Charging pile grouping method and system based on multilayer perceptron neural network

Abstract

The invention relates to a charging pile grouping method and system based on a multilayer perceptron neural network, and the method specifically comprises the steps: calculating a boundary condition which needs to be satisfied by an individual charging pile error coefficient under a preset misjudgment rate and a missed judgment rate when a preset measurement guarantee scheme is adopted to determine an out-of-tolerance individual in a to-be-detected charging pile group; in combination with existing charging pile data, a pre-trained multi-layer perceptron neural network is adopted to obtain an incidence relation between the charging pile explicit tag and the error coefficient; analyzing the explicit labels of the to-be-grouped charging piles by using the obtained incidence relation, predicting error coefficients of the to-be-grouped charging piles, namely inputting the explicit labels of the to-be-grouped charging piles into a pre-trained multilayer perceptron neural network model, and calculating and outputting predicted values of the error coefficients of the charging piles; and grouping the to-be-grouped charging piles according to the boundary conditions of the error coefficients of the individual charging piles and the error coefficients of the to-be-grouped charging piles.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle charging pile verification and grouping, and in particular to a charging pile grouping method and system based on a multi-layer perceptron neural network. Background Art

[0002] The rapid growth of the scale of charging piles has brought great challenges to the calibration work. The current manual calibration method has become increasingly difficult to meet the needs of comprehensive calibration of charging piles. To this end, a large number of researchers have begun to explore charging pile error analysis and calibration technologies based on data-driven methods, such as the calculation method of charging pile error coefficients based on energy conservation, and the calibration scheme based on the interaction between standard vehicles and charging piles. However, the existing technical solutions are difficult to effectively guarantee the accuracy of error analysis, and it is also difficult to achieve comprehensive and dynamic error control.

[0003] The metrology assurance program is to use verification standards and transfer standards to comprehensively assess the verification and testing system of the metrology room, use mathematical statistics methods to control the verification quality of the metrology units participating in the metrology assurance program, and analyze and evaluate the total uncertainty of the verification process, timely discover problems, feedback information, minimize the transfer error and ensure the quality of the value transfer. It is an important technology in the field of metrology that uses closed-loop feedback control ideas to achieve error control. It is widely used in the verification and error control of resistance standards, battery standards, etc. Applying it to the verification of charging piles can effectively achieve comprehensive and dynamic error control and judgment of out-of-tolerance piles. The metrology assurance program requires that the objects involved in the verification have the same error distribution, otherwise it is difficult to ensure the effectiveness of the program. In actual charging piles, it is almost difficult to ensure that all charging piles have the same error distribution, so it is inevitable that the metrology assurance program will produce misjudgments and missed judgments. If the error coefficients of the applied charging piles are too different, it may lead to a large misjudgment rate and missed judgment rate, and the application effect of the metrology assurance program will be very poor.

[0004] In the prior art CN118011308B, a charging pile metering misalignment detection method, system, terminal and medium are disclosed, which establish a metering error identification model based on the collected charging station power measurement values, the power measurement values ​​of each charging pile, and the AC / DC conversion efficiency fitting value of each charging pile; use a recursive damped least squares algorithm to solve the metering error identification model to obtain the metering error of each charging pile; and judge whether the metering error exceeds the metering misalignment threshold.

[0005] In the prior art, error calculation is performed on each charging pile separately. Although it can ensure that each charging pile can meet the error range, it cannot ensure that the error distribution in the same batch of charging piles is the same. In this way, the accuracy of the detection results cannot be guaranteed in the measurement assurance scheme. Therefore, the charging piles need to be grouped to ensure that the same batch of charging piles have the same error distribution. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention proposes a charging pile grouping method and system based on a multi-layer perceptron neural network.

[0007] The technical solution of the present invention is as follows:

[0008] On the one hand, the present invention proposes a charging pile grouping method based on a multi-layer perceptron neural network, and the specific steps include:

[0009] When using the preset metering assurance scheme to determine the out-of-tolerance individuals in the group of charging piles to be tested, calculate the boundary conditions that the error coefficient of the individual charging pile must meet under the preset misjudgment rate and missed judgment rate;

[0010] Combined with the existing charging pile data, a pre-trained multi-layer perceptron neural network is used to obtain the correlation between the charging pile explicit label and the error coefficient;

[0011] The obtained association relationship is used to analyze the explicit labels of the charging piles to be grouped, and the error coefficients of the charging piles to be grouped are predicted, that is, the explicit labels of the charging piles to be grouped are input into the pre-trained multi-layer perceptron neural network model, and the predicted values ​​of the error coefficients of the charging piles are calculated and output;

[0012] The charging piles to be grouped are grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of the charging piles to be grouped.

[0013] As a preferred implementation, when the preset measurement guarantee scheme is used to determine the out-of-tolerance individuals in the charging pile group to be tested, the boundary condition steps to be satisfied by calculating the error coefficient of the individual charging pile under the preset misjudgment rate and missed judgment rate are specifically as follows:

[0014] The error coefficient x of each charging pile i The mean value x t As the virtual verification standard value of the measurement assurance plan, the conditional probability is calculated:

[0015] P1≤γ1;

[0016] P2≤γ2;

[0017] Wherein, γ1 and γ2 are the maximum allowable values ​​of the misjudgment rate and missed judgment rate when the measurement assurance scheme is adopted, P1 is the probability that the mean error coefficient of each charging pile passes the t-test and the error coefficient of each charging pile has an out-of-tolerance, and P2 is the probability that the mean error coefficient of each charging pile fails the t-test and the error coefficient of each charging pile does not have an out-of-tolerance;

[0018] According to the conditional probability calculation of the above two equations, the boundary conditions of the error coefficient of the individual charging pile are obtained to meet the requirements of the misjudgment rate and missed judgment rate:

[0019] f(x i )≤0;

[0020] In the formula, the f() function represents the functional relationship that the error coefficient of each charging pile needs to satisfy.

[0021] As a preferred implementation, the step of using a multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient is specifically as follows:

[0022] The explicit label of the charging pile is used as an input feature, wherein the explicit label includes manufacturer, production date, model, batch, installation location, indoor / outdoor, AC / DC, monthly usage frequency, and monthly charging power;

[0023] Perform one-hot encoding on the categorical variables in the display labels and standardize the numerical variables in the display labels to obtain the preprocessed input feature vectors;

[0024] A pre-trained multi-layer perceptron neural network model is used, with the pre-processed input feature vector as input, to output the predicted value of the charging pile error coefficient.

[0025] As a preferred implementation, the step of grouping the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging pile and the error coefficient of each charging pile to be grouped is specifically as follows:

[0026] The prediction error coefficient of each charging pile is taken as input, and the prediction error coefficient of each charging pile is clustered using the Gaussian mixture model; the Gaussian probability density function of the error coefficient of each charging pile is calculated, specifically:

[0027]

[0028] In the formula, z k is the hidden variable of the kth Gaussian distribution (k=1,2,…,Q, Q is the number of Gaussian distributions), z k =1 means that charging pile j belongs to the kth Gaussian distribution, zk=0 means it does not; Σ k , μ k are the covariance matrix and mean of the kth Gaussian component respectively; y j Indicates the error coefficient of each charging pile;

[0029] Calculate the posterior probability that the error coefficient of the jth charging pile belongs to the kth Gaussian distribution. The specific calculation formula is as follows:

[0030]

[0031] In the formula, π l is the weight of the lth Gaussian distribution, satisfying

[0032] Then update the model parameters π k ,Σ k , μ k :

[0033]

[0034] Finally, according to the posterior probability P(z k |y j ), the error coefficient of each charging pile is assigned to the group with the maximum a posteriori probability, thus completing the grouping of the charging piles to be grouped.

[0035] On the other hand, the present invention proposes a charging pile grouping system based on a multi-layer perceptron neural network, comprising:

[0036] The boundary condition calculation module calculates the boundary conditions that the error coefficient of the individual charging pile needs to meet under the preset misjudgment rate and missed judgment rate when determining the out-of-tolerance individual in the charging pile group to be tested using the preset measurement assurance scheme;

[0037] The correlation calculation module combines the existing charging pile data and uses a pre-trained multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient;

[0038] The error coefficient calculation module uses the obtained association relationship to analyze the explicit labels of the charging piles to be grouped, and predicts the error coefficients of the charging piles to be grouped, that is, the explicit labels of the charging piles to be grouped are input into the pre-trained multi-layer perceptron neural network model, and the predicted values ​​of the error coefficients of the charging piles are calculated and output;

[0039] The charging pile grouping module groups the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of each charging pile to be grouped.

[0040] As a preferred implementation, when the preset measurement guarantee scheme is used to determine the out-of-tolerance individuals in the charging pile group to be tested, the boundary condition steps to be satisfied by calculating the error coefficient of the individual charging pile under the preset misjudgment rate and missed judgment rate are specifically as follows:

[0041] The error coefficient x of each charging pile i The mean value x t As the virtual verification standard value of the measurement assurance plan, the conditional probability is calculated:

[0042] P1≤γ1;

[0043] P2≤γ2;

[0044] Where γ1 and γ2 are the maximum values ​​of the misjudgment rate and missed judgment rate when the measurement assurance scheme is adopted, P1 is the probability of an error coefficient exceeding the tolerance of each charging pile, and P2 is the probability of no error coefficient exceeding the tolerance of each charging pile.

[0045] According to the conditional probability calculation of the above two equations, the boundary conditions of the error coefficient of the individual charging pile are obtained to meet the requirements of the misjudgment rate and missed judgment rate:

[0046] f(x i )≤0;

[0047] In the formula, the f() function represents the functional relationship that the error coefficient of each charging pile needs to satisfy.

[0048] As a preferred implementation, the step of using a multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient is specifically as follows:

[0049] The explicit label of the charging pile is used as an input feature, wherein the explicit label includes manufacturer, production date, model, batch, installation location, indoor / outdoor, AC / DC, monthly usage frequency, and monthly charging power;

[0050] Perform one-hot encoding on the categorical variables in the display labels and standardize the numerical variables in the display labels to obtain the preprocessed input feature vectors;

[0051] A pre-trained multi-layer perceptron neural network model is used, with the pre-processed input feature vector as input, to output the predicted value of the charging pile error coefficient.

[0052] As a preferred implementation, the step of grouping the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging pile and the error coefficient of each charging pile to be grouped is specifically as follows:

[0053] The prediction error coefficient of each charging pile is taken as input, and the prediction error coefficient of each charging pile is clustered using the Gaussian mixture model; the Gaussian probability density function of the error coefficient of each charging pile is calculated, specifically:

[0054]

[0055] In the formula, z k is the hidden variable of the kth Gaussian distribution (k=1,2,…,Q, Q is the number of Gaussian distributions), z k =1 means that charging pile j belongs to the kth Gaussian distribution, zk=0 means it does not; Σ k , μ k are the covariance matrix and mean of the kth Gaussian component respectively; y j Indicates the error coefficient of each charging pile;

[0056] Calculate the posterior probability that the error coefficient of the jth charging pile belongs to the kth Gaussian distribution. The specific calculation formula is as follows:

[0057]

[0058] In the formula, π l is the weight of the lth Gaussian distribution, satisfying

[0059] Then update the model parameters π k ,Σ k , μ k :

[0060]

[0061] Finally, according to the posterior probability P(z k |y j ), the error coefficient of each charging pile is assigned to the group with the maximum a posteriori probability, thus completing the grouping of the charging piles to be grouped.

[0062] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a charging pile grouping method based on a multi-layer perceptron neural network as described in any embodiment of the present invention.

[0063] The present invention has the following beneficial effects:

[0064] 1. The present invention divides charging piles with similar conditions into the same group by calculating the boundary conditions of the error coefficient of each individual charging pile, so as to facilitate the subsequent calculation of the error coefficient.

[0065] 2. The present invention is based on the display label data of each charging pile, and normalizes the data, and calculates the error coefficient of each charging pile through a pre-trained multi-layer perceptron neural network, so that the calculation result is more accurate.

[0066] 3. The present invention uses a Gaussian mixture model to cluster the error coefficient data of each charging pile, and groups the charging piles to be grouped according to the clustering results, and divides the charging piles with similar error coefficients into the same group. This ensures that when the metering assurance scheme is applied in groups with similar error coefficients, the misjudgment rate and missed judgment rate can be controlled to a certain extent, thereby ensuring its effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0068] 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.

[0069] 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.

[0070] 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.

[0071] 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.

[0072] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0073] Embodiment 1:

[0074] See also Figure 1 , a charging pile grouping method based on a multi-layer perceptron neural network, the specific steps include:

[0075] Step S101, when using a preset measurement assurance scheme to determine the out-of-tolerance individual in the charging pile group to be tested, calculate the boundary conditions that the error coefficient of the individual charging pile needs to meet under the preset misjudgment rate and missed judgment rate;

[0076] Step S102, combining the existing charging pile data, using a pre-trained multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient;

[0077] Step S103, using the obtained association relationship to analyze the explicit labels of the charging piles to be grouped, predicting the error coefficients of the charging piles to be grouped, that is, inputting the explicit labels of the charging piles to be grouped into a pre-trained multi-layer perceptron neural network model, and calculating and outputting the predicted values ​​of the error coefficients of the charging piles;

[0078] Step S104 , grouping the charging piles to be grouped according to the boundary conditions of the error coefficients of the individual charging piles and the error coefficients of the charging piles to be grouped.

[0079] In this embodiment, in actual application, in order to ensure the accuracy of the metering assurance scheme, it is first necessary to meet the requirement that the applicable objects of the metering assurance scheme are charging piles with similar error coefficients; therefore, it is necessary to group the charging piles to a certain extent, and the application of the metering assurance scheme in groups with similar error coefficients can control the misjudgment rate and missed judgment rate to a certain extent, thereby ensuring its effectiveness.

[0080] However, it is impossible to obtain the error coefficients of all charging piles in practice to determine the misjudgment rate and missed judgment rate of the metering assurance scheme. To this end, the present invention proposes to construct the correlation law between the explicit indicators and the error coefficients of the charging piles by means of small sample training, so as to directly predict the error coefficients through the explicit indicators, and then group the charging piles, so as to ensure that the misjudgment rate and missed judgment rate of the metering assurance scheme applied in the grouped charging piles are limited to an acceptable range.

[0081] As a preferred implementation of this embodiment, the step of calculating the boundary conditions that the error coefficient of the individual charging pile needs to meet under the preset false positive rate and missed positive rate is specifically as follows:

[0082] The error coefficient x of each charging pile i The mean value x t As the virtual verification standard value of the measurement assurance plan, the conditional probability is calculated:

[0083] P1≤γ1;

[0084] P2≤γ2;

[0085] Where γ1 and γ2 are the maximum values ​​of the misjudgment rate and missed judgment rate when the measurement assurance scheme is adopted, P1 is the probability of an error coefficient exceeding the tolerance of each charging pile, and P2 is the probability of no error coefficient exceeding the tolerance of each charging pile.

[0086] In this embodiment, the settings of γ1 and γ2 need to be set by on-site inspection personnel according to actual conditions.

[0087] According to the conditional probability calculation of the above two equations, the boundary conditions of the error coefficient of the individual charging pile are obtained to meet the requirements of the misjudgment rate and missed judgment rate:

[0088] f(x i )≤0;

[0089] In the formula, the f() function represents the functional relationship that the error coefficient of each charging pile needs to satisfy.

[0090] As a preferred implementation of this embodiment, the step of using a multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient is specifically as follows:

[0091] The explicit label of the charging pile is used as an input feature, wherein the explicit label includes manufacturer, production date, model, batch, installation location, indoor / outdoor, AC / DC, monthly usage frequency, and monthly charging power;

[0092] Perform one-hot encoding on the categorical variables in the display labels and standardize the numerical variables in the display labels to obtain the preprocessed input feature vectors;

[0093] A pre-trained multi-layer perceptron neural network model is used, with the pre-processed input feature vector as input, to output the predicted value of the charging pile error coefficient.

[0094] In this embodiment, the parameters for pre-training the multi-layer perceptron neural network model are set as follows:

[0095] A multi-layer perceptron neural network is used, where the input layer corresponds to the processed explicit label, and the hidden layer contains one or more layers, and the number of nodes in each layer is h1,h2,…,h k , the activation function is the rectified linear unit (ReLU), and the output layer is the predicted value of the error coefficient;

[0096] The mean square error is used as the loss function, and the adaptive moment estimation optimizer (Adam optimizer) is used for training until the loss function converges or the early stopping condition is met.

[0097] As a preferred implementation of this embodiment, the step of using the obtained association relationship to analyze the explicit labels of each charging pile to be grouped and predicting the error coefficient of each charging pile to be grouped is specifically as follows:

[0098] The explicit labels of the charging piles to be grouped are input into the trained multi-layer perceptron neural network model, and the predicted value of the error coefficient of each charging pile is calculated and output.

[0099] In this embodiment, the output error coefficient y j As the prediction error coefficient of the charging pile, y j =g(Y), where Y={Y1,Y2,…,Y M} is the explicit label set of the charging piles to be grouped, where Y j (j=1, 2, ..., M, M is the number of charging piles to be grouped) is the input feature vector of the jth charging pile to be grouped after preprocessing.

[0100] As a preferred implementation of this embodiment, the step of grouping the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of the charging piles to be grouped is specifically as follows:

[0101] The prediction error coefficient of each charging pile is taken as input, and the prediction error coefficient of each charging pile is clustered using the Gaussian mixture model; the Gaussian probability density function of the error coefficient of each charging pile is calculated, specifically:

[0102]

[0103] In the formula, z k is the hidden variable of the kth Gaussian distribution (k=1,2,…,Q, Q is the number of Gaussian distributions), z k =1 means that charging pile j belongs to the kth Gaussian distribution, zk=0 means it does not; Σ k , μ k are the covariance matrix and mean of the kth Gaussian component respectively; y j Indicates the error coefficient of each charging pile;

[0104] Calculate the posterior probability that the error coefficient of the jth charging pile belongs to the kth Gaussian distribution. The specific calculation formula is as follows:

[0105]

[0106] In the formula, π l is the weight of the lth Gaussian distribution, satisfying

[0107] Then update the model parameters π k ,Σ k , μ k :

[0108]

[0109] Finally, according to the posterior probability P(z k |y j ), the error coefficient of each charging pile is assigned to the group with the maximum a posteriori probability, thus completing the grouping of the charging piles to be grouped.

[0110] Embodiment 2:

[0111] A charging pile grouping system based on a multi-layer perceptron neural network, comprising:

[0112] The boundary condition calculation module calculates the boundary conditions that the error coefficient of the individual charging pile needs to meet under the preset misjudgment rate and missed judgment rate when determining the out-of-tolerance individual in the charging pile group to be tested using the preset measurement assurance scheme;

[0113] The correlation calculation module combines the existing charging pile data and uses a pre-trained multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient;

[0114] The error coefficient calculation module uses the obtained association relationship to analyze the explicit labels of the charging piles to be grouped, and predicts the error coefficients of the charging piles to be grouped, that is, the explicit labels of the charging piles to be grouped are input into the pre-trained multi-layer perceptron neural network model, and the predicted values ​​of the error coefficients of the charging piles are calculated and output;

[0115] The charging pile grouping module groups the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of each charging pile to be grouped.

[0116] As a preferred implementation of this embodiment, the step of calculating the boundary conditions that the error coefficient of the individual charging pile needs to meet under the preset false positive rate and missed positive rate is specifically as follows:

[0117] The error coefficient x of each charging pile i The mean value x t As the virtual verification standard value of the measurement assurance plan, the conditional probability is calculated:

[0118] P1≤γ1;

[0119] P2≤γ2;

[0120] Where γ1 and γ2 are the maximum values ​​of the misjudgment rate and missed judgment rate when the measurement assurance scheme is adopted, P1 is the probability of an error coefficient exceeding the tolerance of each charging pile, and P2 is the probability of no error coefficient exceeding the tolerance of each charging pile.

[0121] According to the conditional probability calculation of the above two equations, the boundary conditions of the error coefficient of the individual charging pile are obtained to meet the requirements of the misjudgment rate and missed judgment rate:

[0122] f(x i )≤0;

[0123] In the formula, the f() function represents the functional relationship that the error coefficient of each charging pile needs to satisfy.

[0124] As a preferred implementation of this embodiment, the step of using a multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient is specifically as follows:

[0125] The explicit label of the charging pile is used as an input feature, wherein the explicit label includes manufacturer, production date, model, batch, installation location, indoor / outdoor, AC / DC, monthly usage frequency, and monthly charging power;

[0126] Perform one-hot encoding on the categorical variables in the display labels and standardize the numerical variables in the display labels to obtain the preprocessed input feature vectors;

[0127] A pre-trained multi-layer perceptron neural network model is used, with the pre-processed input feature vector as input, to output the predicted value of the charging pile error coefficient.

[0128] As a preferred implementation of this embodiment, the step of using the obtained association relationship to analyze the explicit labels of each charging pile to be grouped and predicting the error coefficient of each charging pile to be grouped is specifically as follows:

[0129] The explicit labels of the charging piles to be grouped are input into the trained multi-layer perceptron neural network model, and the predicted value of the error coefficient of each charging pile is calculated and output.

[0130] As a preferred implementation of this embodiment, the step of grouping the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of the charging piles to be grouped is specifically as follows:

[0131] The prediction error coefficient of each charging pile is taken as input, and the prediction error coefficient of each charging pile is clustered using the Gaussian mixture model; the Gaussian probability density function of the error coefficient of each charging pile is calculated, specifically:

[0132]

[0133] In the formula, z k is the hidden variable of the kth Gaussian distribution (k=1,2,…,Q, Q is the number of Gaussian distributions), z k =1 means that charging pile j belongs to the kth Gaussian distribution, zk=0 means it does not; Σ k , μ k are the covariance matrix and mean of the kth Gaussian component respectively; y j Indicates the error coefficient of each charging pile;

[0134] Calculate the posterior probability that the error coefficient of the jth charging pile belongs to the kth Gaussian distribution. The specific calculation formula is as follows:

[0135]

[0136] In the formula, π l is the weight of the lth Gaussian distribution, satisfying

[0137] Then update the model parameters π k ,Σ k , μ k :

[0138]

[0139] Finally, according to the posterior probability P(z k |y j), the error coefficient of each charging pile is assigned to the group with the maximum a posteriori probability, thus completing the grouping of the charging piles to be grouped.

[0140] 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

1. A charging pile grouping method based on a multi-layer perceptron neural network, characterized in that: The specific steps include: When using the preset metering assurance scheme to determine the out-of-tolerance individuals in the group of charging piles to be tested, calculate the boundary conditions that the error coefficient of the individual charging pile must meet under the preset misjudgment rate and missed judgment rate; Combined with the existing charging pile data, a pre-trained multi-layer perceptron neural network is used to obtain the correlation between the charging pile explicit label and the error coefficient; The obtained association relationship is used to analyze the explicit labels of the charging piles to be grouped, and the error coefficients of the charging piles to be grouped are predicted, that is, the explicit labels of the charging piles to be grouped are input into the pre-trained multi-layer perceptron neural network model, and the predicted values ​​of the error coefficients of the charging piles are calculated and output; The charging piles to be grouped are grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of the charging piles to be grouped.

2. A charging pile grouping method based on a multi-layer perceptron neural network according to claim 1, characterized in that: When the preset metering assurance scheme is used to determine the out-of-tolerance individuals in the group of charging piles to be tested, the boundary condition steps to be satisfied by calculating the error coefficient of the individual charging pile under the preset misjudgment rate and missed judgment rate are specifically as follows: The error coefficient x of each charging pile i The mean value x t As the virtual verification standard value of the measurement assurance plan, the conditional probability is calculated: P1≤γ1; P2≤γ2; Wherein, γ1 and γ2 are the maximum allowable values ​​of the misjudgment rate and missed judgment rate when the measurement assurance scheme is adopted, P1 is the probability that the mean error coefficient of each charging pile passes the t-test and the error coefficient of each charging pile has an out-of-tolerance, and P2 is the probability that the mean error coefficient of each charging pile fails the t-test and the error coefficient of each charging pile does not have an out-of-tolerance; According to the conditional probability calculation of the above two equations, the boundary conditions of the error coefficient of the individual charging pile are obtained to meet the requirements of the misjudgment rate and missed judgment rate: f(x i )≤0; In the formula, the f() function represents the functional relationship that the error coefficient of each charging pile needs to satisfy.

3. The charging pile grouping method based on a multi-layer perceptron neural network according to claim 1 is characterized in that: The step of using a multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient is specifically as follows: The explicit label of the charging pile is used as an input feature, wherein the explicit label includes manufacturer, production date, model, batch, installation location, indoor / outdoor, AC / DC, monthly usage frequency, and monthly charging power; Perform one-hot encoding on the categorical variables in the display labels and standardize the numerical variables in the display labels to obtain the preprocessed input feature vectors; A pre-trained multi-layer perceptron neural network model is used, with the pre-processed input feature vector as input, to output the predicted value of the charging pile error coefficient.

4. The charging pile grouping method based on a multi-layer perceptron neural network according to claim 1 is characterized in that: The step of grouping the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging pile and the error coefficient of each charging pile to be grouped is specifically as follows: The prediction error coefficient of each charging pile is taken as input, and the prediction error coefficient of each charging pile is clustered using the Gaussian mixture model; the Gaussian probability density function of the error coefficient of each charging pile is calculated, specifically: In the formula, z k is the hidden variable of the kth Gaussian distribution (k=1,2,…,Q, Q is the number of Gaussian distributions), z k =1 means that charging pile j belongs to the kth Gaussian distribution, zk=0 means it does not; Σ k , μ k are the covariance matrix and mean of the kth Gaussian component respectively; y j Indicates the error coefficient of each charging pile; Calculate the posterior probability that the error coefficient of the jth charging pile belongs to the kth Gaussian distribution. The specific calculation formula is as follows: In the formula, π l is the weight of the lth Gaussian distribution, satisfying Then update the model parameters π k ,Σ k , μ k : Finally, according to the posterior probability P(z k |y j ), the error coefficient of each charging pile is assigned to the group with the maximum a posteriori probability, thus completing the grouping of the charging piles to be grouped.

5. A charging pile grouping system based on a multi-layer perceptron neural network, characterized in that: include: The boundary condition calculation module calculates the boundary conditions that the error coefficient of the individual charging pile needs to meet under the preset misjudgment rate and missed judgment rate when determining the out-of-tolerance individual in the charging pile group to be tested using the preset measurement assurance scheme; The correlation calculation module combines the existing charging pile data and uses a pre-trained multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient; The error coefficient calculation module uses the obtained association relationship to analyze the explicit labels of the charging piles to be grouped, and predicts the error coefficients of the charging piles to be grouped, that is, the explicit labels of the charging piles to be grouped are input into the pre-trained multi-layer perceptron neural network model, and the predicted values ​​of the error coefficients of the charging piles are calculated and output; The charging pile grouping module groups the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging piles and the error coefficient of each charging pile to be grouped.

6. The charging pile grouping system based on multi-layer perceptron neural network according to claim 5 is characterized in that: When the preset metering assurance scheme is used to determine the out-of-tolerance individuals in the group of charging piles to be tested, the boundary condition steps to be satisfied by calculating the error coefficient of the individual charging pile under the preset misjudgment rate and missed judgment rate are specifically as follows: The error coefficient x of each charging pile i The mean value x t As the virtual verification standard value of the measurement assurance plan, the conditional probability is calculated: P1≤γ1; P2≤γ2; Where γ1 and γ2 are the maximum values ​​of the misjudgment rate and missed judgment rate when the measurement assurance scheme is adopted, P1 is the probability of an error coefficient exceeding the tolerance of each charging pile, and P2 is the probability of no error coefficient exceeding the tolerance of each charging pile. According to the conditional probability calculation of the above two equations, the boundary conditions of the error coefficient of the individual charging pile are obtained to meet the requirements of the misjudgment rate and missed judgment rate: f(x i )≤0; In the formula, the f() function represents the functional relationship that the error coefficient of each charging pile needs to satisfy.

7. The charging pile grouping system based on multi-layer perceptron neural network according to claim 5 is characterized in that: The step of using a multi-layer perceptron neural network to obtain the correlation between the charging pile explicit label and the error coefficient is specifically as follows: The explicit label of the charging pile is used as an input feature, wherein the explicit label includes manufacturer, production date, model, batch, installation location, indoor / outdoor, AC / DC, monthly usage frequency, and monthly charging power; Perform one-hot encoding on the categorical variables in the display labels and standardize the numerical variables in the display labels to obtain the preprocessed input feature vectors; A pre-trained multi-layer perceptron neural network model is used, with the pre-processed input feature vector as input, to output the predicted value of the charging pile error coefficient.

8. The charging pile grouping system based on multi-layer perceptron neural network according to claim 5 is characterized in that: The step of grouping the charging piles to be grouped according to the boundary conditions of the error coefficient of the individual charging pile and the error coefficient of each charging pile to be grouped is specifically as follows: The prediction error coefficient of each charging pile is taken as input, and the prediction error coefficient of each charging pile is clustered using the Gaussian mixture model; the Gaussian probability density function of the error coefficient of each charging pile is calculated, specifically: In the formula, z k is the hidden variable of the kth Gaussian distribution (k=1,2,…,Q, Q is the number of Gaussian distributions), z k =1 means that charging pile j belongs to the kth Gaussian distribution, zk=0 means it does not; Σ k , μ k are the covariance matrix and mean of the kth Gaussian component respectively; y j Indicates the error coefficient of each charging pile; Calculate the posterior probability that the error coefficient of the jth charging pile belongs to the kth Gaussian distribution. The specific calculation formula is as follows: In the formula, π l is the weight of the lth Gaussian distribution, satisfying Then update the model parameters π k ,Σ k , μ k : Finally, according to the posterior probability P(z k |y j ), the error coefficient of each charging pile is assigned to the group with the maximum a posteriori probability, thus completing the grouping of the charging piles to be grouped.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a charging pile grouping method based on a multi-layer perceptron neural network as described in any one of claims 1 to 4 is implemented.

Images