Electric vehicle charging pile metrological verification method based on improved infectious disease model

Abstract

The invention relates to an electric vehicle charging pile metrological verification method based on an improved infectious disease model. The electric vehicle charging pile metrological verification method comprises the following steps: step 1, constructing a charging pile remote metrological verification model based on an improved susceptibility-infection-susceptibility propagation model; 2, solving the constructed charging pile remote metrological verification model based on the improved SIS propagation model; and step 3, selecting key features influencing the charging electric quantity, fitting the electric vehicle side charging feature data and a charging pile accumulated charging electric energy curve by adopting a cloud reference charging accumulated electric energy calculation method based on a residual convolutional neural network, and obtaining a metering error of the charging pile through an interval charging electric quantity change value. And the metering state of the charging pile is judged based on a binary method to realize charging pile metering situation awareness. And 4, recalibrating the state of the charging pile in the detected area, and presenting the result of each round of detection until the detection is finished. According to the invention, continuous dynamic monitoring and verification of the metering performance of the charging pile can be realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric vehicle charging pile (charging pile) measurement and verification, and relates to an electric vehicle charging pile measurement and verification method, in particular to an electric vehicle charging pile measurement and verification method based on an improved infectious disease model. BACKGROUND

[0002] With the popularity of electric vehicles (EVs), non-vehicle-mounted chargers (charging piles) as important equipment to support electric vehicle charging across regions, the accuracy of their measurement directly relates to the fairness of user charging costs and the management efficiency of power operation enterprises. However, traditional measurement and verification methods mainly rely on offline sampling or online periodic verification, which have the problems of poor flexibility and insufficient real-time performance, and are difficult to meet the increasingly complex needs of charging scenarios. At the same time, as the measurement performance of charging piles gradually degrades in long-term use, the traditional method cannot continuously and dynamically reflect the measurement state of the equipment, which easily leads to the accumulation of measurement errors and negatively affects the quality of electric vehicle charging services. Therefore, there is an urgent need for a remote verification technology that can accurately monitor the measurement performance of charging piles in real time to meet the current industry challenges.

[0003] Currently, methods for remote online verification through massive measurement data have been widely studied. However, early generalized energy conservation methods are mainly used for measurement and verification of smart electric energy meters under the transformer area, and direct transplantation to non-vehicle-mounted charger measurement and verification is not completely applicable. This is because the indication of the non-vehicle-mounted charger is the direct current power after the meter, while the generalized energy conservation method needs to be converted to the alternating current power before the meter to build the conservation equation for calculation, which introduces an estimated error of the conversion efficiency of the first conversion, including the conversion efficiency of the AC / DC power module, line loss power, and power of station equipment (such as power loss of charging pile standby). The estimated error of these influencing parameters is often larger than the measurement error, so it is difficult to guarantee the accuracy of the measurement error solution and the excess error discrimination.

[0004] Under the background of electric vehicle cross-domain mobility characteristics, the complexity of vehicle-pile interaction scenarios brings new opportunities and challenges to remote measurement and verification. On the one hand, the dynamic use characteristics of electric vehicles provide rich operation data for state analysis of measurement equipment; on the other hand, traditional measurement models are difficult to effectively handle such dynamic and multi-variable data characteristics.

[0005] To solve the above problems, the application provides an electric vehicle charging pile measurement and verification method based on an improved infectious disease model.

[0006] After searching, no existing technical disclosure literature identical or similar to the present application has been found. SUMMARY

[0007] To address the shortcomings of existing technologies, this invention proposes a metrological verification method for electric vehicle charging piles based on an improved infectious disease model. This method overcomes the limitations of traditional offline sampling verification and online periodic verification. By constructing a metrological verification propagation model based on vehicle-charging pile interaction scenarios, it enables continuous dynamic monitoring and verification of the metrological performance of charging piles.

[0008] The above-mentioned objective of this invention is achieved through the following technical solution: A metrological verification method for electric vehicle charging piles based on an improved infectious disease model includes the following steps: Step 1: Construct a remote metering verification model for charging piles based on an improved susceptible-infection-susceptible propagation model, based on the vehicle-charging pile interaction scenario; Step 2: Solve the remote metering verification model of charging piles based on the improved SIS propagation model constructed in Step 1 based on the mean field theory, and obtain the continuous and dynamic analysis results of the dynamic behavior of metering verification propagation. Step 3: Based on the continuous and dynamic analysis results of the dynamic behavior of the measurement verification propagation obtained in Step 2, select the key features that affect the charging power, and use the cloud-based benchmark charging cumulative energy calculation method based on residual convolutional neural network to fit the electric vehicle side charging characteristic data and the charging pile cumulative charging energy curve. Obtain the charging pile measurement error through the interval charging power change value, and judge the charging pile measurement status based on the binary method to realize the charging pile measurement situation awareness.

[0009] Step 4: Based on the charging pile metering situation awareness results obtained in Step 3, calculate the degradation intensity of charging piles in non-out-of-tolerance state, determine the specific charging piles that have degraded, recalibrate the status of charging piles in the inspection area, and present the inspection results for each round until the inspection is completed.

[0010] Furthermore, the specific steps of step 1 include: (1) Determine the verification status of the charging pile: In the remote metering verification model of the charging pile based on the improved SIS propagation model, the verification status of the charging pile is defined as: verification status CE, metering non-out-of-tolerance status CI, metering out-of-tolerance status NCI and verification status CS. When a charging pile in the pending verification state (CS) charges a calibration vehicle, a metrological verification action occurs, and the state changes to the verification state (CE). During the charging process, charging time sequence characteristic data and charging interaction data are uploaded to the cloud-based metrological error calculation model to solve for the metrological error value of the charging pile. Based on the measurement error value, it is determined whether the charging pile is out of tolerance, and it jumps to either the non-out-of-tolerance state CI or the out-of-tolerance state NCI. The charging pile that has changed to the non-out-of-tolerance state CI becomes a new source of propagation, continuing to spread its state by charging uncalibrated electric vehicles. As the charging process continues, there is a probability that the measurement accuracy of the charging pile that is already in the non-out-of-tolerance state CI will degrade, so it may be converted back to the pending verification state CS, waiting to be re-verified.

[0011] (2) Based on the verification status of the charging pile determined in step (1), construct a remote metering verification model for the charging pile based on the improved SIS propagation model: model the charging pile as a node in the metering network diagram. v i Using quadruples S ( v i )={ λ ( v i ), state ( v i , t ), τ ( v i ), β ( v i )}represent v i The attributes in the network are used to model the EVs-charging pile charging interaction network, and then the construction of a remote metering verification model for charging piles based on the improved SIS propagation model is completed. in, λ ( v i )represent v i Measurement and verification of propagation rate. state ( v i , t ) indicates a non-vehicle-mounted charger v i In time t The state at any given moment. τ ( v i ) represents the intensity of the degradation from the CI state to the CS state. β ( v i ) is a BOOL type variable. β ( v i The value 1 indicates a transition from the CE state to the CI state. β ( v i)=0 indicates a transition from CE state to NCI state.

[0012] Furthermore, the specific steps of step 2 include: (1) After each round of metrological verification, the number of charging piles in CS, CE, and CI states changes as follows: (1) (2) (3) In the formula, , and These represent the ratios of charging piles in CS, CE, and CI states, respectively; the initial propagation source formation time is taken as the starting time, and each time interval Δt is recorded as one propagation round; λ(t) is the number of propagation rounds. t Wheel measurement verification of average propagation rate, λ( t )=1 / h ∑hi=1λ( v i ,△ t ), v i ∈CI, h Let be the total number of nodes with propagation capability in round t; ( t ) is the first t Wheel measurement non-out-of-tolerance state transition rate, ( t ) is the first t Wheel measurement non-out-of-tolerance state transition rate, ( t )=1 / q ∑hi=1 β ( v i ), v i ∈CE, q The total number of nodes in the CE state in round t; r ( t ) is the first t Wheel metrological verification degradation rate.

[0013] (2) Using mean-field theory, the equations of the SIS model can be solved: (4) Furthermore, charging stations in different states have the following relationships: (5) (6) Based on the above equation, the proportion of charging piles in different states at (t+△t) rounds can be obtained, thus obtaining continuous and dynamic analysis results of the dynamic behavior of measurement verification propagation.

[0014] Furthermore, the specific steps of step 3 include: (1) Based on the continuous and dynamic analysis results of the dynamic behavior of metrological verification propagation obtained in step 2, select the key features that affect the charging capacity; (2) Based on the key features that affect the charging power selected in step (1), a cloud-based benchmark charging cumulative energy calculation method based on the residual convolutional neural network Attention-GRU model is constructed. The electric vehicle side charging feature data and the charging pile cumulative charging energy curve are fitted, and the metering error of the charging pile is obtained through the interval charging power change value. (3) Based on the obtained measurement error estimate, the metering status of the charging pile is judged by the binary method to realize the metering situation perception of the charging pile.

[0015] Furthermore, the specific steps of step 3 (1) include: ① The correlation between time-series characteristic data and standard charging cumulative power data under the SOC change during EV charging was calculated by Pearson correlation coefficient, and the key factors affecting charging power were analyzed. (7) In the formula, x i For the charging timing feature data of the i-th EV, y Accumulate charging point energy data for standard charging piles. and These are the average values ​​of EV charging timing characteristic data and the cumulative charging point energy data of charging piles, respectively.

[0016] ② Select the features that are strongly correlated with the accumulated charge under standard charging, namely SOC, BMS voltage, battery pack maximum temperature, and battery pack minimum temperature, as inputs for the subsequent model.

[0017] Moreover, the specific steps of step 3 (2) include: ① Based on the key features affecting the charging capacity selected in step (1), the information weight values ​​at each time point are first calculated, then the weight values ​​are normalized using the softmax function, and finally the normalized attention weight values ​​are assigned to each time point of the input vector. The calculation process is as follows: (8) (9) In the formula, The input feature tensor; This is a dimensional transformation operation; This is a fully connected operation; These are trainable parameters; This is an element-wise multiplication operation.

[0018] After the attention mechanism, a gated recurrent unit (GRU) is used to extract one-dimensional temporal features. The GRU gated unit includes an input gate, an update gate, and a reset gate. in, The input information of the gating unit at the current moment, This represents the hidden state information passed down from the previous moment. This represents the output information of the gating unit at the current moment. This represents updating the output information of the gate at the current moment. This represents the output information for resetting the door at the current moment.

[0019] The formula for calculating the amount of data that can be retained up to the current time by updating the gate's memory information is as follows: (10) In the formula, This represents a one-dimensional convolution operation; This represents the training parameters of the convolution kernel of the update gate.

[0020] (11) In the formula, This represents the training parameters of the convolution kernel for resetting the gate.

[0021] The output of the gating unit at the current moment is obtained by the following formula: (12) (13) In the formula, This represents the training parameters of the convolution kernel.

[0022] ② The benchmark cumulative charging point energy is reconstructed through this network. When EVs verify charging piles, the actual cumulative charging energy can be predicted based on the trained model. The actual cumulative charging energy is compared with the charging energy value indicated on the pile side, realizing the metering status perception of the CE status node in the metering verification network, and thus completing the fitting of the electric vehicle side charging characteristic data and the charging pile cumulative charging energy curve.

[0023] ③ Calculate the change in accumulated charging energy by analyzing the output sequence of actual accumulated charging energy of the EV. The corresponding indicated energy change value of the tested charging pile Calculate the estimated relative measurement error of the charging pile under inspection; (14) Moreover, the specific method of step (3) of step 3 is as follows: After obtaining the estimated measurement error, a measurement relative error threshold is set. Determine C E Whether the metering status has changed is determined and calibrated using a binary method, thus realizing the metering status awareness of the charging pile; (15) In the formula, The standard measurement relative error for secondary charging piles, .

[0024] Furthermore, the specific steps of step 4 include: (1) Based on the charging pile metering situation awareness results obtained in step 3, the degradation intensity of charging piles in non-out-of-tolerance state is calculated by using the dynamic calculation method of metering verification degradation rate. (twenty four) In the formula, For a fixed degradation rate.

[0025] (2) Each CI calculates its own metering performance degradation intensity according to formula (10), and determines the specific degraded charging pile individual based on the ranking of metering performance degradation intensity and the number of degradations.

[0026] (25) In the formula, The cumulative number of times a charging station has been marked as CI status until the current propagation test identifies it as charging an electric vehicle. The threshold number of charging cycles; For node v i Metrological verification of degradation intensity; The initial metrological performance degradation intensity; This is an estimate of the measurement error; This represents the mean of the estimated measurement error values; This represents the variance of the estimated measurement error.

[0027] (3) After each round of measurement status solution, the status of the charging piles in the inspection area is recalibrated, and the inspection results of each round are presented until the inspection is completed.

[0028] The advantages and beneficial effects of this invention are as follows: 1. This invention proposes a metrological verification method for electric vehicle charging piles based on an improved infectious disease model. By constructing a metrological verification propagation model based on vehicle-pile interaction scenarios, the metrological status of charging piles is abstracted as a "susceptible-infected-susceptible (SIS)" propagation process (the metrological status evolves / propagates in the network with vehicle-pile interaction). This effectively overcomes the drawbacks of traditional charging pile metrological verification methods based on generalized energy conservation theory, such as the unavoidable line loss and AC-DC conversion loss causing out-of-tolerance identification errors, and the inability to verify isolated charging piles or those coexisting with other electrical equipment under the same meter in the same distribution area. Furthermore, it possesses the rapid verification capability of the SIS infectious disease index spread.

[0029] 2. This invention proposes a metering verification method for electric vehicle charging piles based on an improved infectious disease model. The method uses mean field theory to solve the propagation model in step 1, enabling real-time perception of the dynamic evolution of the charging pile metering status over time, and achieving remote, sustainable dynamic monitoring and flexible analysis.

[0030] 3. This invention proposes a metering verification method for electric vehicle charging piles based on an improved infectious disease model, and proposes a cloud-based benchmark charging cumulative energy calculation method using Attention-GRU. This method achieves accurate fitting of the nonlinear relationship between "electric vehicle side charging characteristic data - charging pile cumulative charging energy curve", thereby solving for the estimated value of charging pile metering error and improving the accuracy of charging pile metering error identification.

[0031] 4. This invention proposes a metrological verification method for electric vehicle charging piles based on an improved infectious disease model. It constructs an energy decay model to quantify the intensity of metrological performance degradation and implements a dynamic exit mechanism for "metrological non-out-of-tolerance state" based on dynamic judgment logic, which more comprehensively and objectively reflects the charging pile's operating status and the degradation attributes of metrological accuracy. Attached Figure Description

[0032] Figure 1 This is a framework diagram of the electric vehicle charging pile metrological verification method based on an improved infectious disease model according to the present invention; Figure 2 This is a diagram illustrating the baseline charging cumulative energy calculation framework based on Attention-GRU of the present invention. Figure 3 This is a graph showing the dynamic change of measurement error as the number of verification rounds varies according to the present invention. Figure 4 The diagram shows the calculation results of the SIS-charging pile metering verification model of the present invention. Detailed Implementation

[0033] The structure of the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that these embodiments are descriptive and not limiting.

[0034] A metrological verification method for electric vehicle charging piles based on an improved infectious disease model includes the following steps: Step 1: Based on the vehicle-charging pile interaction scenario, a remote metering verification model for charging piles based on the improved Susceptible-Infective-Susceptible (SIS) propagation model was constructed. The specific steps of step 1 include: (1) Determining the calibration status of charging piles: In the remote metrological calibration model of charging piles based on the improved SIS propagation model, the calibration status of charging piles is defined as the calibration state (CE), the non-out-of-tolerance state (CI), the out-of-tolerance state (NCI), and the state to be calibrated (CS). When a charging pile in the CS state charges a calibration vehicle, a metrological calibration action occurs, and the state changes to the CE state. During the charging process, the charging time sequence characteristic data and charging interaction data are uploaded to the cloud metrological error calculation model to solve for the metrological error value of the charging pile. Based on the metrological error value, it is determined whether it is out of tolerance, and jumps to the non-out-of-tolerance state (CI) or the out-of-tolerance state (NCI). The charging pile that has changed to the CI state becomes a new propagation source, and continues to propagate its state by charging uncalibrated electric vehicles. As the charging process continues, there is a probability that the metrological accuracy of the charging pile that is already in the CI state will degrade, so it may change back to the CS state and wait to be calibrated again.

[0035] (2) Based on the verification status of the charging pile determined in step (1), construct a remote metering verification model for the charging pile based on the improved SIS propagation model: model the charging pile as a node in the metering network diagram. v i Using quadruples S ( v i )={ λ ( v i ), state ( v i , t ), τ ( v i ), β ( v i )}represent v i The attributes in the network are used to model the EVs-charging pile charging interaction network, and then the construction of a remote metering verification model for charging piles based on the improved SIS propagation model is completed. in, λ ( v i )representv i Measurement and verification of propagation rate. state ( v i , t ) indicates a non-vehicle-mounted charger v i In time t The state at any given moment. τ ( v i ) represents the intensity of the degradation from the CI state to the CS state. β ( v i ) is a BOOL type variable. β ( v i The value 1 indicates a transition from the CE state to the CI state. β ( v i )=0 indicates a transition from CE state to NCI state.

[0036] The working principle of step 1 is as follows: (1) SIS-Remote Metering Verification Model for Charging Piles In practical applications, charging piles include those concentrated at various stations and those scattered independently within a jurisdiction – known as "dispersed piles." Regardless of the type, multiple electric vehicles (EVs) are charged during use. Simultaneously, the same EV may charge multiple charging piles sequentially within a certain timeframe. Therefore, the many-to-many interaction between EVs and charging piles naturally establishes an EV-charging pile charging interaction network, which can be used for measurement value transfer, achieving non-invasive metrological verification. Based on this, a novel remote metrological verification model for charging piles is constructed using the naturally formed relationship between EVs and charging piles in the many-to-many interaction process to achieve non-invasive metrological verification. Charging piles calibrated by legal metrological verification institutions are mapped as the source of infection in the model, calibrating their measurement error to zero, meaning their charging measurements are all benchmark values, and the EV is the transmission medium. However, unlike traditional SIS biological models, the tested charging piles exist in two states: exceeding and not exceeding the tolerance. A "cloud-edge" collaborative metrological error calculation model is used to determine whether the tested charging pile exceeds the tolerance. Considering that out-of-tolerance charging piles can cause economic losses to trading users and grid operators, their charging services should be suspended until they are calibrated on-site by verification personnel. Only charging piles with non-out-of-tolerance metering have the ability to continue spreading the problem. This invention defines the improved SIS infectious disease model as the SIS-charging pile model. The comparison between the SIS-charging pile verification principle and the SIS model elements is shown in Table 1.

[0037] Table 1 Comparison of SIS-Charging Pile Verification Principles and SIS Model Elements

[0038] In the SIS-charging pile model, the calibration status of a charging pile is defined as: Calibration (CE), Measurement Non-Tolerance (CI), Measurement Out-of-Tolerance (NCI), and Calibration-Awaiting (CS). When a charging pile in the CS state charges a calibration vehicle, a metrological calibration occurs, and the status transitions to CE. During charging, charging time-series characteristic data and charging interaction data are uploaded to a cloud-based metrological error calculation model to determine the metrological error value of the charging pile. Based on the metrological error value, it is determined whether the value is out of tolerance, and the charging pile transitions to either the Measurement Non-Tolerance (CI) or the Measurement Out-of-Tolerance (NCI) state. Charging piles that transition to the CI state become new propagation sources, continuing to propagate their status by charging uncalibrated electric vehicles. As the charging process continues, charging piles already in the CI state may experience a degradation in metrological accuracy, and therefore may revert to the CS state, awaiting further metrological calibration.

[0039] Modeling the EVs-charging pile charging interaction network: Modeling charging piles as nodes in the metering network graph. v i Using quadruples S ( v i )={ λ ( v i ), state ( v i , t ), τ ( v i ), β ( v i )}represent v i Attributes in the network. λ ( v i )represent v i Measurement and verification of propagation rate. state ( v i , t ) indicates a non-vehicle-mounted charger v i In time t The state at any given moment. τ ( v i ) represents the intensity of the degradation from the CI state to the CS state. β (v i ) is a BOOL type variable. β ( v i The value 1 indicates a transition from the CE state to the CI state. β ( v i )=0 indicates a transition from CE state to NCI state.

[0040] Step 2: Solve the remote metering verification model of charging piles based on the improved SIS propagation model constructed in Step 1 based on the mean field theory, and obtain the continuous and dynamic analysis results of the dynamic behavior of metering verification propagation. The specific steps of step 2 include: (1) After each round of metrological verification, the number of charging piles in CS, CE, and CI states changes as follows: (1) (2) (3) In the formula, , and These represent the ratios of charging piles in CS, CE, and CI states, respectively; the initial propagation source formation time is taken as the starting time, and each time interval Δt is recorded as one propagation round; λ(t) is the number of propagation rounds. t Wheel measurement verification of average propagation rate, λ( t )=1 / h ∑hi=1λ( v i ,△ t ), v i ∈CI, h Let be the total number of nodes with propagation capability in round t; ( t ) is the first t Wheel measurement non-out-of-tolerance state transition rate, ( t ) is the first t Wheel measurement non-out-of-tolerance state transition rate, ( t )=1 / q ∑hi=1 β ( v i ), v i ∈CE, q The total number of nodes in the CE state in round t; r ( t ) is the first tWheel metrological verification degradation rate.

[0041] (2) Using mean-field theory, the equations of the SIS model can be solved: (4) Furthermore, charging stations in different states have the following relationships: (5) (6) Based on the above equation, the proportion of charging piles in different states at (t+△t) rounds can be obtained, thus obtaining continuous and dynamic analysis results of the dynamic behavior of measurement verification propagation.

[0042] Step 3: Based on the continuous and dynamic analysis results of the dynamic behavior of the measurement verification propagation obtained in Step 2, select the key features that affect the charging power, and use the cloud-based benchmark charging cumulative energy calculation method based on residual convolutional neural network to fit the electric vehicle side charging characteristic data and the charging pile cumulative charging energy curve. Obtain the charging pile measurement error through the interval charging power change value, and judge the charging pile measurement status based on the binary method to realize the charging pile measurement situation awareness.

[0043] The specific steps of step 3 include: (1) Based on the continuous and dynamic analysis results of the dynamic behavior of metrological verification propagation obtained in step 2, select the key features that affect the charging capacity; The charging timing characteristic data includes up to 10 components: SOC, minimum battery pack temperature, maximum battery pack temperature, maximum cell voltage, minimum cell voltage, BMS voltage, BMS current, maximum allowable BMS temperature, maximum allowable BMS voltage, and maximum allowable BMS current.

[0044] The specific steps of step 3, step (1) include: ① The correlation between time-series characteristic data and standard charging cumulative power data under the SOC change during EV charging was calculated by Pearson correlation coefficient, and the key factors affecting charging power were analyzed. (7) In the formula, x i For the charging timing feature data of the i-th EV, y Accumulate charging point energy data for standard charging piles. and These are the average values ​​of EV charging timing characteristic data and the cumulative charging point energy data of charging piles, respectively.

[0045] ② Select the features that are strongly correlated with the accumulated charge under standard charging, namely SOC, BMS voltage, battery pack maximum temperature, and battery pack minimum temperature, as inputs for the subsequent model.

[0046] (2) Based on the key features that affect the charging power selected in step (1), a cloud-based benchmark charging cumulative energy calculation method based on the residual convolutional neural network Attention-GRU model is constructed. The electric vehicle side charging feature data and the charging pile cumulative charging energy curve are fitted, and the metering error of the charging pile is obtained through the interval charging power change value. The specific steps of step 3, step (2) include: ① Based on the key features affecting the charging capacity selected in step (1), the information weight values ​​at each time point are first calculated, then the weight values ​​are normalized using the softmax function, and finally the normalized attention weight values ​​are assigned to each time point of the input vector. The calculation process is as follows: (8) (9) In the formula, The input is the EV time-series feature tensor; This is a dimensional transformation operation; This is a fully connected operation; These are trainable parameters; This is an element-wise multiplication operation.

[0047] After the attention mechanism, a gated recurrent unit (GRU) is used to extract one-dimensional time-series features, fully exploring the charging time-series features of EVs. The GRU gated unit includes an input gate, an update gate, and a reset gate. in, The input information of the gating unit at the current moment, This represents the hidden state information passed down from the previous moment. This represents the output information of the gating unit at the current moment. This represents updating the output information of the gate at the current moment. This represents the output information for resetting the door at the current moment.

[0048] The formula for calculating the amount of data that can be retained up to the current time by updating the gate's memory information is as follows: (10) In the formula, This represents a one-dimensional convolution operation; This represents the training parameters of the convolution kernel of the update gate.

[0049] (11) In the formula, This represents the training parameters of the convolution kernel for resetting the gate.

[0050] The output of the gating unit at the current moment is obtained by the following formula: (12) (13) In the formula, This represents the training parameters of the convolution kernel.

[0051] ② The benchmark cumulative charging point energy is reconstructed through this network. When EVs verify charging piles, the actual cumulative charging energy can be predicted based on the trained model. The actual cumulative charging energy is compared with the charging energy value indicated on the pile side, realizing the metering status perception of the CE status node in the metering verification network, and thus completing the fitting of the electric vehicle side charging characteristic data and the charging pile cumulative charging energy curve.

[0052] ③ Calculate the change in accumulated charging energy by analyzing the output sequence of actual accumulated charging energy of the EV. The corresponding indicated energy change value of the tested charging pile Calculate the estimated relative measurement error of the charging pile under inspection; (14) (3) Based on the obtained measurement error estimate, the metering status of the charging pile is judged by the binary method to realize the metering situation awareness of the charging pile. The specific method for step (3) of step 3 is as follows: After obtaining the estimated measurement error, a measurement relative error threshold is set. Determine C E Whether the metering status has changed is determined and calibrated using a binary method, thus realizing the metering status awareness of the charging pile; (15) In the formula, The standard measurement relative error for secondary charging piles, .

[0053] The working principle of step 3 is as follows: (1) Constructing a charging cumulative energy calculation method based on the Attention-GRU model. First, convolution operation is used to fuse the input four-dimensional EV charging time series feature data. In order to extract important information from the charging time series, an attention mechanism is introduced to highlight the important parts of the fused features.

[0054] First, the information weights at each time step are calculated. Then, the weights are normalized using the softmax function. Finally, the normalized attention weights are assigned to the input vector at each time step. The calculation process is as follows: (16) (17) In the formula, The input is the EV time-series feature tensor; This is a dimensional transformation operation; This is a fully connected operation; These are trainable parameters; This is an element-wise multiplication operation.

[0055] After the attention mechanism, a gated recurrent unit (GRU) is used to extract EV charging time-series feature data. GRU, a type of recurrent neural network (RNN), is a variant of the Long Short-Term Memory (LSTM) network. Compared to LSTM networks, GRU reduces one "gating" gate, thus reducing network parameters and computational complexity, while still addressing the gradient vanishing problem that commonly occurs during RNN training. The GRU gating unit includes an input gate, an update gate, and a reset gate. The input information of the gating unit at the current moment, This represents the hidden state information passed down from the previous moment. This represents the output information of the gating unit at the current moment. This represents updating the output information of the gate at the current moment. This represents the output information for resetting the door at the current moment.

[0056] The formula for calculating the amount of data that can be retained up to the current time by updating the gate's memory information is as follows: (18) In the formula, This represents a one-dimensional convolution operation; This represents the training parameters of the convolution kernel of the update gate.

[0057] (19) In the formula, This represents the training parameters of the convolution kernel for resetting the gate.

[0058] The output of the gating unit at the current moment is obtained by the following formula: (20) (twenty one) In the formula, This represents the training parameters of the convolution kernel.

[0059] The baseline cumulative charging energy is reconstructed through this network. When EVs are calibrating charging piles, the actual cumulative charging energy can be predicted based on the trained model and compared with the charging energy readings at the pile side, thus realizing the metering status perception of the CE status nodes in the metering verification network.

[0060] (2) Measurement state estimation The change in accumulated charging energy is calculated by using the output sequence of actual accumulated charging energy of the EV. The corresponding indicated energy change value of the tested charging pile Calculate the estimated relative measurement error of the charging pile under inspection.

[0061] (twenty two) After obtaining the estimated measurement error, a measurement relative error threshold is set. Determine C E Whether the measurement state has changed is determined and calibrated using the binary method: (twenty three) In the formula, The standard measurement relative error for secondary charging piles, .

[0062] Step 4: Based on the charging pile metering situation awareness results obtained in Step 3, calculate the degradation intensity of charging piles in non-out-of-tolerance state, determine the specific charging piles that have degraded, recalibrate the status of charging piles in the inspection area, and present the inspection results for each round until the inspection is completed.

[0063] The specific steps of step 4 include: (1) Based on the charging pile metering situation awareness results obtained in step 3, the degradation intensity of charging piles in non-out-of-tolerance state is calculated by using the dynamic calculation method of metering verification degradation rate. (twenty four) In the formula, For a fixed degradation rate.

[0064] (2) Each CI calculates its own metering performance degradation intensity according to formula (10), and determines the specific degraded charging pile individual based on the ranking of metering performance degradation intensity and the number of degradations.

[0065] (25) In the formula, The cumulative number of times a charging station has been marked as CI status until the current propagation test identifies it as charging an electric vehicle. The threshold number of charging cycles; For node v i Metrological verification of degradation intensity; The initial metrological performance degradation intensity; This is an estimate of the measurement error; This represents the mean of the estimated measurement error values; This represents the variance of the estimated measurement error.

[0066] (3) After each round of measurement status solution, the status of the charging piles in the inspection area is recalibrated, and the inspection results of each round are presented until the inspection is completed.

[0067] The working principle of step 4 is as follows: The degradation rate of charging pile metering performance is a characterization of the rate at which the metering performance of a charging pile in the CI (Completely Inspected) state deteriorates. This process is considered to be an evolution over time, with changes in metering error due to factors such as aging of the internal metering modules over time. Therefore, a dynamic calculation method for the degradation rate of metering verification is designed, where the degradation rate gradually increases with the number of verification rounds.

[0068] (26) In the formula, For a fixed degradation rate.

[0069] Analysis of the factors affecting the degradation of charging pile metering performance reveals that it is mainly influenced by the cumulative charging frequency of the charging pile. That is, within a time interval Δt, the higher the charging frequency, the higher the probability of degradation of its metering accuracy. This behavior is similar to the free space energy degradation model. In addition, considering that charging piles closer to the out-of-tolerance boundary have a higher probability of degrading into out-of-tolerance charging piles, a method for calculating the degradation intensity of charging pile metering performance is proposed. An initial metering performance degradation intensity factor is added to the algorithm, assuming that the initial metering performance degradation intensity follows a Gaussian distribution based on the estimated metering error. Each CI calculates its own metering performance degradation intensity according to equation (27), and the specific degraded charging pile is determined based on the ranking of metering performance degradation intensity and the number of degradations.

[0070] (27) In the formula, The cumulative number of times a charging station has been marked as CI status until the current propagation test identifies it as charging an electric vehicle. The threshold number of charging cycles; For node v i Metrological verification of degradation intensity; The initial metrological performance degradation intensity; This is an estimate of the measurement error; This represents the mean of the estimated measurement error values; This represents the variance of the estimated measurement error.

[0071] Example 1 This invention discloses a metrological verification method for electric vehicle charging piles based on an improved infectious disease model. Utilizing the flexible mobility of EVs, it establishes an EV-charging pile charging interaction network through a many-to-many interaction process between the EV and the charging pile, enabling value transfer and metrological verification. The architecture diagram of the proposed electric vehicle charging pile metrological verification method based on the improved infectious disease model is shown below. Figure 1 As shown. Initially, the legal metrology unit manually verifies c selected charging piles to calibrate their measurement error to zero, meaning their charging measurements are all benchmark values, and the status changes from CE to CI. The manually calibrated charging piles will provide charging services to multiple EVs for a period of time. After the charging process is completed, the EV's charging time-series characteristic data and the charging pile's charging interaction data are uploaded to the cloud. The charging capacity of the verified non-out-of-tolerance charging piles can be used to verify the EV's charging capacity. Furthermore, by using Attention-GRU intelligent training, the charging time-series characteristic curves of the same model EV are calculated from the EV's charging time-series characteristic curves, allowing for the calibration of the next charging pile to be inspected. This fully utilizes the interaction between charging piles and EVs and the flexible mobility of EVs to achieve rapid dissemination of charging verification. For charging piles suspected of being out of tolerance, considering the economic losses they may cause to trading users and grid operators, they are marked as NCI and their charging service is stopped until on-site calibration by personnel.

[0072] The specific implementation steps are as follows: Step 1: Select the area to be calibrated, determine the number of charging piles to be manually calibrated by the Metrology Institute, and perform manual calibration.

[0073] Step 2: Calibrate the metering status of all charging piles in the inspection area.

[0074] Step 3: Analyze the massive amount of charging interaction data of electric vehicles and off-board chargers obtained from the cloud platform, identify key features related to the accumulated charging energy, and extract sample data with a uniform data length.

[0075] Step 4: Construct the Attention-GRU model according to equations (8) to (13), as follows: Figure 2 As shown, the selected key charging features are used as model input, feature fusion is achieved through convolution operation, attention weights are mined through the attention mechanism, deep mining and prediction are performed through GRU network, and the model is trained through the charging interaction data of electric vehicles and standard charging piles to achieve accurate prediction of the cumulative energy of benchmark charging.

[0076] Step 5 According to equations (14) to (15), the charging piles undergoing metrological verification upload the data of their charging interaction with the verification vehicle to the cloud model to solve the baseline cumulative charging energy change, calculate the estimated value of the metrological error, and determine the metrological status change through the metrological error threshold.

[0077] Step 6. Based on equations (1) to (7) and (24) to (25), calculate the number of charging piles in different states in the inspected area, calculate the degradation intensity of charging piles in non-out-of-tolerance states, and determine the specific charging piles that have degraded based on the degradation intensity ranking and the number of degraded individuals.

[0078] Step 7: After each round of measurement status calculation, recalibrate the status of charging piles within the inspection area and present the inspection results for each round, such as... Figure 3 As shown, the results of each round of inspection for charging piles numbered 1 to 15 are presented, including the dynamic process of degradation and re-inspection, until the inspection is completed. Figure 4 As shown, this presents the number of charging piles in different final states of the inspected area.

[0079] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A metrological verification method for electric vehicle charging piles based on an improved infectious disease model, characterized in that: Includes the following steps: Step 1: Construct a remote metering verification model for charging piles based on an improved susceptible-infection-susceptible propagation model, based on the vehicle-charging pile interaction scenario; Step 2: Solve the remote metering verification model of charging piles based on the improved SIS propagation model constructed in Step 1 based on the mean field theory, and obtain the continuous and dynamic analysis results of the dynamic behavior of metering verification propagation. Step 3: Based on the continuous and dynamic analysis results of the dynamic behavior of the measurement verification propagation obtained in Step 2, select the key features that affect the charging power, and use the cloud-based benchmark charging cumulative energy calculation method based on residual convolutional neural network to fit the electric vehicle side charging feature data and the charging pile cumulative charging energy curve. Obtain the charging pile measurement error through the interval charging power change value, and judge the charging pile measurement status based on the binary method to realize the charging pile measurement situation awareness. Step 4: Based on the charging pile metering situation awareness results obtained in Step 3, calculate the degradation intensity of charging piles in non-out-of-tolerance state, determine the specific charging piles that have degraded, recalibrate the status of charging piles in the inspection area, and present the inspection results for each round until the inspection is completed.

2. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 1, characterized in that: The specific steps of step 1 include: (1) Determine the verification status of the charging pile: In the remote metering verification model of the charging pile based on the improved SIS propagation model, the verification status of the charging pile is defined as: verification status CE, metering non-out-of-tolerance status CI, metering out-of-tolerance status NCI and verification status CS. When a charging pile in the pending verification state (CS) charges a calibration vehicle, a metrological verification action occurs, and the state changes to the verification state (CE). During the charging process, charging time sequence characteristic data and charging interaction data are uploaded to the cloud-based metrological error calculation model to solve for the metrological error value of the charging pile. The charging pile is judged to be out of tolerance based on the measurement error value, and jumps to the measurement non-tolerance state CI or the measurement out-of-tolerance state NCI. The charging pile that has been transformed into the measurement non-tolerance state CI becomes a new source of propagation. By charging the uncalibrated electric vehicles, it continues to propagate its state. As the charging process continues, there is a probability that the measurement accuracy of the charging pile that is already in the measurement non-tolerance state CI will degrade. Therefore, it may be transformed back into the verification pending state CS and wait to be re-verified. (2) Based on the verification status of the charging pile determined in step (1), construct a remote metering verification model for the charging pile based on the improved SIS propagation model: model the charging pile as a node in the metering network diagram. v i Using quadruples S ( v i )={ λ ( v i ), state ( v i , t ), τ ( v i ), β ( v i )}represent v i The attributes in the network are used to model the EVs-charging pile charging interaction network, and then the construction of a remote metering verification model for charging piles based on the improved SIS propagation model is completed. in, λ ( v i )represent v i Measurement and verification of propagation rate; state ( v i , t ) indicates a non-vehicle-mounted charger v i In time t The state at any given moment; τ ( v i ) represents the intensity of the degradation from the CI state to the CS state; β ( v i ) is a BOOL type variable. β ( v i The value 1 indicates a transition from the CE state to the CI state. β ( v i )=0 indicates a transition from CE state to NCI state.

3. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 1, characterized in that: The specific steps of step 2 include: (1) After each round of metrological verification, the number of charging piles in CS, CE, and CI states changes as follows: (1)； (2)； (3)； In the formula, , and The percentages of charging piles in CS, CE, and CI states are respectively; the initial propagation source formation time is taken as the starting time, and each time interval Δt is recorded as one propagation round; λ(t) is the number of propagation rounds. t Wheel measurement verification of average propagation rate, λ( t )=1 / h ∑hi=1λ( v i ,△ t ), v i ∈CI, h Let be the total number of nodes with propagation capability in round t; ( t ) is the first t Wheel measurement non-out-of-tolerance state transition rate, ( t ) is the first t Wheel measurement non-out-of-tolerance state transition rate, ( t )=1 / q ∑hi=1 β ( v i ), v i ∈CE, q The total number of nodes in the CE state in round t; r ( t ) is the first t Wheel metrological verification degradation rate; (2) Using mean-field theory, the equations of the SIS model can be solved: (4)； Furthermore, charging stations in different states have the following relationships: (5)； (6)； Based on the above equation, the proportion of charging piles in different states at (t+△t) rounds can be obtained, thus obtaining continuous and dynamic analysis results of the dynamic behavior of measurement verification propagation.

4. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 1, characterized in that: The specific steps of step 3 include: (1) Based on the continuous and dynamic analysis results of the dynamic behavior of metrological verification propagation obtained in step 2, select the key features that affect the charging capacity; (2) Based on the key features that affect the charging power selected in step (1), a cloud-based benchmark charging cumulative energy calculation method based on the residual convolutional neural network Attention-GRU model is constructed. The electric vehicle side charging feature data and the charging pile cumulative charging energy curve are fitted, and the metering error of the charging pile is obtained through the interval charging power change value. (3) Based on the obtained measurement error estimate, the metering status of the charging pile is judged by the binary method to realize the metering situation perception of the charging pile.

5. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 4, characterized in that: The specific steps of step 3, step (1) include: ① The correlation between time-series characteristic data and standard charging cumulative power data under the SOC change during EV charging was calculated by Pearson correlation coefficient, and the key factors affecting charging power were analyzed. (7)； In the formula, x i For the charging timing feature data of the i-th EV, y The standard accumulated charging point energy data for charging piles, and These are the average values ​​of EV charging time-series characteristic data and the cumulative charging point energy data of charging piles, respectively. ② Select the features that are strongly correlated with the accumulated charge under standard charging, namely SOC, BMS voltage, battery pack maximum temperature, and battery pack minimum temperature, as inputs for the subsequent model.

6. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 4, characterized in that: The specific steps of step 3, step (2) include: ① Based on the key features affecting the charging capacity selected in step (1), the information weight values ​​at each time point are first calculated, then the weight values ​​are normalized using the softmax function, and finally the normalized attention weight values ​​are assigned to each time point of the input vector. The calculation process is as follows: (8)； (9) ； In the formula, The input feature tensor; This is a dimensional transformation operation; This is a fully connected operation; These are trainable parameters; This is an element-wise multiplication operation; After the attention mechanism, a gated recurrent unit (GRU) is used to extract one-dimensional temporal features. The GRU gated unit includes an input gate, an update gate, and a reset gate. in, The input information of the gating unit at the current moment, This represents the hidden state information passed down from the previous moment. This represents the output information of the gating unit at the current moment. This represents updating the output information of the gate at the current moment. This represents the output information for resetting the door at the current moment; The formula for calculating the amount of data that can be retained up to the current time by updating the gate's memory information is as follows: (10)； In the formula, This represents a one-dimensional convolution operation; This represents the training parameters of the convolution kernel for updating the gate; (11)； In the formula, The training parameters of the convolution kernel representing the reset gate; The output of the gating unit at the current moment is obtained by the following formula: (12)； (13)； In the formula, Represents the training parameters of the convolution kernel; ② The benchmark cumulative charging point energy is reconstructed through this network. When EVs verify charging piles, the actual cumulative charging energy can be predicted based on the trained model and compared with the charging energy reading on the pile side. This enables the CE status node in the metering verification network to perceive the metering status, thereby completing the fitting of the electric vehicle side charging characteristic data and the charging pile cumulative charging energy curve. ③ Calculate the change in accumulated charging energy by analyzing the output sequence of actual accumulated charging energy of the EV. The corresponding indicated energy change value of the tested charging pile Calculate the estimated relative measurement error of the charging pile under inspection; (14)。 7. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 4, characterized in that: The specific method for step (3) of step 3 is as follows: After obtaining the estimated measurement error, a measurement relative error threshold is set. Determine C E Whether the metering status has changed is determined and calibrated using a binary method, thus realizing the metering status awareness of the charging pile; (15)； In the formula, The standard measurement relative error for secondary charging piles, .

8. The method for metrological verification of electric vehicle charging piles based on an improved infectious disease model according to claim 1, characterized in that: The specific steps of step 4 include: (1) Based on the charging pile metering situation awareness results obtained in step 3, the degradation intensity of charging piles in non-out-of-tolerance state is calculated by using the dynamic calculation method of metering verification degradation rate. (24)； In the formula, For a fixed degradation rate; (2) Each CI calculates its own metering performance degradation intensity according to formula (10), and determines the specific degraded charging pile individual based on the ranking of metering performance degradation intensity and the number of degradations; (25) ； In the formula, The cumulative number of times a charging station has been marked as CI status until the current propagation test identifies it as charging an electric vehicle. The threshold number of charging cycles; For node v i Metrological verification of degradation intensity; The initial metrological performance degradation intensity; This is an estimate of the measurement error; This represents the mean of the estimated measurement error values; The variance of the estimated measurement error; (3) After each round of measurement status solution, the status of the charging piles in the inspection area is recalibrated, and the inspection results of each round are presented until the inspection is completed.

Images