample _xi , where 0<i ≤n; each column vector is the feature quantity xj of the corresponding observation sample, where 0<j≤m.

Step S201, standardize the data.

The formula used for normalization is as follows:

为标准化后的特征量，

为特征量均值，s(x_j)为特征量标准差。In the formula,

is the standardized feature quantity,

is the feature quantity mean, and s(x _j ) is the feature quantity standard deviation.

Step S202 , using the standardized data, calculate the contribution rate of each stress.

得到协方差矩阵为P，即：Let the normalized matrix be

The covariance matrix is obtained as P, that is:

Then the eigenvalue λ _i of the covariance matrix P and its eigenvector _ei are calculated. Through the similar diagonalization of the matrix, the similar diagonal matrix D with P is obtained, and the elements on the diagonal in D are arranged in descending order of eigenvalues, that is, D=diag(λ ₁ , λ ₂ ,...,λ _n ), then P=EDE ^T , where E is the set of eigenvalues λ _i corresponding to the eigenvectors ei, and E is a standard orthogonal matrix, that is, E=diag(e ₁ , e ₂ ,..., _en ), through the formula ( 6) to obtain each principal component vector m ₁ , m ₂ , ..., m _n .

Among them, the contribution rate corresponding to the kth principal component is expressed as:

λ _k represents the eigenvalue of the kth principal component, and λ _i corresponds to the eigenvector _ei .

Step S203 , selecting several stresses with a larger contribution ratio as typical stresses.

The contribution rate is sorted from high to low, and several principal components are selected as the typical stress.

According to a large number of experiments, the cumulative contribution rate of the first three or the first four principal components can reach more than 95%, so when taking the typical stress, only three or four kinds of stress need to be selected, and the main influencing factors can be analyzed. These stresses include: temperature, humidity, air pressure, regional voltage.

Step S300: Process the typical stress data and error data, remove abnormal data, and obtain processed data. It can increase the accuracy of training data in the subsequent process, and exclude data that is obviously unreasonable during sampling.

According to some preferred embodiments of the present application, this step specifically includes:

Step S301, arranging the error data in the smart meter according to the collection sequence to obtain x(t), and calculating the average value μ and the standard deviation σ;

Step S302, removing abnormal data in the error data.

Using the "3σ criterion" commonly used in statistics, the data points in the data segment that are not in the range of [μ-3σ, μ+3σ] are determined as outliers and eliminated.

Step S303 , performing sequence abnormality correction using mean value interpolation.

The excluded data will vacate a position, average the data before and after this position, and fill it to the original position. The formula is as follows:

Step S304 , transform the stress sample sequence using a normalization method.

Using the z-score method (normalization method), transform the sample sequence x ₁ , x ₂ , ..., x _n of stress, the formula is as follows:

The transformed sequence is y ₁ , y ₂ , ..., y _n , the mean of the sequence is 0, the variance is 1, and it is dimensionless. Get processed data.

Step S400, sending the processed data into a neural network for training.

Under the three-layer neural network structure, set the preprocessed typical stress y=[y ₁ , y ₂ ,..., y _n ] as the input vector, and the smart energy meter error z=[z ₁ , z ₂ ,..., z _m ] is the output vector, the actual error t is the expected output vector, w is the connection weight of the neural network, and the sigmoid function is used as the activation function in the intermediate layer and the output layer.

When the output result of the network, that is, the error z of the smart energy meter does not meet the setting requirements, the data processing of the network enters the back-propagation link. At this time, the error signal modifies the weights and thresholds of each layer unit from the back to the front. In this paper, the gradient descent is used. The principle of the method completes the process of back propagation. The error function of the output layer is set to E, and its expression is:

Among them, w is all weights, tk is the actual value of the smart energy meter error at the kth iteration, zk is the estimated value of the smart energy meter error after the kth iteration, t is the actual value of the smart energy meter error, z is the estimation error of the smart energy meter.

According to some preferred embodiments of the present application, the so-called actual value t of the smart meter error is an accurate error obtained from a large amount of experimental data, which can be considered as the actual smart meter error corresponding to the stress condition.

In the initial stage of the model, the initial value of the weight is set to a random value within a certain range. The Hessian matrix is decomposed into the product of the Jacobians matrix, and then its inverse matrix is obtained, thereby reducing the complexity of the calculation and updating the weights faster. The calculation formula is:

In the formula, ΔW is the variation of the network weight; h is the number of iterations; J _h is the Jacobians matrix of the error function of the h-th iteration; μ _h is a constant greater than zero; E is the identity matrix; e _h is the h-th return pass error.

The process of model training is to reduce the error value through multiple iterations, and the weight vector is updated at the ith iteration as:

w _(i+1) = w _(i) + Δw _(i) (12)

When the output error function value satisfies the following formula:

E(h)≤e _m (13)

Among them, _em is the pre-set termination criterion, which can be set according to the accuracy requirements. When Equation (13) is satisfied, it means that when the end condition is reached, the iterative calculation stops and the training of the neural network model is completed.

That is to say, when the difference between the estimated error k trained by the model and the actual smart meter error t is small enough, it means that the model has a fairly high accuracy during training.

The trained model can mine the nonlinear relationship between the stress characteristic value acting on the electric energy meter and the error value, and can predict the error more accurately under certain conditions.

According to some preferred embodiments of the present application, the present application can rely on the above model to predict the error conditions under different stress conditions, and then correct the error of the electric meter according to the above error, which can be specifically described as:

Step S500, estimating the error offset of the smart energy meter according to the environment and working conditions of the smart energy meter.

When analyzing the error of the smart electric energy meter, the method of the invention fully considers the influence of various stresses on the error in actual use, and considers the additional errors caused by various stresses, thereby realizing the correction of the error of the smart electric energy meter.

Specifically include:

Step S501, using the trained neural network to predict the estimation error.

The estimated error z caused by the stress error of the smart meter in the current operating state is predicted by the model. The estimated error z generated by the stress error is the final output of the above model after iterative optimization of parameters for several times and satisfying the termination criterion.

Step S502 , adding the estimated error to the reference error of the smart meter as an error offset of the meter.

The estimated error z obtained in step S501 and the reference error Δe ₀ of the electric energy meter under standard conditions are added to obtain the error offset Δe of the electric energy meter. The formula can be expressed as:

Δe=z+Δe ₀ (14)

The reference error Δe ₀ of the electric energy meter is a statistical error under ideal conditions, and the reference error Δe ₀ is also a random error, which is the rated error of a certain electric energy meter under standard conditions, which is essentially a level of accuracy. Because the errors of different electric energy meters are not necessarily the same under the same working conditions, the random errors of electric energy meters should be considered.

The estimated error z caused by the stress error under the actual conditions is caused by the inconsistency between the stress conditions of the working environment and the ideal conditions. According to the model in this scheme, the stress condition of the electric energy meter to be measured is taken as the input variable, and the Estimate the error z, and calculate the error offset Δe of the electric energy meter according to the above formula 14.

Step S600, using the trained neural network to establish the relationship between the error offsets under different stress conditions.

The trained neural network model can establish the relationship between the error offset Δe and the stress condition, which can be shown by drawing, and the change trend can be seen intuitively.

Embodiment two,

In order to verify the validity of the above-mentioned embodiment, on the basis of the first embodiment, the data is brought into the description:

Step A100: Acquire stress data and error data of the smart meter.

It is consistent with the method described in Embodiment 1, and is omitted here.

Step A200, determining the typical stress in the smart meter stress data.

In a certain experiment, among 60725 valid data samples, each sample has 8 sample data, the sample data matrix is X _60725×8 , and each row vector is 1 observation data sample x _i , where 0<i≤60725; each column vector is the feature quantity xj of the corresponding observation sample, where 0<j≤8.

Step A201, standardize the data.

The formula is as follows:

为标准化后的特征量，

为特征量均值，s(x_j)为特征量标准差。In the formula,

is the standardized feature quantity,

is the feature quantity mean, and s(x _j ) is the feature quantity standard deviation.

Step A202: Calculate the contribution rate of each stress by using the standardized data.

得到协方差矩阵为P，即：Let the normalized matrix be

The covariance matrix is obtained as P, that is:

Calculate the eigenvalue λ _i of the covariance matrix P and its eigenvector _ei , namely P=EDE ^T where D is a diagonal matrix arranged in descending order of eigenvalues, D=diag(λ ₁ , λ ₂ ,...,λ ₆₀₇₂₅ ), E is the set of eigenvalues λ _i corresponding to the eigenvectors e _i , and E is a standard orthogonal matrix, E=diag(e ₁ , e ₂ , ..., e ₆₀₇₂₅ ), obtained by the linear change of formula (17) Each principal component vector m ₁ , m ₂ , ..., m ₆₀₇₂₅ .

The contribution rate corresponding to the kth principal component is:

Step A203: Select several stresses with the largest contribution ratio as typical stresses.

The contribution rate is sorted from high to low, and several principal components are selected as the typical stress.

Experiments show that the cumulative contribution rate of the first four principal components has reached 95%. These four principal components are temperature, humidity, air pressure and voltage stress, which are used as the input variables of the following model.

Step A300: Process typical stress data and error data, remove abnormal data, and obtain processed data.

Step A301: Arrange the error data in the smart meter according to the collection sequence to obtain x(t), and calculate the average value μ and the standard deviation σ;

Step A302, setting a rejection threshold.

Using the "3σ criterion" commonly used in statistics, the data points in the data segment that are not in the range of [μ-3σ, μ+3σ] are determined as outliers and eliminated.

Step A303 , performing order abnormality correction using mean value interpolation.

The excluded data will vacate a position, average the data before and after this position, and fill it to the original position. The formula is as follows:

Step A304, use the normalization method to transform the sample sequence

Using the z-score method (normalization method), transform the sample sequence x ₁ , x ₂ , ..., x _n , the formula is as follows:

The transformed sequence is y ₁ , y ₂ , ..., y _n , the mean of the sequence is 0, the variance is 1, and it is dimensionless.

Get processed data.

Step A400: Send the processed data to a neural network for training.

The steps are the same as those in Embodiment 1, and are omitted here.

Step A500: Calculate the error offset of the smart energy meter according to the environment and working conditions of the smart energy meter.

The steps are the same as those in Embodiment 1, and are omitted here.

Step A600 , using the trained neural network to establish the relationship between the error offsets under different stress conditions.

The present invention takes the stress error curve obtained by one experiment as an example, which are the influence curve of temperature on the error of the electric meter (as shown in Figure 3); the influence curve of humidity on the error of the electric meter (as shown in Figure 4); the influence curve of air pressure on the error of the electric meter ( As shown in Figure 5) and the influence of voltage in different regions on the meter error (as shown in Figure 6).

The above method uses the trained neural network to establish the relationship between typical stress and error, and displays it in the form of images, which can intuitively analyze the errors caused by different stresses.

An embodiment of the second aspect of the present invention provides a smart meter stress error analysis device, including:

The data collection system can obtain the stress data and the error data of the smart meter;

a typical stress analysis system capable of determining the typical stress in the stress data of the smart meter;

a model training module, capable of sending the processed data into a neural network for training to obtain a trained neural network;

The relationship analysis module can use the trained neural network to establish the error data relationship under different stress conditions.

In this embodiment of the present application, the stress error analysis method of a smart meter described in the first embodiment is presented in the form of a computer device, and the BP neural network optimized by LM can be used to analyze the corresponding relationship between the stress and error in the smart meter, The change curve is drawn, so as to realize the establishment of the corresponding relationship between the stress and the error by means of the above method, and achieve the effect of the error analysis of the smart meter.

Yet another embodiment of the present application provides a terminal, including: a memory, a processor, and a computer program stored in the memory and running on the processor, when the processor executes the computer program, the above-mentioned method for analyzing stress error of a smart meter is implemented .

Specifically, the processor may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logical blocks, modules and circuits described in connection with this disclosure. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

Specifically, the processor is connected to the memory through a bus, and the bus may include a path for transferring information. The bus can be a PCI bus or an EISA bus or the like. The bus can be divided into address bus, data bus, control bus and so on.

The memory can be ROM or other types of static storage devices that can store static information and instructions, RAM or other types of dynamic storage devices that can store information and instructions, or EEPROM, CD-ROM or other optical disk storage, optical disk storage ( including compact discs, laser discs, compact discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or capable of carrying or storing desired program code in the form of instructions or data structures and capable of being stored by a computer any other medium taken, but not limited to this.

Optionally, the memory is used to store the code of the computer program for executing the solution of the present application, and the execution is controlled by the processor.

The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned embodiments. Within the scope of knowledge possessed by those of ordinary skill in the technical field, various modifications can be made without departing from the purpose of the present invention. kind of change.

Claims (11)

1. The stress error analysis method for the intelligent ammeter is characterized by comprising the following steps:

acquiring stress data and error data of the intelligent electric meter;

determining typical stress data in the stress data of the intelligent electric meter;

sending the typical stress data and the error data thereof into a neural network for training to obtain a trained neural network;

and establishing error data relationships under different stress conditions by using the trained neural network.

2. The method for analyzing stress error of the smart meter according to claim 1, wherein the step of determining typical stress data in the stress data of the smart meter comprises:

carrying out standardization processing on the data;

calculating the contribution rate of each stress by using the data subjected to the standardization processing;

several stresses with relatively large contribution ratios are selected as typical stresses.

3. The method for analyzing stress error of a smart meter according to claim 2, wherein the sum of the typical stress contributions exceeds 95%.

4. The method for analyzing stress error of the smart meter according to claim 1, wherein after the step of determining typical stress data in the stress data of the smart meter, the method further comprises the steps of:

and processing the typical stress data and the error data thereof, and removing abnormal data to obtain processed data.

5. The method for analyzing stress error of a smart meter according to claim 4, wherein the step of processing the typical stress data and the error data thereof to remove abnormal data and obtain processed data comprises:

arranging error data in the intelligent ammeter according to an acquisition time sequence to obtain x (t), and solving an average value mu and a standard deviation sigma;

rejecting abnormal data in the error data;

correcting sequence abnormality by utilizing average value interpolation;

and transforming the sample sequence of the stress by using a normalization method to obtain processed data.

6. The method for analyzing stress error of a smart meter according to claim 1, wherein the neural network is a BP neural network based on LM optimization.

7. The method for analyzing stress error of a smart meter according to claim 6, wherein the step of sending the processed data to a neural network for training to obtain a trained neural network comprises:

setting the stress parameters as input vectors;

setting the error of the intelligent ammeter as an output vector;

the actual error is set to the desired output vector.

8. The method for analyzing stress error of a smart meter according to claim 1, further comprising:

predicting an estimation error by using the trained neural network;

and adding the estimated error to the reference error of the intelligent electric meter to be used as the error offset of the intelligent electric meter.

9. The utility model provides a smart electric meter stress error analysis device which characterized in that includes:

the data collection module can acquire stress data and error data of the intelligent ammeter;

the typical stress analysis module can determine typical stress data in the stress data of the intelligent ammeter;

the model training module can send the typical stress data and the error data thereof into a neural network for training to obtain a trained neural network;

and the relation analysis module can establish error data relation under different stress conditions by utilizing the trained neural network.

10. A terminal, comprising: memory, processor and computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method of any of claims 1 to 8.

11. A computer-readable medium, characterized in that the computer medium has stored therein computer software, which when executed, can implement the method for analyzing stress error of a smart meter as described in any one of claims 1 to 8.

Images