Electric vehicle charging out-of-tolerance measurement modeling method, system, equipment and medium
By constructing a charging environment topology and a neural network constant differential equation model, the problems of nonlinearity in metering error and low response rate of electric vehicle charging facilities are solved, achieving high-precision metering and prediction, and supporting intelligent management and fair billing of electric vehicle charging systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-31
- Publication Date
- 2026-03-20
AI Technical Summary
Existing metering models for electric vehicle charging facilities do not fully consider the interaction between battery properties and complex charging environments, resulting in nonlinear metering errors and low response rates, which affect the fairness of user billing and the accuracy of grid load forecasting.
By acquiring operational data from the vehicle, charging pile, and cloud, a charging environment topology is constructed and the energy conservation relationship is analyzed. By combining the neural network constant differential equation model to learn the nonlinear dynamic characteristics of the measurement error, accurate measurement of the charging process is achieved.
It significantly improves the metering accuracy and real-time performance of electric vehicle charging facilities under complex operating conditions, enhances metering response rate and prediction accuracy, and supports intelligent management and fair billing of charging systems.
Smart Images
Figure CN121706526A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of metering error technology for electric vehicle charging facilities, and in particular to a method, system, equipment, and medium for metering error modeling of electric vehicle charging. Background Technology
[0002] As electric vehicles become increasingly popular, more and more users are using them as their primary mode of daily transportation. To meet the ever-growing charging demand, various regions are accelerating the construction of charging piles and stations. With the proliferation of these charging facilities, the issue of accurate metering becomes increasingly prominent. Metering results not only directly affect the accuracy of recording users' actual electricity consumption but also impact the fairness of electricity billing. Furthermore, the power grid relies on this data to understand load changes and plan power dispatch; inaccurate metering will lead to deviations in forecasts, affecting operational decisions.
[0003] During actual charging, the characteristics of the battery itself change continuously with the charging stage, its age, and ambient temperature. Coupled with external factors such as grid voltage fluctuations, line impedance, and equipment heat generation, these factors often lead to non-linear errors in the electricity meter. This error is particularly pronounced under high-power fast charging or low-temperature conditions. In such cases, charging facilities become sluggish in responding to metering deviations exceeding permissible limits, resulting in a "low out-of-tolerance metering response rate." This can lead to discrepancies in how much users should pay, potentially resulting in overpayment or underpayment. If such problems persist in the data collected by the grid, the accuracy of load forecasting will also be affected.
[0004] Therefore, traditional models with fixed parameters are no longer sufficient to handle such complex operating conditions. What is needed now is a modeling method that can dynamically adjust based on battery status, environmental conditions, and equipment operation, resulting in more accurate measurements and faster response. Currently, such dynamic modeling schemes capable of identifying and correcting out-of-tolerance measurements in real time are relatively lacking and urgently require further research and improvement. Summary of the Invention
[0005] In view of the aforementioned existing problems, the present invention is proposed.
[0006] Therefore, this invention provides a method, system, device, and medium for electric vehicle charging out-of-tolerance measurement modeling to solve the problems of low out-of-tolerance measurement response rate and insufficient prediction accuracy caused by existing electric vehicle charging facility measurement models not fully considering the interaction between battery properties and complex charging environment, and being affected by nonlinear error interference.
[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for modeling the charging tolerance measurement of electric vehicles, comprising: By acquiring vehicle-side operation data, charging pile-side operation data, and cloud-based operation data, and by constructing a charging environment topology and analyzing the energy conservation relationship of the charging station area, the metering error prediction results of electric vehicle charging facilities are obtained. Based on the measurement error prediction results, the measurement uncertainty is calculated by obtaining the voltage estimation error and combining it with statistical theory, resulting in a set of quantitative indicators. By using a set of quantitative indicators as features to input into a neural ordinary differential equation model, and using the neural ordinary differential equation model to learn the differential law of the evolution of the input features over time, a matching result for the nonlinear dynamic characteristics of the metering error during the charging process is obtained. Based on the matching results, the out-of-tolerance measurement results of the electric vehicle charging facilities are obtained by training the ordinary differential equation model and inputting the real-time operation data of the electric vehicle charging facilities into the trained model.
[0008] As a preferred embodiment of the electric vehicle charging deviation measurement modeling method described in this invention, the analysis of the energy conservation relationship of the charging station area includes: The total power supply measured by the main meter in the transformer substation is broken down into the power consumption of each electric vehicle charging facility, the power conversion loss inside each charging facility, the power consumption of other fixed loads in the transformer substation, and the sum of line transmission losses. Based on the principle of energy conservation, calculate the theoretical actual amount of electricity flowing through each electric vehicle charging facility; Based on the deviation between the theoretical actual electricity value and the meter reading of the charging pile, the metering error prediction result of the electric vehicle charging facility is obtained.
[0009] As a preferred embodiment of the electric vehicle charging deviation measurement modeling method described in this invention, the analysis of the energy conservation relationship of the charging station area further includes: Based on the vehicle-side data, charging pile-side data, and cloud-based data, a charging environment topology is constructed. The topology includes the physical connection relationships and service correspondence relationships between vehicles, charging piles, and the regional power grid. In the charging environment topology, vehicle data belonging to the same service relationship are associated with charging pile data, and the power grid load data of the area where the charging pile is located is bound. Based on the structured data after association and binding, within the framework of the topology, the energy flow and distribution relationship of the charging station area is analyzed according to the conservation principle that the total input power is equal to the sum of the power consumed by each branch.
[0010] The beneficial effect of this preferred technical solution is that by constructing and performing data association and energy conservation analysis based on the vehicle-pile-cloud collaborative charging environment topology, it achieves accurate mapping of the power flow of each unit in the charging area, providing an accurate data basis for quantifying the metering error of electric vehicle charging facilities.
[0011] As a preferred embodiment of the electric vehicle charging error measurement modeling method of the present invention, the step of obtaining the voltage estimation error includes: Real-time collection of voltage measurements from electric vehicle charging facilities and vehicle battery management systems; The voltage measurement value at the charging facility terminal is compared with the voltage measurement value of the vehicle battery management system; The voltage estimation error is obtained by dynamically correcting the comparison results and the measurement error prediction results.
[0012] As a preferred embodiment of the electric vehicle charging deviation measurement modeling method of the present invention, the calculation of measurement uncertainty includes: A multidimensional error distribution matrix is constructed based on the voltage estimation error; The normality of the error distribution matrix is tested and parameters are estimated to determine the mean and variance of the error data distribution. Based on the statistical confidence interval theory, and combined with the preset confidence level and distribution parameters, the measurement uncertainty is calculated.
[0013] As a preferred embodiment of the electric vehicle charging deviation measurement modeling method of the present invention, the training of the neural network differential equation model includes: Construct a training dataset that takes a set of quantitative indicators as input and outputs the true values of historical deviation measurements; Define a composite loss function that includes a prediction error term and a model complexity penalty term; The gradient descent algorithm is used to iteratively optimize the network parameters of the neural ordinary differential equation model until the loss function converges.
[0014] The beneficial effects of this preferred technical solution are that by constructing a training dataset with quantitative indicators as input and historical true values as output, and by using a composite loss function that includes a penalty term for prediction error and model complexity for gradient descent iterative optimization, accurate training of the parameters of the neural ordinary differential equation model is achieved, which effectively improves the model's ability to learn and predict the dynamic characteristics of out-of-tolerance metering of charging facilities.
[0015] As a preferred embodiment of the electric vehicle charging tolerance measurement modeling method described in this invention, the method for obtaining the tolerance measurement results of electric vehicle charging facilities includes: Set the neural network weights and bias parameters in the converged neural ordinary differential equation model to a fine-tunable state. During the model inference phase, real-time operational data of electric vehicle charging facilities are used as initial conditions input into the model. By solving the system of ordinary differential equations defined by the model, continuous out-of-tolerance measurement numerical outputs can be obtained at any specific charging time.
[0016] Secondly, the present invention provides an electric vehicle charging tolerance measurement modeling system, comprising: The error screening module is used to acquire vehicle-side operation data, charging pile-side operation data, and cloud-based operation data. By constructing the charging environment topology and analyzing the energy conservation relationship of the charging station area, it obtains the metering error prediction results of electric vehicle charging facilities. The uncertainty quantification module is used to obtain a set of quantification indicators by acquiring the voltage estimation error and combining it with statistical theory to calculate the measurement uncertainty based on the measurement error prediction results. The dynamic characteristic matching module is used to input a set of quantitative indicators as features into a neural ordinary differential equation model, and use the neural ordinary differential equation model to learn the differential law of the evolution of the input features over time, so as to obtain the matching result of the nonlinear dynamic characteristics of the metering error during the charging process. The out-of-tolerance measurement output module is used to obtain the out-of-tolerance measurement results of electric vehicle charging facilities by training a neural network constant differential equation model based on the matching results and inputting the real-time operation data of electric vehicle charging facilities into the trained model.
[0017] Thirdly, the present invention provides an electronic device, comprising: Memory, used to store programs; A processor is configured to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the electric vehicle charging tolerance measurement modeling method.
[0018] Fourthly, the present invention provides a computer-readable storage medium, comprising: when the program is executed by a processor, the steps of implementing the electric vehicle charging tolerance measurement modeling method.
[0019] The beneficial effects of this invention are as follows: By acquiring vehicle-side operation data, charging pile-side operation data, and cloud-based operation data, and analyzing the energy flow in the charging area based on the energy conservation relationship, this invention achieves preliminary and accurate prediction of metering errors in electric vehicle charging facilities, solving the problem of low error response rate caused by neglecting multi-source data interaction in traditional methods. By extracting voltage estimation errors from the metering error prediction results and calculating measurement uncertainty using statistical theory, a set of quantitative indicators characterizing error characteristics is formed, achieving refined quantification of the static distribution characteristics of metering errors and providing highly reliable input features for dynamic modeling. By using the set of quantitative indicators as input, the differential law of error characteristics evolving over time is learned using a neural ordinary differential equation model, and iterative training and parameter optimization are performed using a loss function to achieve dynamic matching and correction of highly nonlinear metering errors during charging. Finally, a high-precision dynamic model capable of outputting out-of-tolerance metering results at specific times is constructed, significantly improving metering accuracy and system robustness. Attached Figure Description
[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a basic flowchart illustrating a method for measuring and modeling the charging deviation of an electric vehicle according to an embodiment of the present invention. Figure 2 This is a topology diagram of the charging environment of an electric vehicle charging facility, provided by an embodiment of the present invention, for an electric vehicle charging tolerance measurement modeling method. Figure 3 A nonlinear dynamic characteristic matching diagram based on a neural ODE model is provided as an embodiment of the present invention for a method for modeling the metering deviation of electric vehicle charging. Figure 4 An experimental platform diagram of a method for measuring and modeling charging deviations in electric vehicles, provided in one embodiment of the present invention; Figure 5 A graph showing the prediction results of metering errors for electric vehicle charging facilities, provided by an embodiment of the present invention, for an electric vehicle charging out-of-tolerance metering modeling method. Figure 6 The figure shows the experimental results of the model error response rate of a method for modeling the out-of-tolerance metering of electric vehicle charging, which is provided in an embodiment of the present invention. Detailed Implementation
[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0022] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for metering and modeling electric vehicle charging deviations is provided, comprising: S100: Acquire vehicle-side operation data, charging pile-side operation data, and cloud-based operation data. By constructing the charging environment topology and analyzing the energy conservation relationship of the charging station area, the metering error prediction results of electric vehicle charging facilities are obtained. S200: Based on the measurement error prediction results, the measurement uncertainty is calculated by obtaining the voltage estimation error and combining it with statistical theory to obtain a set of quantitative indicators; S300: By using the set of quantitative indicators as features as inputs to the neural ordinary differential equation model, the neural ordinary differential equation model is used to learn the differential law of the evolution of the input features over time, and the matching result of the nonlinear dynamic characteristics of the metering error during the charging process is obtained. S400: Based on the matching results, the abnormal measurement results of electric vehicle charging facilities are obtained by training the ordinary differential equation model and inputting the real-time operation data of electric vehicle charging facilities into the trained model.
[0023] It should be noted that existing metering models for electric vehicle charging facilities face a series of challenges during operation. These include the fact that the charging process is affected by multiple factors such as battery chemical characteristics, ambient temperature, charging power fluctuations, and vehicle-charging pile interaction protocols, resulting in highly nonlinear and dynamic time-varying metering errors that are difficult to accurately capture using traditional static or linear models. Furthermore, the models lack deep integration of multi-source information, including data from the vehicle-side battery management system, real-time charging pile operation status, and cloud-based grid load, leading to biased metering error analysis and low out-of-range response rates. In addition, existing methods often neglect the energy conservation relationship at the transformer substation level and the topological connections between devices, making it impossible to accurately pinpoint the source of errors. This results in the accumulation of metering deviations, affecting user billing fairness and the accuracy of grid load forecasting.
[0024] Therefore, addressing the issues of low out-of-tolerance response rate and insufficient prediction accuracy in existing electric vehicle charging facility metering models due to insufficient consideration of the interaction between battery properties and complex charging environments, and interference from nonlinear errors, this paper proposes a solution through steps S100-S400. This solution integrates multi-source operational data from the vehicle, charging pile, and cloud platforms to construct a charging environment topology and perform preliminary prediction of metering errors based on energy conservation principles. It then extracts voltage estimation errors and measurement uncertainties to form quantitative characteristics. Finally, it introduces a neural network constant differential equation model to accurately match and iteratively optimize the nonlinear dynamic characteristics of the errors. This effectively solves the problems of low out-of-tolerance response rate and insufficient prediction accuracy caused by neglecting multi-source collaboration, nonlinear dynamic evolution, and topological correlation in existing metering models. This significantly improves the metering accuracy and real-time performance of electric vehicle charging facilities under complex operating conditions, providing robust technical support for intelligent management and fair billing of charging systems.
[0025] Example 2, refer to Figures 2-3 As one embodiment of the present invention, based on the previous embodiment, a method for metering and modeling electric vehicle charging tolerances is provided, comprising: In this embodiment, in step S100, the vehicle-side operating data refers to the vehicle battery status information, the charging pile-side operating data refers to the output power of the electric vehicle charging facility and the error value of the metering module, and the cloud-based operating data refers to the regional power grid load data. Based on the acquired data, a charging environment topology is established, such as... Figure 2 As shown.
[0026] In this embodiment of the application, the construction of the charging environment topology in step S100 includes establishing a graph structure based on vehicle-side, charging pile-side, and cloud-side data, which includes vehicles, charging piles, and regional power grid nodes, as well as their physical connection relationships and service correspondence relationships. Under this topology framework, vehicle and charging pile data under the same service relationship are associated, and the power grid load data of the area where the charging pile is located is bound.
[0027] In an optional implementation, the charging environment topology structure constructed in step S100 can also include a vehicle table, a charging pile table, and a power grid load table. The data relationship between the three can be established by using service ID and timestamp as foreign keys, and SQL join table queries can be used to obtain multi-source operating data under the same charging event for energy conservation analysis.
[0028] In an optional implementation, the charging environment topology structure constructed in step S100 can also embed a unified service ID and timestamp tag in the data streams of the vehicle, the charging pile, and the cloud. The data is received through a message queue and matched in real time based on the tag. Multi-source data belonging to the same service relationship are dynamically aggregated to form a data unit for energy conservation analysis.
[0029] In this embodiment of the application, step S100, analyzing the energy conservation relationship of the charging station area, includes: The total power supply measured by the main meter in the transformer substation is broken down into the power consumption of each electric vehicle charging facility, the power conversion loss inside each charging facility, the power consumption of other fixed loads in the transformer substation, and the sum of line transmission losses. Based on the principle of energy conservation, calculate the theoretical actual amount of electricity flowing through each electric vehicle charging facility; Based on the deviation between the theoretical actual electricity value and the meter reading of the charging pile, the metering error prediction result of the electric vehicle charging facility is obtained.
[0030] In this embodiment of the application, step S100, analyzing the energy conservation relationship of the charging station area, further includes: Based on the vehicle-side data, charging pile-side data, and cloud-based data, a charging environment topology is constructed. The topology includes the physical connection relationships and service correspondence relationships between vehicles, charging piles, and the regional power grid. In the charging environment topology, vehicle data belonging to the same service relationship are associated with charging pile data, and the power grid load data of the area where the charging pile is located is bound. Based on the structured data after association and binding, within the framework of the topology, the energy flow and distribution relationship of the charging station area is analyzed according to the conservation principle that the total input power is equal to the sum of the power consumed by each branch.
[0031] In this embodiment, the energy conservation relationship is quantified using formula (1): (1) In the formula, This represents the serial number of the electric vehicle charging facility. Represents the total number of electric vehicle charging facilities. This represents the actual amount of electricity flowing through the main meter of the transformer substation. This represents the total electricity supply from all electric vehicle charging facilities within the designated area. This represents the charge level of the DC metering module inside the designated electric vehicle charging facility. This represents the power loss value of the internal power module of a specified electric vehicle charging facility. This represents the total electricity consumption within the transformer substation area, excluding electric vehicle charging facilities. This represents the total line loss within the transformer area.
[0032] In this embodiment of the application, based on the energy conservation analysis results, the measurement error between the master table and the sub-tables is calculated using formula (2): (2) In the formula, The meter readings represent the measurement values displayed in the main meter for the substation area. Represents the specified sub-meter measurement value. This represents the measurement error in the summary table. This represents the measurement error of the specified sub-meter. This represents the actual amount of electricity flowing through the designated electric vehicle charging facility.
[0033] Finally, the above results are substituted into formula (3) to verify and output the error prediction results: (3) In the formula, This represents the fixed loss value.
[0034] In this embodiment of the application, the fixed loss value It is obtained by statistically averaging the historical data of the empty operation of the transformer substation.
[0035] In this embodiment of the application, obtaining the voltage estimation error in step S200 includes: Real-time collection of voltage measurements from electric vehicle charging facilities and vehicle battery management systems; The voltage measurement value at the charging facility terminal is compared with the voltage measurement value of the vehicle battery management system; The voltage estimation error is obtained by dynamically correcting the comparison results and the measurement error prediction results.
[0036] In this embodiment, the voltage estimation error is calculated using formula (4): (4) In the formula, This represents the voltage estimation error of electric vehicle charging facilities. The voltmeter reading represents the voltage of an electric vehicle charging facility. The value of the BMS voltmeter represents the electric vehicle's voltage. This represents the actual correction value (this value is the predicted measurement error value obtained based on S100).
[0037] In this embodiment of the application, the calculation of measurement uncertainty in step S200 includes: A multidimensional error distribution matrix is constructed based on the voltage estimation error; The normality of the error distribution matrix is tested and parameters are estimated to determine the mean and variance of the error data distribution. Based on the statistical confidence interval theory, and combined with the preset confidence level and distribution parameters, the measurement uncertainty is calculated.
[0038] In this embodiment of the application, the step S200 of calculating the measurement uncertainty includes constructing a multidimensional error distribution matrix based on the voltage estimation error, performing parameter estimation after passing the normality test to determine the mean and variance of the error distribution, and then calculating the measurement uncertainty by combining the statistical confidence interval theory of the preset confidence level.
[0039] In an alternative implementation, the measurement uncertainty calculation in step S200 can also be based on a multidimensional error distribution matrix constructed from the voltage estimation error, directly using the quantiles of the sample data (such as 2.5% and 97.5%) to determine the confidence interval, and using the width of the interval as the measurement uncertainty, without relying on the specific parameter assumptions of the error distribution.
[0040] In an optional implementation, the measurement uncertainty calculation in step S200 can also be based on constructing a multidimensional error distribution matrix based on the voltage estimation error, setting the prior distribution of the error distribution parameters, updating the posterior distribution using Bayes' formula in combination with the observation data, and using the standard deviation or confidence interval width of the posterior distribution as the measurement uncertainty.
[0041] In this embodiment of the application, an error matrix is established using formula (5) to estimate the standard vehicle voltage error value: (5) In the formula, This represents the estimated standard vehicle voltage error value. This represents the voltage reading of the corresponding electric vehicle charging facility. This represents the voltage indication error value of the corresponding electric vehicle charging facility.
[0042] And the statistical parameters corresponding to the measurement uncertainty are calculated using formula (6): (6) In the formula, This represents the mean of the corresponding normal distribution. This represents the variance of the corresponding normal distribution.
[0043] In this embodiment, a neural ODE model is established, and the results obtained above are used as feature input values to achieve feature interaction and matching of nonlinear dynamic characteristics. The specific process is as follows: Figure 3 As shown.
[0044] In this embodiment of the application, step S400, training the neural ordinary differential equation model, includes: Construct a training dataset that takes a set of quantitative indicators as input and outputs the true values of historical deviation measurements; Define a composite loss function that includes a prediction error term and a model complexity penalty term; The gradient descent algorithm is used to iteratively optimize the network parameters of the neural ordinary differential equation model until the loss function converges.
[0045] In this embodiment of the application, the training of the ordinary differential equation model in step S400 includes using a set of quantitative indicators as input features, constructing an ordinary differential equation system with the state change rate defined by a neural network, minimizing the composite loss function containing the prediction error and model complexity penalty terms through gradient descent to perform parameter iterative optimization, and solving the differential equation system in the inference stage to output the continuous deviation measurement result at any time.
[0046] In an optional implementation, the training of the ordinary differential equation model in step S400 can also use the set of quantitative indicators as input features to construct an ordinary differential equation system, and use an adaptive optimizer such as Adam or RMSProp to replace the gradient descent method. The learning rate is dynamically adjusted according to the historical gradient of the parameters to iteratively optimize the network parameters until the composite loss function converges, and then the model outputs a continuous deviation measurement result.
[0047] In an optional implementation, the training of the neural ordinary differential equation model in step S400 can also be carried out by dividing the training set and the validation set when training the model based on the set of quantized indicators, monitoring the loss value on the validation set, and terminating the training in advance when the loss value no longer decreases for several consecutive rounds, so as to prevent overfitting and ensure that the model obtains the best generalization performance.
[0048] In this embodiment, the set of quantification indicators consists of voltage estimation error values and their corresponding measurement uncertainties, forming a two-dimensional feature vector; after the real-time running data is processed by the feature extraction module, a feature vector consistent with that of the training phase is generated, and it is used as the initial state vector h(t) of the model.
[0049] In this embodiment, the loss function is defined by formula (8): (8) In the formula, This represents the total number of samples used for training. Represents the true value. The predicted value represents the model's value. This represents the total number of weights in the model. This represents the weight value that controls regularization. This represents the corresponding hyperparameter value.
[0050] In this embodiment of the application, obtaining the out-of-tolerance measurement result of the electric vehicle charging facility in step S400 includes: Set the neural network weights and bias parameters in the converged neural ordinary differential equation model to a fine-tunable state. During the model inference phase, real-time operational data of electric vehicle charging facilities are used as initial conditions input into the model. By solving the system of ordinary differential equations defined by the model, continuous out-of-tolerance measurement numerical outputs can be obtained at any specific charging time.
[0051] In this embodiment of the application, the output method of the out-of-tolerance measurement result in step S400 includes setting the parameters of the neural ordinary differential equation model after training convergence to a fine-tunable state, using the real-time operation data of electric vehicle charging facilities as the initial condition input to the model, and obtaining the continuous out-of-tolerance measurement numerical output for any specific charging time by numerically solving its defined ordinary differential equation system.
[0052] In an optional implementation, the output method of the out-of-tolerance measurement result in step S400 can also divide the charging process into fixed time intervals, use real-time running data as the initial input, use the model to predict the out-of-tolerance measurement value at the next moment in each time step, and generate a discrete measurement result sequence of the entire charging process through recursive calculation.
[0053] In an optional implementation, the output method of the out-of-tolerance measurement result in step S400 can also use a sequence model (such as LSTM or Transformer) instead of the neural ordinary differential equation model, taking the multi-source operation data of vehicle-pile-cloud over a period of time as input, and directly regressing to predict the out-of-tolerance measurement result at a specific time in the future or during the entire charging period.
[0054] In this embodiment, the differential function f in the neural network constant differential equation model is a three-layer fully connected neural network with an input layer dimension of d, a hidden layer neuron count of h, an activation function of ReLU, and an output layer dimension consistent with the dimension of the state vector h(t).
[0055] In this embodiment, the dynamic evolution of out-of-tolerance measurement is described by formulas (7) and (9): (7) (9) In the formula, Represents time, Represents the change over time. The initial state vector representing the out-of-tolerance measurement values of electric vehicle charging facilities. This represents the final dynamic result of the out-of-tolerance measurement. Represents the specific charging time. These represent the parameter values that allow the neural network to be trained.
[0056] Example 3, referring to Figures 4-6Table 1 illustrates one embodiment of the present invention, which provides a method for modeling the metering deviation of electric vehicle charging. To verify the beneficial effects of the present invention, scientific demonstration is conducted through specific implementation methods and their effects.
[0057] The specific details of this embodiment are as follows: To verify the effectiveness and practicality of the electric vehicle charging out-of-tolerance measurement dynamic model established in this invention, an electric vehicle charging station with a long operating history was selected as the experimental scenario. The 16 DC charging piles in this station exhibit varying degrees of measurement deviation due to long-term operation, demonstrating typical out-of-tolerance measurement characteristics. The experimental platform architecture is as follows: Figure 4 As shown, the system includes a vehicle-side battery management system (BMS) data acquisition module, a pile-side power and metering data acquisition device, a cloud-based regional power grid load access interface, and an edge computing server for deploying this model. The experimental parameters are set as follows: training batch size is 64, training epochs are 500, initial learning rate is 1×10⁻³, and the learning rate decay strategy is 0.5 times decay every 100 epochs.
[0058] To ensure that the model established in this invention can perform out-of-tolerance measurement of electric vehicle charging facilities, this experiment verifies the feasibility of the model. The model was applied to perform out-of-tolerance measurement on 12 charging piles at a charging site. The measurement results displayed in the model allow for real-time monitoring of information such as voltage, current, power, and temperature of the charging piles. Furthermore, the model can correct the out-of-tolerance measurement results based on key information such as battery status, charging demand, and charging amount, thereby obtaining key information such as the battery status, charging demand, and charging amount of the electric vehicle currently charging. The out-of-tolerance measurement results of the electric vehicle charging facility are calculated with high accuracy. This indicates that the model established in this invention has high feasibility.
[0059] In practical applications, electrical energy fluctuates during measurement and transmission, leading to an error of approximately ±1% in the measurement results due to environmental influences. Therefore, a relative error within 1% is considered normal. To ensure the accuracy of the measurement results from the dynamic model for out-of-tolerance metering of electric vehicle charging facilities, this invention's model was used as the experimental group. The remote metering method for off-board electric vehicle chargers based on Bayesian theory served as control group 1, and the DC charging energy metering method incorporating wavelet threshold denoising and smoothing processing served as control group 2. The sampling period for the sub-meters was set to 15 minutes, and the number of sampling points was set to 350. By comparing the relative errors between the predicted and actual results of the three methods, the accuracy of the prediction results was verified. The experimental results are as follows: Figure 5As shown in the figure above, the relative errors of the prediction results of the experimental group models fluctuated within ±1%, while the relative errors of the prediction results of the two control group models all exceeded ±1%, with the control group model 1 even reaching 2%. These experimental results demonstrate that the model established in this invention can accurately predict the metering error of electric vehicle charging facilities, and the relative error of the prediction results can be controlled below 1%, indicating high accuracy.
[0060] To verify the accuracy of the out-of-tolerance measurement results of the electric vehicle charging facilities established by the model of this invention, this experiment used three models to perform out-of-tolerance measurement on 16 electric vehicle charging facilities and obtained the measurement operation error. The results are shown in Table 1.
[0061] Table 1 Comparison of Measurement Operation Errors Exceeding Tolerance
[0062] The above results show that the model established in this invention has a small error in out-of-tolerance measurement. To verify the accuracy of the three models in calculating the out-of-tolerance measurement results of electric vehicle charging facilities, the error response rate is introduced as a metric. The out-of-tolerance measurement error data of electric vehicle charging facilities are calculated based on the models in Table 1 to reflect the reliability of the models in calculating the out-of-tolerance measurement results of electric vehicle charging facilities. The calculation formula is as follows: (10) In the formula, The set value representing the amount of power deviation. This represents the change in the metering error of electric vehicle charging facilities calculated by the model. This represents the model error response rate.
[0063] Experimental results of error response rates for the three models are as follows: Figure 6 As shown in the figure, the blind test results of 20 electric vehicle charging facilities show that the error response rate of the electric vehicle charging facilities established by the present invention is above 95%, while the equipment error response rates of the two control group models are lower than those of the model established by the present invention. This indicates that the model established by the present invention can more accurately capture the dynamic error characteristics of electric vehicle charging facilities and significantly improve the accuracy and robustness of out-of-tolerance measurement of electric vehicle charging facilities.
[0064] Example 4 is an embodiment of the present invention. This embodiment differs from the first embodiment in that it provides an electric vehicle charging deviation measurement modeling system.
[0065] It should be noted that the technical solution of the electric vehicle charging tolerance measurement modeling system is based on the same concept as the above-mentioned electric vehicle charging tolerance measurement modeling method. For details not described in detail in the technical solution of the electric vehicle charging tolerance measurement modeling system in this embodiment, please refer to the description of the above-mentioned electric vehicle charging tolerance measurement modeling method.
[0066] This embodiment provides an electric vehicle charging deviation measurement modeling system, comprising: The error screening module is used to acquire vehicle-side operation data, charging pile-side operation data, and cloud-based operation data. By constructing the charging environment topology and analyzing the energy conservation relationship of the charging station area, it obtains the metering error prediction results of electric vehicle charging facilities. The uncertainty quantification module is used to obtain a set of quantification indicators by acquiring the voltage estimation error and combining it with statistical theory to calculate the measurement uncertainty based on the measurement error prediction results. The dynamic characteristic matching module is used to input a set of quantitative indicators as features into a neural ordinary differential equation model, and use the neural ordinary differential equation model to learn the differential law of the evolution of the input features over time, so as to obtain the matching result of the nonlinear dynamic characteristics of the metering error during the charging process. The out-of-tolerance measurement output module is used to obtain the out-of-tolerance measurement results of electric vehicle charging facilities by training a neural network constant differential equation model based on the matching results and inputting the real-time operation data of electric vehicle charging facilities into the trained model.
[0067] This embodiment also provides an electronic device applicable to a method for modeling the metering deviation of electric vehicle charging, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a method for measuring and modeling the charging tolerance of an electric vehicle, as proposed in the above embodiments.
[0068] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a method for modeling the metering deviation of electric vehicle charging as proposed in the above embodiments.
[0069] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for implementing electric vehicle charging deviation measurement modeling proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.
[0070] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.
[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for measuring and modeling charging tolerances in electric vehicles, characterized in that, include: By acquiring vehicle-side operation data, charging pile-side operation data, and cloud-based operation data, and by constructing a charging environment topology and analyzing the energy conservation relationship of the charging station area, the metering error prediction results of electric vehicle charging facilities are obtained. Based on the measurement error prediction results, the measurement uncertainty is calculated by obtaining the voltage estimation error and combining it with statistical theory, resulting in a set of quantitative indicators. By using a set of quantitative indicators as features to input into a neural ordinary differential equation model, and using the neural ordinary differential equation model to learn the differential law of the evolution of the input features over time, a matching result for the nonlinear dynamic characteristics of the metering error during the charging process is obtained. Based on the matching results, the out-of-tolerance measurement results of the electric vehicle charging facilities are obtained by training the ordinary differential equation model and inputting the real-time operation data of the electric vehicle charging facilities into the trained model.
2. The electric vehicle charging tolerance measurement modeling method as described in claim 1, characterized in that: The analysis of the energy conservation relationship in the charging station area includes: The total power supply measured by the main meter in the transformer substation is broken down into the power consumption of each electric vehicle charging facility, the power conversion loss inside each charging facility, the power consumption of other fixed loads in the transformer substation, and the sum of line transmission losses. Based on the principle of energy conservation, calculate the theoretical actual amount of electricity flowing through each electric vehicle charging facility; Based on the deviation between the theoretical actual electricity value and the meter reading of the charging pile, the metering error prediction result of the electric vehicle charging facility is obtained.
3. The electric vehicle charging deviation measurement modeling method as described in claim 1 or 2, characterized in that: The analysis of the energy conservation relationship in the charging station area also includes: Based on the vehicle-side data, charging pile-side data, and cloud-based data, a charging environment topology is constructed. The topology includes the physical connection relationships and service correspondence relationships between vehicles, charging piles, and the regional power grid. In the charging environment topology, vehicle data belonging to the same service relationship are associated with charging pile data, and the power grid load data of the area where the charging pile is located is bound. Based on the structured data after association and binding, within the framework of the topology, the energy flow and distribution relationship of the charging station area is analyzed according to the conservation principle that the total input power is equal to the sum of the power consumed by each branch.
4. The electric vehicle charging tolerance measurement modeling method as described in claim 3, characterized in that: The voltage estimation error includes: Real-time collection of voltage measurements from electric vehicle charging facilities and vehicle battery management systems; The voltage measurement value at the charging facility terminal is compared with the voltage measurement value of the vehicle battery management system; The voltage estimation error is obtained by dynamically correcting the comparison results and the measurement error prediction results.
5. The electric vehicle charging deviation measurement modeling method as described in claim 4, characterized in that: The calculation of measurement uncertainty includes: A multidimensional error distribution matrix is constructed based on the voltage estimation error; The normality of the error distribution matrix is tested and parameters are estimated to determine the mean and variance of the error data distribution. Based on the statistical confidence interval theory, and combined with the preset confidence level and distribution parameters, the measurement uncertainty is calculated.
6. The electric vehicle charging deviation measurement modeling method as described in claim 5, characterized in that: The training of the constant differential equation model includes: Construct a training dataset that takes a set of quantitative indicators as input and outputs the true values of historical deviation measurements; Define a composite loss function that includes a prediction error term and a model complexity penalty term; The gradient descent algorithm is used to iteratively optimize the network parameters of the neural ordinary differential equation model until the loss function converges.
7. The electric vehicle charging deviation measurement modeling method as described in claim 6, characterized in that: The obtained out-of-tolerance measurement results for electric vehicle charging facilities include: Set the neural network weights and bias parameters in the converged neural ordinary differential equation model to a fine-tunable state. During the model inference phase, real-time operational data of electric vehicle charging facilities are used as initial conditions input into the model. By solving the system of ordinary differential equations defined by the model, continuous out-of-tolerance measurement numerical outputs can be obtained at any specific charging time.
8. A metering and modeling system for electric vehicle charging deviations, using the method described in any one of claims 1-7, characterized in that, include: The error screening module is used to acquire vehicle-side operation data, charging pile-side operation data, and cloud-based operation data. By constructing the charging environment topology and analyzing the energy conservation relationship of the charging station area, it obtains the metering error prediction results of electric vehicle charging facilities. The uncertainty quantification module is used to obtain a set of quantification indicators by acquiring the voltage estimation error and combining it with statistical theory to calculate the measurement uncertainty based on the measurement error prediction results. The dynamic characteristic matching module is used to input a set of quantitative indicators as features into a neural ordinary differential equation model, and use the neural ordinary differential equation model to learn the differential law of the evolution of the input features over time, so as to obtain the matching result of the nonlinear dynamic characteristics of the metering error during the charging process. The out-of-tolerance measurement output module is used to obtain the out-of-tolerance measurement results of electric vehicle charging facilities by training a neural network constant differential equation model based on the matching results and inputting the real-time operation data of electric vehicle charging facilities into the trained model.
9. An electronic device, characterized in that, include: Memory, used to store programs; A processor for loading the program to perform the steps of the method as claimed in any one of claims 1-7.
10. A computer-readable storage medium storing a program, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.