CN115308674A - Method and system for evaluating epitope running state of automatic verification assembly line of electric energy meter - Google Patents

Abstract

The invention discloses a method and a system for evaluating the epitope running state of an electric energy meter automatic verification assembly line, which comprises the following steps: acquiring historical test data of the epitopes on the verification production line under different running states, removing batch effects of the historical test data on the basis of normal epitope data, and constructing feature vectors of the epitopes under different running states; constructing a twin neural network, updating network parameters by setting boundary values when the twin neural network is trained based on the feature vectors of all epitopes, and constructing a state evaluation model by introducing a state judgment threshold; performing parameter optimization on the state evaluation model by adopting an Ulva gull optimization algorithm and taking the model evaluation reliability as a fitness function to obtain an optimized state evaluation model; and performing state evaluation on the epitope to be tested by adopting the optimized state evaluation model. The method effectively identifies the abnormal state of the epitope caused by performance degradation, and solves the problems of insufficient reliability and high labor cost in the traditional manual investigation.

Description

Method and system for evaluating the operating status of electric energy meter automatic verification assembly line meter position

technical field

The invention relates to the technical field of on-line monitoring of electric power metering, in particular to a method and system for evaluating the running state of a meter position of an automatic verification assembly line of an electric energy meter.

Background technique

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

The automatic verification assembly line of the electric energy meter provides guarantee for the normal operation of the smart electric energy meter. However, during the long-term operation of the assembly line, the frequent access of the intelligent electric energy meter to the meter position will cause deformation of the mechanical crimping terminal of the meter position; at the same time, long-term electrified operation will accelerate the mechanical The rate of oxidation of the material on the surface of the crimped terminal, causing the terminal to corrode. The deformation and corrosion of the mechanical crimping link of the meter will directly affect the reliability of the error test results, and then affect the verification quality of the smart energy meter. However, the current manual periodic detection method cannot respond in time to the abnormal working conditions that occur during the operation and maintenance interval of the pipeline.

Contents of the invention

In order to solve the above problems, the present invention proposes a method and system for evaluating the operation state of epitopes in the automatic verification pipeline of electric energy meters, which can effectively identify the abnormal state of epitopes caused by performance degradation, and overcome the problems of insufficient reliability and high labor costs in traditional manual inspections. To solve the problem, realize the online discrimination of the epitope status of the verification pipeline.

In order to achieve the above object, the present invention adopts the following technical solutions:

In the first aspect, the present invention provides a method for evaluating the operating state of the electric energy meter automatic verification assembly line, including:

Obtain the historical test data of the epitope on the verification pipeline under different operating states, remove the batch effect on the historical test data based on the normal epitope data, and construct the feature vector of each epitope under different operating states;

Construct a twin neural network, update the network parameters by setting boundary values when training the twin neural network based on the eigenvectors of each epitope, and build a state evaluation model by introducing a state judgment threshold;

Using the black tern optimization algorithm, taking the model evaluation reliability as the fitness function, the parameters of the state evaluation model are optimized, and the optimized state evaluation model is obtained;

The optimized state evaluation model is used to evaluate the state of the epitope to be tested.

As an optional implementation, based on the normal epitope data, the average center method is used to remove the batch effect;

为第i个表位第k项试验批次效应去除后的试验结果；

为第k项试验批次效应的误差。Among them, _εik is the test result of the i-th epitope k-th test,

is the test result after removing the test batch effect of the i-th epitope and the k-th test batch effect;

is the error of the batch effect of the kth test.

As an optional implementation, the Siamese neural network is trained using a contrastive loss function L _loss :

Among them, margin is the boundary value, D(x ₁ , x ₂ ) is the similarity measure of samples x ₁ and x ₂ , y is the label corresponding to the support set, and Q is the number of support sets.

As an optional implementation manner, the fitness function K is:

Among them, H ₀ is the observed coincidence rate, a ₁ , a ₂ , and a ₃ respectively represent the number of samples that are actually normal, warning, and abnormal samples that are predicted to be normal, warning, and abnormal samples, S is the total number of samples, and _He Indicates the chance coincidence rate, b ₁ , b ₂ , b ₃ are the real numbers of normal, warning, and abnormal samples respectively, and c ₁ , c ₂ , c ₃ are the numbers of predicted normal, warning, and abnormal samples, respectively.

As an optional implementation manner, the parameter optimization process is to optimize boundary values and state judgment thresholds.

As an optional implementation, a nonlinear control parameter A is proposed in the black tern optimization algorithm:

Among them, f _c is the parameter to control the frequency of A, r is the number of iterations, and R is the maximum number of iterations.

As an optional implementation, the sample to be tested and the known state sample of the epitope to be tested are used to generate a sample pair to be tested, the similarity between the sample to be tested and the known state sample is calculated, and the average value of the similarity under each operating state is taken as For the final similarity, the epitope status of the sample to be tested is judged according to the final similarity and the status judgment threshold.

In the second aspect, the present invention provides a system for evaluating the operating state of the electric energy meter automatic verification assembly line, including:

The training set acquisition module is configured to obtain the historical test data of epitopes in different operating states on the verification pipeline, remove the batch effect on the historical test data based on the normal epitope data, and construct each table under different operating states bit eigenvector;

The model building module is configured to construct a twin neural network, update the network parameters by setting boundary values when training the twin neural network based on the feature vectors of each epitope, and construct a state evaluation model by introducing a state judgment threshold;

The parameter optimization module is configured to use the black tern optimization algorithm and use the model evaluation reliability as the fitness function to optimize the parameters of the state evaluation model to obtain the optimized state evaluation model;

The status evaluation module is configured to use the optimized status evaluation model to evaluate the status of the epitope to be tested.

In a third aspect, the present invention provides an electronic device, including a memory, a processor, and computer instructions stored in the memory and run on the processor. When the computer instructions are executed by the processor, the method described in the first aspect is completed. .

In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the method described in the first aspect is completed.

Compared with prior art, the beneficial effect of the present invention is:

The invention proposes a method and system for evaluating the running state of the automatic verification assembly line of electric energy meters, which uses a twin neural network to distinguish the operating state of the electric energy meter automatic verification assembly line, and solves the problem of small sample learning.

The present invention proposes a method and system for evaluating the operating state of an electric energy meter automatic verification assembly line. The black tern optimization algorithm is used to iteratively optimize the hyperparameter boundary value and threshold introduced by the twin deep neural network and the state judgment mechanism, and the model is improved. Accuracy and reliability of assessments.

The present invention proposes a method and system for evaluating the running state of the electric energy meter automatic verification assembly line, adopts batch effect elimination, eliminates the distribution difference of test data that may exist in different batches of electric energy meters, and constructs a time-series feature with error expectation Feature vector.

Advantages of additional aspects of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention, and the schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations to the present invention.

Fig. 1 is the flow chart of the method for evaluating the running state of the electric energy meter automatic verification assembly line meter position provided by Embodiment 1 of the present invention;

FIG. 2 is a state assessment model diagram provided by Embodiment 1 of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the present invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terminology used here is only for describing specific embodiments, and is not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural, and it should also be understood that the terms "comprising" and "having" and any variations thereof are intended to cover a non-exclusive Comprising, for example, a process, method, system, product, or device comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include steps or units not explicitly listed or for these processes, methods, Other steps or units inherent in a product or equipment.

In the case of no conflict, the embodiments and the features in the embodiments of the present invention can be combined with each other.

Example 1

This embodiment proposes a method for evaluating the running status of the electric energy meter automatic verification assembly line, as shown in Figure 1, including:

Obtain the historical test data of the epitope on the verification pipeline under different operating states, remove the batch effect on the historical test data based on the normal epitope data, and construct the feature vector of each epitope under different operating states;

Construct a twin neural network, update the network parameters by setting boundary values when training the twin neural network based on the eigenvectors of each epitope, and build a state evaluation model by introducing a state judgment threshold;

Using the black tern optimization algorithm, taking the model evaluation reliability as the fitness function, the parameters of the state evaluation model are optimized, and the optimized state evaluation model is obtained;

The optimized state evaluation model is used to evaluate the state of the epitope to be tested.

In this embodiment, the running status includes normal, warning and abnormal. Considering that different batches of electric energy meters may have differences in the distribution of test data, in order to eliminate the impact of batch differences, batch effect removal processing is performed on historical test data, specifically:

Based on the normal epitope data, the "average center method" was used to remove the batch effect:

为第i个表位第k项误差试验批次效应去除后的误差试验结果；Among them, _εik is the error test result measured by the i-th epitope k-th item error test,

is the error test result after removing the batch effect of the i-th epitope k-th error test;

为误差试验批次效应的误差，n为同一检定单元中正常表位的数量，ε_jk为第j个表位第k项误差试验测得的误差结果，其中每个表位进行k项误差试验，k＝1,2,...,10。in,

is the error of the batch effect of the error test, n is the number of normal epitopes in the same test unit, _εjk is the error result measured by the j-th epitope k-item error test, and each epitope is subjected to k-item error tests , k=1,2,...,10.

In this embodiment, based on the test results after the batch effect is removed, the time-series feature vector x of each epitope under normal, alarm, abnormal and other operating states is constructed with the data of T days as the time window;

表示表位i对第v个电能表进行第k项误差试验得到的批次效应去除后的误差结果，t为当前时刻，T天的数据为t与其之前的T-1天共同构成；Among them, m is the number of electric energy meters detected by episite i,

Represents the error result after the batch effect is removed from the kth error test of the vth electric energy meter by episite i, t is the current moment, and the data of T days is composed of t and the previous T-1 day;

为第t天表位i对m个电能表进行第k项误差试验而计算得到的误差期望，

为T-u天前当天表位i对m个电能表进行第k项误差试验而计算得到的误差期望。

is the error expectation calculated by performing the k-th error test on m electric energy meters for the meter position i on the t-th day,

It is the error expectation calculated by performing the k-th error test on m electric energy meters for meter position i on the day before Tu.

In this embodiment, the Siamese neural network is constructed with the feature vectors of each epitope as the training set, and the state evaluation model includes the Siamese neural network and a threshold judgment module; as shown in FIG. 2 .

The twin neural network includes two sub-neural networks with the same model parameters. Its purpose is to measure the similarity of the input samples after the feature vectors extracted by the sub-neural network, so that the features extracted by the sub-neural network are more discriminative; specifically:

(1) Construct an input sample pair (x ₁ , x ₂ ) based on the feature vector of each epitope in the pipeline after the batch effect is removed;

Based on the eigenvectors of each epitope in the pipeline after the batch effect is removed, the training set and the test set are constructed, and some samples are selected from the training set to construct sample pairs to form a support set. Randomly select a data to form a data pair, and then form a support set;

The data composition forms of the support set are: normal and normal, normal and warning, normal and abnormal, warning and warning, warning and abnormal, abnormal and abnormal, so as to obtain the input sample pair (x ₁ , x ₂ ).

(2) The Network module in the twin neural network adopts the 1-D CNN-LSTM sub-network model, based on the 1-D CNN-LSTM sub-network model, the feature vector G(x ₁ of the input sample pair (x ₁ , x ₂ ) is obtained ) and G(x ₂ ):

Take samples x ₁ and x ₂ as input data, and input them into two identical neural network modules Network in the feature extraction model respectively. These two neural network modules have shared parameter weight W and bias b, and the feature extraction module takes the two The input samples x ₁ and x ₂ are respectively mapped to the same feature space of the sub-network, and the feature vectors G(x ₁ ) and G(x ₂ ) of the two input samples x ₁ and x ₂ are obtained.

(3) Similarity calculation;

The similarity measure is used to calculate the distance between two sample feature vectors in a sample pair to evaluate the similarity between two samples. In this embodiment, the Euclidean distance is used, and the input sample pair (x ₁ , x ₂ ) similarity measure is:

D(x ₁ ,x ₂ )＝||G(x ₁ )-G(x ₂ )|| (6)

(4) The twin neural network uses the contrastive loss function L _loss for model training:

Among them, margin is the set boundary value. When the value of D(x ₁ , x ₂ ) is less than the boundary value margin, the weight W of the twin neural network needs to be adjusted; y is the label corresponding to the support set, and Q is the number of support sets .

When the input samples are samples of the same input category of the two sub-networks, the corresponding label is 1, so that the G(x ₁ ) and G(x ₂ ) distance functions of the feature map vectors of the last layer are as close as possible; when the input samples are of different categories When sample, the corresponding label is 0, so that the feature vector distance function is as far away as possible.

In this embodiment, in order to realize the judgment of the epitope state, the threshold judgment module is connected after the twin neural network, and the judgment of the epitope state of the sample to be tested is completed through the state judgment threshold β;

The sample to be tested and the known state sample are used to generate a sample pair to be tested, which is input into the twin neural network to calculate the similarity between the sample to be tested and the known state sample, and the average value of the sample similarity in each state is taken as the final sample Similarity D _W ; if D _W <β, they are of the same type; otherwise, they are of the same type.

In this embodiment, two new hyperparameters, margin and β, are introduced into the state evaluation model based on the twin neural network and the state judgment mechanism. The changes of the two have a greater impact on the evaluation performance of the state evaluation model, so In this embodiment, the sooty tern optimization algorithm is adopted, and the reliability of the model evaluation is used as the fitness function to iteratively optimize the margin and β, so as to improve the reliability of the model state evaluation.

To map the initial position of the sooty tern to the boundary value margin and the state judgment threshold β, the steps are as follows:

(1) Take the reliability K of model evaluation as the fitness function:

a₁、a₂、a₃分别代表实际为正常、告警、异常的样本经过模型预测为正常、告警、异常的样本的个数，S为总样本个数；H_e表示机遇符合率，

其中b₁、b₂、b₃分别为正常、告警、异常样本的真实个数，c₁、c₂、c₃分别为模型预测为正常、告警、异常的样本个数。Among them, _H0 is the observed coincidence rate,

a ₁ , a ₂ , and a ₃ respectively represent the number of samples that are actually normal, warning, and abnormal after being predicted by the model to be normal, warning, and abnormal. S is the total number of samples; He represents the chance _coincidence rate,

Among them, b ₁ , b ₂ , and b ₃ are the real numbers of normal, warning, and abnormal samples, respectively, and c ₁ , c ₂ , and c ₃ are the numbers of normal, warning, and abnormal samples predicted by the model.

(2) Sooty Tern Optimization Algorithm (STO), migration and attacking prey are unique behaviors of sooty terns, during migration, sooty terns move to the strongest sooty tern in the group, then, other sooty terns start To update their initial positions, it is necessary to avoid collisions between sooty terns; specifically, the update method of sooty tern optimization algorithm (STO) parameters is as follows:

① Initialize the population size num, the maximum number of iterations R, and randomly initialize the initial position of the sooty tern;

②Evaluate the fitness function: Calculate the fitness of each sooty tern's position according to the fitness function, find out the optimal fitness value by comparison, and determine the best position P _best (r) of the population;

③The sooty tern performs the migration operation:

Among them, C _s represents the new position that will not collide with other terns; P _s (r) represents the current position of the sooty tern, A is the movement mode of the sooty tern in a given space; r represents the number of iterations; R represents The maximum number of iterations; f _c is used to control the frequency of A, where f _c takes a value of 2 to control A from 2 to 0; B is a random variable; M _s is the process of moving the current position to the optimal position; P _best (r) is the global optimal position of the sooty tern; randvalue is a random number between 0 and 1; L _s is the trajectory of updating the current position to the optimal position.

However, the linear control parameters cannot characterize the actual convergence process, which is nonlinear. Therefore, this embodiment proposes a nonlinear control parameter A:

In this method, the value of A presents a non-linear change trend in the process of decreasing, which can better improve the global optimization ability. Each iteration can not only avoid the position conflict between black terns, but also better balance Explore and develop.

④ Black tern attack operation:

Sooty terns can increase flight height through wings during migration, and can also adjust speed and attack angle. When attacking prey, the sooty tern's hovering behavior in the air can be defined:

ω是定义螺旋形状的常数，本实施例设置为1，e是自然对数的底数；Among them, x′, y′, z′ are the positions of simulated terns circling in three-dimensional space, λ is the radius of the spiral; θ is a random angle in [0,2π]; w ₀ ,

ω is the constant that defines spiral shape, and the present embodiment is set to 1, and e is the base number of natural logarithm;

⑤The update location of Black Tern:

P _s (r+1)=(L _s ×(x′+y′+z′))×P _best (r) (12)

Among them, P _s (r+1) is the updated position of the sooty tern, L _s is the trajectory of updating the current position to the optimal position, and P _best (r) is the best position of the population;

⑥ Calculate the fitness value of each sooty tern individual position, and record the global optimal value:

表示乌燕鸥位置更新后模型评估的可靠度，

为乌燕鸥位置更新后的观察符合率，

a_1更、a_2更、a_3更分别为乌燕鸥位置更新后实际为正常、告警、异常的样本经过模型预测为正常、告警、异常的样本的个数，S_更为乌燕鸥位置更新后的总样本个数；

表示乌燕鸥位置更新后的机遇符合率，

其中b_1更、b_2更、b_3更分别为正常、告警、异常样本的真实个数，c_1更、c_2更、c_3更分别为模型预测为正常、告警、异常的样本个数。in,

Indicates the reliability of the model evaluation after the Sooty Tern's position is updated,

is the observed coincidence rate after the location update of the sooty tern,

a _{1 update} , a _{2 update} , and a _{3 update} are respectively the number of samples that are actually normal, warning, and abnormal after the position update of the sooty tern, and the number of samples that are predicted to be normal, warning, and abnormal by the model, and S _is the position of the sooty tern The total number of samples after the update;

Indicates the chance coincidence rate after the Sooty Tern's position is updated,

Among them, b _{1 update} , b _{2 update} , and b _{3 update} are the real numbers of normal, warning, and abnormal samples respectively, and c _{1 update} , c _{2 update} , and c _{3 update} are the number of samples predicted to be normal, warning, and abnormal by the model respectively. .

Equation (13) is the specific calculation process of taking the updated individual position of Sooty Tern (that is, at the time of r+1 iterations) into Equation (8) to calculate the fitness of each individual and select the optimal fitness.

Find the global optimum of fitness values in a population of sooty terns, and record the location:

[K _best P _best (r+1)]＝max(K) (14)

⑦Judge whether the reliability is greater than the set threshold: if yes, use the boundary value margin and threshold β of the current sooty tern position map as the best hyperparameters of the model; if not, update the parameters of the sooty tern optimization algorithm and the best position, repeat steps ③-⑦ until the maximum number of iterations is met.

In this embodiment, based on the trained state evaluation model, the feature vector of the epitope to be tested is used as input, and the state of each epitope in the pipeline is discriminated.

Example 2

This embodiment provides a system for evaluating the running state of the electric energy meter automatic verification assembly line, including:

The training set acquisition module is configured to obtain the historical test data of epitopes in different operating states on the verification pipeline, remove the batch effect on the historical test data based on the normal epitope data, and construct each table under different operating states bit eigenvector;

The model building module is configured to construct a twin neural network, update the network parameters by setting boundary values when training the twin neural network based on the feature vectors of each epitope, and construct a state evaluation model by introducing a state judgment threshold;

The parameter optimization module is configured to use the black tern optimization algorithm and use the model evaluation reliability as the fitness function to optimize the parameters of the state evaluation model to obtain the optimized state evaluation model;

The status evaluation module is configured to use the optimized status evaluation model to evaluate the status of the epitope to be tested.

It should be noted here that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that, as a part of the system, the above-mentioned modules can be executed in a computer system such as a set of computer-executable instructions.

In further embodiments, there is also provided:

An electronic device includes a memory, a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the method described in Embodiment 1 is completed. For the sake of brevity, details are not repeated here.

It should be understood that in this embodiment, the processor can be a central processing unit CPU, and the processor can also be other general-purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic devices , discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.

The memory may include read-only memory and random access memory, and provide instructions and data to the processor, and a part of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

A computer-readable storage medium is used for storing computer instructions, and when the computer instructions are executed by a processor, the method described in Embodiment 1 is completed.

The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software module may be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware. To avoid repetition, no detailed description is given here.

Those skilled in the art can appreciate that the units of the examples described in this embodiment, that is, the algorithm steps, can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be regarded as exceeding the scope of the present application.

Although the specific implementation of the present invention has been described above in conjunction with the accompanying drawings, it does not limit the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, those skilled in the art do not need to pay creative work Various modifications or variations that can be made are still within the protection scope of the present invention.

Claims (10)

1. The method for evaluating the epitope running state of the automatic verification assembly line of the electric energy meter is characterized by comprising the following steps of:

acquiring historical test data of the epitopes on the verification production line under different running states, removing batch effects of the historical test data on the basis of normal epitope data, and constructing feature vectors of the epitopes under different running states;

constructing a twin neural network, updating network parameters by setting a boundary value when training the twin neural network based on the characteristic vector of each epitope, and constructing a state evaluation model by introducing a state judgment threshold;

performing parameter optimization on the state evaluation model by adopting an Ulva gull optimization algorithm and taking the model evaluation reliability as a fitness function to obtain an optimized state evaluation model;

and performing state evaluation on the epitope to be tested by adopting the optimized state evaluation model.

2. The method for evaluating the epitope running state of the automatic verification assembly line of the electric energy meter according to claim 1, characterized in that the batch effect removal is performed by adopting an average center method based on normal epitope data;

wherein epsilon _ik Test results for the kth test for the ith epitope，

The test result after the kth test batch effect of the ith epitope is removed;

error for the k-th trial batch effect.

3. The method for evaluating the epitope running state of the automatic verification assembly line of an electric energy meter according to claim 1, wherein the twin neural network adopts a contrast loss function L _loss Training is carried out:

wherein margin is a boundary value, D (x) ₁ ,x ₂ ) Is a sample x ₁ 、x ₂ Y is the label corresponding to the support set, and Q is the number of the support sets.

4. The method for evaluating the epitope running state of the automatic verification assembly line of the electric energy meter according to claim 1, wherein the fitness function K is as follows:

wherein H ₀ To observe the coincidence rate, a ₁ 、a ₂ 、a ₃ Respectively representing the number of samples which are actually normal, alarm and abnormal and are predicted to be normal, alarm and abnormal, S is the total number of samples, H _e Representing chance coincidence rate, b ₁ 、b ₂ 、b ₃ The actual number of normal, alarm and abnormal samples, c ₁ 、c ₂ 、c ₃ The numbers of samples predicted to be normal, alarm and abnormal are respectively.

5. The method for evaluating the epitope running state of the automatic verification assembly line of an electric energy meter according to claim 1, wherein the parameter optimizing process is optimizing a boundary value and a state judgment threshold value.

6. The method for evaluating the epitope running state of the automatic verification assembly line of the electric energy meter according to claim 1, wherein a nonlinear control parameter A is provided in the Woofer optimization algorithm:

wherein, f _c For the parameters controlling the frequency a, R is the number of iterations and R is the maximum number of iterations.

7. The method for evaluating the epitope running state of the electric energy meter automatic verification assembly line according to claim 1, characterized in that a pair of samples to be tested is generated by a sample to be tested of the epitope to be tested and a sample in a known state, the similarity between the sample to be tested and the sample in the known state is calculated, the mean value of the similarities in all running states is taken as the final similarity, and the epitope state judgment of the sample to be tested is carried out according to the final similarity and a state judgment threshold.

8. Electric energy meter automated verification assembly line epitope running state evaluation system which characterized in that includes:

the training set acquisition module is configured to acquire historical test data of the epitopes on the verification production line in different running states, remove batch effects on the historical test data on the basis of normal epitope data, and construct feature vectors of the epitopes in different running states;

the model building module is configured to build a twin neural network, update network parameters by setting boundary values when the twin neural network is trained on the basis of the feature vectors of all epitopes, and build a state evaluation model by introducing a state judgment threshold;

the parameter optimizing module is configured to adopt an gull-shaped optimization algorithm, take the model evaluation reliability as a fitness function, and conduct parameter optimizing on the state evaluation model to obtain an optimized state evaluation model;

and the state evaluation module is configured to carry out state evaluation on the epitope to be tested by adopting the optimized state evaluation model.

9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the method of any of claims 1-7.

10. A computer-readable storage medium storing computer instructions which, when executed by a processor, perform the method of any one of claims 1 to 7.

Images