A digital multimeter electric energy metering function calibration system and method

Abstract

The application discloses a digital multimeter electric energy metering function verification system and method, wherein the verified digital multimeter is connected with a standard device; the verified digital multimeter and the standard device are respectively connected with a synchronous clock module; output signals of the verified digital multimeter and the standard device are transmitted to a data acquisition system; the data acquisition system comprises a signal conditioning module, an ADC acquisition module and a data preprocessing module; the signal conditioning module is used for amplitude adjustment and filter preprocessing of the output signals; the ADC acquisition module is used for realizing synchronous sampling; and the data preprocessing module is used for completing data format conversion and preliminary noise reduction; the data acquisition system is connected with an error calculation module; the error calculation module stores signals transmitted from the data acquisition system and performs error analysis; the system and the method can realize intelligent verification of the multimeter, can realize early warning of error risks, and meet the demand of automatic verification.

Description

Technical Field

[0001] This invention belongs to the field of electrical measurement technology and relates to a system and method for verifying the energy metering function of a digital multimeter. Background Technology

[0002] Currently, existing digital meter power verification technologies mainly rely on traditional verification methods. These methods often employ single time-frequency analysis algorithms (such as conventional S-transform and Fourier transform) to extract harmonic parameters. In complex harmonic and strong electromagnetic interference scenarios, the extraction errors of harmonic amplitude and phase are significant, leading to inaccurate calculation of the standard total power and consequently affecting the reliability of the final error assessment. Existing error calculations mostly use static statistical methods, which can only obtain the final error result and cannot fit the dynamic nonlinear relationship between the measured power of the meter under test and the standard power. This makes it difficult to achieve accurate output of instantaneous errors and early prediction of error trends, and is particularly unsuitable for complex verification conditions such as transient loads and pulse loads. Furthermore, the verification process has a low level of intelligence. After the error calculation is completed, manual analysis is required to determine the cause of the error exceeding the standard. This is not only time-consuming and labor-intensive, prone to human error, but also unable to provide early warning of error risks, making it difficult to meet the industry's development needs for automated verification. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a digital multimeter energy metering function verification system and method. This system and method can realize intelligent verification of multimeters and can realize early warning of error risks, thus meeting the needs of automated verification.

[0004] To achieve the above objectives, the present invention discloses a digital multimeter energy metering function verification system, including a digital multimeter to be verified, a standard device, a data acquisition system, an error calculation module, and a synchronization clock module for synchronizing the clock. The digital multimeter under test is connected to a standard device; both the digital multimeter under test and the standard device are connected to a synchronous clock module; the output signals of the digital multimeter under test and the standard device are transmitted to a data acquisition system; the data acquisition system includes a signal conditioning module, an ADC acquisition module, and a data preprocessing module; the signal conditioning module is used to adjust the amplitude and perform filtering preprocessing on the output signal, the ADC acquisition module is used to achieve synchronous sampling, and the data preprocessing module is used to complete data format conversion and preliminary noise reduction; the data acquisition system is connected to an error calculation module, which stores the signal sent from the data acquisition system and performs error analysis.

[0005] This invention discloses a method for verifying the energy metering function of a digital multimeter, comprising the following steps: 11) Set up a verification scheme, select the verification point of the digital multimeter to be verified according to the requirements of the verification procedure of the electricity meter, and generate a verification signal containing fundamental wave, harmonic wave and dynamic load characteristics through the standard signal source of the standard device according to the preset verification conditions. The verification quantity includes voltage vector U(ω,t) and current vector signal I(ω,t). 12) Synchronously load the calibration value of the standard signal source of the standard device into the data acquisition system; 13) The data acquisition system synchronously acquires the voltage and current signals after loading through the ADC acquisition module to obtain the original signals; 14) The acquired raw signal is processed by the signal conditioning module using the Kalman filter algorithm for noise reduction to suppress signal distortion caused by power grid electromagnetic interference; 15) Extract the instantaneous amplitude and phase information from the original signal through the data preprocessing module, and construct a three-dimensional data matrix of time-amplitude-phase.

[0006] This invention discloses a method for verifying the energy metering function of a digital multimeter, comprising the following steps: 21) The basic error of the electrical energy of the digital multimeter to be verified shall be expressed in the form of relative error; 22) A standard device with an accuracy level greater than or equal to 0.05 is used. The standard source output of the standard device is the current and voltage signals required for verification. An improved S-transform + CNN1D + attention mechanism model is used to perform time-frequency joint analysis on the voltage and current signals collected by the data, and the vector components of the fundamental wave and each harmonic are decomposed to achieve noise reduction and calibration of the harmonic components.

[0007] Furthermore, the CNN1D+attention mechanism model takes the time-frequency matrix of the S-transform as input, performs inference on the time-frequency matrix after S-transform decomposition, and analyzes and outputs the precise U of the 50th harmonic. n0 φ nu I n0 φ ni The amplitude and phase parameters are used to achieve noise reduction and calibration of harmonic components, where the voltage vector of the nth harmonic is U. n =U n0 ∠φ nu The current vector is I n =I n0 ∠φ ni U n0 I n0 Let φ be the amplitude of the nth harmonic. nu φ ni These are the phase angles of the nth harmonic.

[0008] Furthermore, the CNN1D+ attention mechanism model includes convolutional layers, attention modules, pooling layers, and fully connected layers connected in sequence. The convolutional layers are set to 64 channels, 128 channels, and 256 channels in sequence. The pooling layers use max pooling. The fully connected layers are set to 512 neurons and 100 neurons. A Dropout layer with a ratio of 0.2 is embedded to suppress overfitting.

[0009] Furthermore, it also includes: calculating the total measured power P of the digital multimeter being calibrated based on the principle of vector power calculation. m and standard total power P s .

[0010] Furthermore, the standard total power P s for: P s =Σ(U n0 ×I n0 ×cos(φ nu -φ ni (n=1,2,...,50) The measured total power P of the digital multimeter being calibrated m Obtained by reading its actual output value.

[0011] Furthermore, it also includes: constructing an LSTM + fully connected layer dynamic error model, inputting time-series harmonic power, time, and load factor into the LSTM + fully connected layer dynamic error model, outputting the instantaneous error ε(t), and calculating the instantaneous error ε(t) and the weighted average error ε. av9 .

[0012] Furthermore, the instantaneous error ε(t) and the weighted average error ε av9 They are respectively: ε(t)=(P (t)-P (t)) / P (t)×100% ε av9 =(1 / T)∫0 |(P(t)-P(t)) / P(t)|×100%dt Where T is the verification duration, P m(t) and P s(t) These represent the measured power and standard power of the digital multimeter being calibrated at time t, respectively.

[0013] Furthermore, the LSTM+fully connected layer dynamic error model uses a 2-layer LSTM network to extract temporal features, with an input dimension of 51. The number of neurons in the LSTM hidden layer is 128, and the fully connected layer is set with 64 neurons and 1 neuron respectively. A Dropout layer with a ratio of 0.3 is embedded to suppress overfitting.

[0014] Furthermore, it also includes: using the XGBoost classification model, based on the output of the LSTM + fully connected layer dynamic error model, to determine the reasons for the error exceeding the standard.

[0015] The present invention has the following beneficial effects: The digital multimeter energy metering function verification system and method of the present invention, in specific operation, integrates deep learning algorithms to optimize verification accuracy, and simultaneously outputs the traditional calculation error and the error corrected by deep learning, so as to realize the intelligent verification of multimeters. It not only meets the compliance requirements of metrological verification, but also improves the verification accuracy, meets the needs of automated verification, and is easy to promote and apply in the field. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a system structure diagram of the present invention; Figure 2 This is a flowchart of data acquisition in this invention; Figure 3 This is a flowchart of the error calculation process in this invention.

[0018] Among them, 1 is the digital multimeter being verified, 2 is the standard device, 3 is the data acquisition system, 4 is the error calculation module, 5 is the synchronization clock module, and 6 is the cloud management platform. Detailed Implementation

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

[0020] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0023] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0024] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0026] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0027] Example 1 refer to Figure 1 The digital multimeter energy metering function verification system of the present invention includes a digital multimeter to be verified 1, a standard device 2, a data acquisition system 3, an error calculation module 4, a synchronization clock module 5, and a cloud management platform 6; The digital multimeter under test 1 is connected to the standard device 2; both the digital multimeter under test 1 and the standard device 2 are connected to the synchronous clock module 5; the parameter data of the digital multimeter under test 1 and the standard device 2 are transmitted to the data acquisition system 3; the data acquisition system 3 includes a signal conditioning module, an ADC acquisition module, and a data preprocessing module; the signal conditioning module is used to adjust the amplitude and perform filtering preprocessing on the signal, the ADC acquisition module realizes high-speed synchronous sampling, and the data preprocessing module completes data format conversion and preliminary noise reduction; the data acquisition system 3 acquires the information measured by the digital multimeter under test 1 and the standard device 2 and sends it to the error calculation module 4; the error calculation module 4 stores the data sent by the data acquisition system 3 and performs error analysis, the standard device 2 sets the verification scheme, controls the start and stop of verification, and synchronously displays the error data; Example 2 refer to Figure 2 In this embodiment, the synchronization clock module 5 provides a unified clock reference for the digital multimeter 1 being verified and the standard device 2. The specific process is as follows: 1) The clock reference source of the synchronization clock module 5 provides a unified clock reference for the digital multimeter 1 to be calibrated and the standard device 2, achieving a clock accuracy of ±0.5ppm.

[0028] 2) The time synchronization between the digital multimeter 1 being verified and the standard device 2 is completed by using a two-way timestamp exchange method to ensure the consistency of signal acquisition phase.

[0029] 3) Initialize vector analysis parameters, set the fundamental frequency, harmonic analysis range (2nd-50th harmonics) and dynamic verification operating condition parameters (including steady state, pulse load, harmonic interference and other complex operating conditions).

[0030] Example 3 The process of generating the verification signal and acquiring data is as follows: 1) Set up a verification scheme, select the verification point of the digital multimeter 1 to be verified according to the requirements of the verification procedure of the electricity meter, and generate a verification signal containing fundamental wave, harmonic wave and dynamic load characteristics through the standard signal source of the standard device 2 according to the preset verification conditions. The verification quantity includes voltage vector U(ω,t) and current vector signal I(ω,t). 2) Synchronously load the calibration value of the standard signal source of standard device 2 into data acquisition system 3; 3) Data acquisition system 3 synchronously acquires the voltage and current signals after loading through the ADC acquisition module to obtain the original signal. The sampling frequency is set to above 10kHz to meet the requirements of harmonic analysis. 4) The acquired raw signal is processed by the signal conditioning module using the Kalman filter algorithm for noise reduction to suppress signal distortion caused by power grid electromagnetic interference; 5) Extract the instantaneous amplitude and phase information from the original signal through the data preprocessing module, and construct a three-dimensional data matrix of time-amplitude-phase.

[0031] Example 4 refer to Figure 3 The error calculation process is as follows: 1) The basic error of the electrical energy of the digital multimeter 1 being verified is expressed in the form of relative error, which is used to determine the basic error of the electrical energy of the digital multimeter 1 being verified with an accuracy class of 0.2. 2) A standard device 2 with an accuracy level greater than or equal to 0.05 is used. The standard source output of the standard device 2 provides the required current and voltage signals for verification. An improved S-transform + CNN1D + attention mechanism model is used to perform time-frequency joint analysis on the acquired voltage and current signals, decomposing them to obtain the vector components of the fundamental frequency and each harmonic. Using the time-frequency matrix of the S-transform as input, inference is performed on the time-frequency matrix after S-transform decomposition, and the accurate U of the 50th harmonic is analytically output. n0 φ nu I n0 φ ni The amplitude and phase parameters are used to achieve noise reduction and calibration of harmonic components, where the voltage vector of the nth harmonic is U. n =U n0 ∠φ nu The current vector is I n =I n0 ∠φ ni U n0 I n0 Let φ be the amplitude of the nth harmonic. nu φ ni These are the phase angles of the nth harmonic; The CNN1D+ attention mechanism model consists of sequentially connected convolutional layers, attention modules, pooling layers, and fully connected layers. The attention module is constructed using adaptive average pooling and fully connected layers to enhance the feature weights of key harmonic components. The model structure parameters are as follows: the number of input channels is 2 (corresponding to voltage and current signals respectively), the output dimension is 100 (corresponding to 50 parameters each for amplitude and phase of the 50th harmonic), the convolutional layers are set to 64 channels (5×1 kernel), 128 channels (3×1 kernel), and 256 channels (3×1 kernel), the pooling layer uses max pooling (stride 2), and the fully connected layers are set to 512 neurons and 100 neurons. A Dropout layer with a ratio of 0.2 is embedded to suppress overfitting. 3) Based on the principle of vector power calculation, calculate the total measured power P of the digital multifunction table 1 being verified. m Compared with standard total power P s Among them, the standard total power P s This is the vector sum of the active power of each harmonic, i.e.: P s =Σ(U n0 ×I n0 ×cos(φ nu -φ ni (n=1,2,...,50) The measured total power P of the verified digital multimeter is shown in Table 1. m Obtained by reading its actual output value; 4) Construct an LSTM + fully connected layer dynamic error model, inputting time-series harmonic power, time, load factor, and other features, outputting instantaneous error ε(t), and optimizing the average error calculation, calculating the instantaneous error ε(t) and the weighted average error ε. av9 To address the timing characteristics of electrical energy verification, LSTM is introduced to extract dynamic features. Simultaneously, the average error calculation is optimized through dynamic weighting of error fluctuation variance. ε(t)=(P (t)-P (t)) / P (t)×100% ε av9 =(1 / T)∫0 |(P(t)-P(t)) / P(t)|×100%dt Where T is the verification duration, P m(t) and P s(t) These are the measured power and standard power of the digital multimeter 1 at time t, respectively; The LSTM+fully connected layer dynamic error model uses a 2-layer LSTM network to extract temporal features. The input dimension is 51 (corresponding to the active power and time parameters of the 50th harmonic). The number of neurons in the LSTM hidden layer is 128, and the fully connected layer is set with 64 neurons and 1 neuron respectively. A Dropout layer with a ratio of 0.3 is embedded to suppress overfitting.

[0032] The average error optimization calculation process is as follows: Based on the instantaneous error sequence ε(t) output by the LSTM + fully connected layer dynamic error model, the absolute value of the deviation between the error at each time step and the average error is calculated, and the weight coefficients are constructed accordingly. The larger the deviation, the smaller the weight. The traditional fixed integral is replaced by weighted summation to reduce the impact of error mutation points caused by acquisition noise on the average error calculation. The training data of the model needs to cover multiple operating scenarios with 5%~120% load, 0.5~1.0 power factor, and 0~30% THD. The time-series harmonic power, time, and error label are recorded synchronously for each operating scenario. 5) After calculating the error, use the XGBoost classification model to determine the cause of the error exceeding the standard, such as fundamental frequency measurement error, third harmonic interference, and noise in the acquisition system, and issue an early warning. When the basic error of the acquisition point does not meet the requirements, the basic error of the digital multimeter's energy is determined to be unqualified, and an alarm signal is issued; when the basic error of the acquisition point meets the requirements, a calibration report is issued. The input features of the XGBoost classification model include: the power proportion of each harmonic, the difference in harmonic parameters before and after correction by the CNN1D+ attention mechanism model, the fluctuation variance of instantaneous error, the total harmonic distortion (THD) of voltage / current, and the power factor. The output is the category of reasons for error exceeding the standard and the acceptance risk level.

[0033] The cloud management platform 6 is used for centralized storage, management and analysis of verification data, supports querying, exporting and printing of verification reports; it has error trend statistics and equipment status assessment functions, and can track the energy metering performance of the verified digital multimeter 1 throughout its entire life cycle.

[0034] This invention has the following characteristics: This invention utilizes a CNN1D + attention mechanism model, taking an improved S-transform time-frequency matrix as input, to accurately denoise and calibrate harmonic components, and can efficiently output the amplitude U of each harmonic. n0 Phase φ nu / φ ni The parameters effectively address the issue of large harmonic parameter extraction errors in existing improved S-transforms under complex harmonic and strong noise scenarios, which in turn leads to a decrease in the standard total power P. s Addressing the technical pain point of inaccurate calculations, this study significantly improves the accuracy of harmonic parameter extraction, ensuring the standard total power P. sThe calculations are accurate, providing high-precision data support for subsequent error calculations.

[0035] This invention utilizes an LSTM + fully connected layer to construct a dynamic error model, incorporating time-series features such as harmonic power, time, and load factor, enabling real-time output of the instantaneous error ε(t). Furthermore, it optimizes the average error ε through weighted integration. av9 The calculation effectively overcomes the limitations of traditional error calculation methods, which can only statistically analyze error results and cannot fit the measured power P of the tested table. m (t) and standard power P s The shortcomings of dynamic nonlinear relationship between (t) and the inability to predict error trend in advance are particularly applicable to power energy verification under transient load and complex working conditions, so as to realize accurate dynamic error modeling and trend prediction and optimize the reliability of error assessment. This invention utilizes the XGBoost classification model to accurately identify the causes of errors exceeding the standard. It can quickly locate specific sources of anomalies such as fundamental frequency measurement error, third harmonic interference, and noise in the acquisition system, and simultaneously output a pass / fail risk warning. Compared with the existing manual judgment mode, it not only reduces the workload and error of manual analysis, but also provides early warning of potential verification risks (such as errors approaching the threshold under low load). It realizes integrated intelligent verification of "error calculation - cause tracing - risk warning", which greatly improves verification efficiency and intelligence level.

[0036] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0037] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0038] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A digital multimeter energy metering function verification system, characterized in that, It includes a digital multimeter to be verified (1), a standard device (2), a data acquisition system (3), an error calculation module (4), and a synchronization clock module (5) for synchronizing the clock. The digital multimeter under test (1) is connected to the standard device (2); the digital multimeter under test (1) and the standard device (2) are respectively connected to the synchronous clock module (5); the output signals of the digital multimeter under test (1) and the standard device (2) are transmitted to the data acquisition system (3); the data acquisition system (3) includes a signal conditioning module, an ADC acquisition module and a data preprocessing module; the signal conditioning module is used to adjust the amplitude and filter the output signal, the ADC acquisition module is used to realize synchronous sampling, and the data preprocessing module is used to complete the data format conversion and preliminary noise reduction; the data acquisition system (3) is connected to the error calculation module (4), and the error calculation module (4) stores the signal sent by the data acquisition system (3) and performs error analysis.

2. A method for verifying the energy metering function of a digital multimeter, characterized in that, Includes the following steps: 11) Set up a verification scheme, select the verification point of the digital multimeter (1) to be verified according to the requirements of the verification procedure of the electricity meter, and generate a verification signal containing fundamental wave, harmonic and dynamic load characteristics through the standard signal source of the standard device (2) according to the preset verification conditions. The verification quantity includes voltage vector U(ω,t) and current vector signal I(ω,t). 12) The calibration value of the standard signal source of the standard device (2) is synchronously loaded into the data acquisition system (3); 13) Data acquisition system (3) synchronously acquires the voltage and current signals after loading through the ADC acquisition module to obtain the original signal; 14) The acquired raw signal is processed by the signal conditioning module using the Kalman filter algorithm for noise reduction to suppress signal distortion caused by power grid electromagnetic interference; 15) Extract the instantaneous amplitude and phase information from the original signal through the data preprocessing module, and construct a three-dimensional data matrix of time-amplitude-phase.

3. A method for verifying the energy metering function of a digital multimeter, characterized in that, Includes the following steps: 21) The basic error of the electrical energy of the digital multimeter (1) to be verified is expressed in the form of relative error; 22) A standard device (2) with an accuracy level greater than or equal to 0.05 is used. The standard source output of the standard device (2) is used to verify the required current and voltage signals. An improved S-transform + CNN1D + attention mechanism model is used to perform time-frequency joint analysis on the voltage and current signals collected by the data, and the vector components of the fundamental wave and each harmonic are decomposed to achieve noise reduction and calibration of the harmonic components.

4. The method for verifying the energy metering function of a digital multimeter according to claim 3, characterized in that, The CNN1D+attention mechanism model takes the time-frequency matrix of the S-transform as input, performs inference on the time-frequency matrix after S-transform decomposition, and analyzes and outputs the precise U of the 50th harmonic. n0 φ nu I n0 φ ni The amplitude and phase parameters are used to achieve noise reduction and calibration of harmonic components, where the voltage vector of the nth harmonic is U. n =U n0 ∠φ nu The current vector is I n =I n0 ∠φ ni U n0 I n0 Let φ be the amplitude of the nth harmonic. nu φ ni These are the phase angles of the nth harmonic.

5. The method for verifying the energy metering function of a digital multimeter according to claim 3, characterized in that, The CNN1D+ attention mechanism model includes convolutional layers, attention modules, pooling layers, and fully connected layers connected in sequence. The convolutional layers are set to 64 channels, 128 channels, and 256 channels in sequence. The pooling layers use max pooling. The fully connected layers are set to 512 neurons and 100 neurons. A Dropout layer with a ratio of 0.2 is embedded to suppress overfitting.

6. The method for verifying the energy metering function of a digital multimeter according to claim 3, characterized in that, Also includes: Based on the principle of vector power calculation, the total measured power P of the digital multimeter (1) under verification is calculated respectively. m and standard total power P s .

7. The method for verifying the energy metering function of a digital multimeter according to claim 6, characterized in that, The standard total power P s for: P s =Σ(U n0 ×I n0 ×cos(φ nu -f ni ))(n=1,2,...,50) The measured total power P of the digital multimeter under test (1) m Obtained by reading its actual output value.

8. The method for verifying the energy metering function of a digital multimeter according to claim 7, characterized in that, Also includes: A dynamic error model with an LSTM and a fully connected layer is constructed. Time-series harmonic power, time, and load factor are input into the LSTM and fully connected layer dynamic error model, and the instantaneous error ε(t) is output. The instantaneous error ε(t) and the weighted average error ε are then calculated. av9 ; The LSTM+fully connected layer dynamic error model uses a 2-layer LSTM network to extract temporal features. The input dimension is 51, the number of neurons in the LSTM hidden layer is 128, and the fully connected layer is set with 64 neurons and 1 neuron respectively. A Dropout layer with a ratio of 0.3 is embedded to suppress overfitting.

9. The method for verifying the energy metering function of a digital multimeter according to claim 8, characterized in that, Instantaneous error ε(t) and weighted average error ε av9 They are respectively: ε(t)=(P (t)-P (t)) / P (t)×100% ε av9 =(1 / T)∫0 |(P (t)-P (t)) / P (t)|×100%dt Where T is the verification duration, P m(t) and P s(t) The measured power and standard power of the digital multimeter (1) at time t are respectively the measured power and standard power.

10. The method for verifying the energy metering function of a digital multimeter according to claim 8, characterized in that, Also includes: Using the XGBoost classification model, and based on the output of the LSTM + fully connected layer dynamic error model, we determine the reasons for the error exceeding the standard.

Images