A remote metering method for charging piles

Abstract

The present invention discloses a remote metering method for charging piles, which relates to the technical field of charging pile metering, including: performing stability analysis on the collected charging data, screening out charging data that meets the standards, then sampling, intercepting constant current stage data, analyzing the correlation between each parameter and charging electric energy, and setting a preset number of parameters as charging process characteristics; selecting the interval range value of the battery SOC in the constant current stage data as the starting point and end point of the model input, constructing a feature sequence according to the charging process characteristics, setting the feature sequence as the model input data set, and training a binary classification model and a multi-classification model based on a deep neural network respectively; and classifying and judging the charging data of a new charging pile according to the trained binary classification model and / or multi-classification model. By adopting a deep neural network classification model, the charging data of the charging pile can be analyzed and judged remotely and at a high frequency, realizing real-time monitoring and remote evaluation of the performance of the charging pile.

Description

Technical Field

[0001] The present invention relates to the technical field of charging pile metering, and in particular to a charging pile remote metering method. Background Art

[0002] As an essential supporting facility for electric vehicles, charging piles are increasingly being built on a larger scale and the amount of transactions they generate is also increasing. If there is an error in the measurement accuracy of electric vehicle charging piles, it will cause a deviation in transaction costs, which will bring direct economic losses to electric vehicle charging piles or power grid operating companies.

[0003] Traditional charging pile calibration requires the use of standard calibration devices and the corresponding calibration procedures, which have complex calibration procedures, long time consumption, and huge work costs. A more efficient, intelligent, and remote charging pile measurement performance calibration method is urgently needed. Summary of the invention

[0004] Based on the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a charging pile remote metering method to solve the above-mentioned technical problems.

[0005] To achieve the above object, the present invention provides the following technical solution: a charging pile remote metering method, comprising:

[0006] Collect charging data of the charging pile according to a preset collection frequency, perform stability analysis on the charging data, screen out charging data that meets the standards, and eliminate abnormal charging data;

[0007] Sampling the charging data after stability analysis, intercepting constant current stage data from the sampled charging data, analyzing the correlation between various parameters in the constant current stage data and the charging electric energy, and setting a preset number of parameters as charging process characteristics according to the correlation;

[0008] The interval range value of the battery SOC in the constant current stage data is selected as the starting point and end point of the model input, the charging pile manufacturer and the meter manufacturer are one-hot encoded, and spliced ​​with the charging pile operation years to the charging process characteristics, and a feature sequence is constructed according to the spliced ​​value of the charging process characteristics, and the feature sequence is set as the model input data set;

[0009] According to the model input data set, a binary classification model and a multi-classification model are trained based on a deep neural network, respectively. The binary classification model is used to determine the category of charging data of the charging pile, and the multi-classification model is used to further subdivide the type of charging data category;

[0010] The charging data of the new charging pile is classified and judged according to the trained binary classification model and / or multi-classification model.

[0011] The present invention is further configured such that the charging data of the charging pile includes: charging pile data, transaction data, electric meter data, charging vehicle data, and charging process data; wherein the parameters of the charging pile data include: charging pile manufacturer, charging pile number, charging pile calibration error, and charging pile commissioning life; the parameters of the transaction data include: transaction order number, charging start time, and charging end time; the parameters of the electric meter data include: electric meter manufacturer data; the parameters of the charging vehicle data include: vehicle model, vehicle age, and vehicle driving history; the parameters of the charging process data include: battery SOC, charging pile output voltage, charging pile output current, charging pile output power, ambient temperature, battery required voltage, battery required current, battery charging mode, single cell maximum voltage, single cell maximum current, battery pack maximum temperature, battery pack minimum temperature, charging power, and timestamp.

[0012] The present invention is further configured to perform stability analysis on the charging data, including: for each charging pile, classifying the charging data of connected charging vehicles with the same vehicle model, vehicle age and mileage belonging to the same range into one category, analyzing the variation range of the charging power of the corresponding charging vehicles under the same variation range of the battery SOC, screening the charging data according to the variation range, and screening out the charging data outside the variation range.

[0013] The present invention is further configured to analyze the variation range of the charging power of the corresponding charging vehicle under the same variation range of the battery SOC, and quantify the variation range of the charging power by calculating the variation rate of the charging power of the corresponding charging vehicle when the battery SOC changes at time t. The calculation logic of the variation rate of the charging power is: ΔE i (t) is the rate of change of the charging power of the i-th charging vehicle, n is the total number of timestamps during the charging process, E ij For the i-th charging vehicle at time stamp t j Charging energy, SOC ij For the i-th charging vehicle at time stamp t j The battery SOC at that time, α is the adjustment index, which is used to adjust the sensitivity of the battery SOC to a small change, sign() is the sign function, which is used to determine the direction of the battery SOC change, and the function result is +1 or -1.

[0014] The present invention is further configured to sample the charging data after stability analysis, and the sampling logic is: extracting the charging data corresponding to the initial value of the battery SOC and the first charging data corresponding to each change in the battery SOC from the charging data after stability analysis;

[0015] The logic of intercepting the constant current stage data is: according to the data points of the sampled charging data, extract the voltage value of the next data point greater than or equal to the voltage value of the previous data point, and the difference between the current value of the next data point and the current value of the previous data point does not exceed the preset current threshold;

[0016] Analyze the correlation between each parameter in the constant current stage data and the charging electric energy, and set a preset number of parameters as charging process characteristics according to the correlation, specifically including: judging whether each parameter changes with time during the charging process, removing parameters that do not change with time, selecting according to whether the remaining parameters in the charging data have a positive or negative correlation with the charging electric energy, and selecting a preset number of parameters as charging process characteristics in descending order according to the correlation.

[0017] The present invention is further configured to select a preset number of parameters as charging process characteristics in descending order according to whether there is a positive or negative correlation between the remaining parameters in the charging data and the charging electric energy, and the calculation logic is: Among them, F i is the charging process feature set, X i The remaining parameters in the charging data after removing the parameters that do not change with time. Top_N is N parameters selected in descending order. Indicates parameter X i Sensitivity to the charging power change rate, δ() is a weighted function used to measure the correlation between the characteristic change rate and the charging power change rate.

[0018] The present invention is further configured to divide the charging data of the charging pile into two categories according to the absolute value of the error between the model input data set and the predicted value, record the absolute value of the charging pile error as error1, record the charging data corresponding to the charging pile whose error1 is less than or equal to the first error threshold as standard pile data, and record the charging data corresponding to the charging pile whose absolute value of error1 is greater than the first error threshold as out-of-tolerance pile data, construct a binary classification model based on the standard pile data and the out-of-tolerance pile data, perform binary classification model training, classify and judge the charging data of the new charging pile according to the trained binary classification model, and judge the corresponding charging pile category according to the classification judgment result.

[0019] The present invention is further configured such that when training a binary classification model, the loss function of the binary classification model is: Among them, Loss binary is the loss function of the binary classification model, M is the number of samples, which means the number of data in the model input data set used for binary classification model training, y i is the actual value, is the predicted value, η is the regularization parameter, which is used to control the weight of the regularization term, W is the weight matrix, which represents the connection weight of each layer in the binary classification model, ||W|| p is the p-norm regularization term of the weight matrix.

[0020] The present invention is further configured to divide the charging data of the charging pile into multiple categories according to the model input data set and the predicted value error value, record the charging pile error value as error2, record the charging data corresponding to the charging pile whose error2 is less than the second error threshold as negative excess pile data, record the charging data corresponding to the charging pile whose error2 is greater than or equal to the second error threshold and less than 0 as negative deviation standard pile data, record the charging data corresponding to the charging pile whose error2 is equal to 0 as standard pile data, record the charging data corresponding to the charging pile whose error2 is greater than 0 and less than or equal to the first error threshold as positive deviation standard pile data, and record the charging data corresponding to the charging pile whose error2 is greater than the first error threshold as positive excess pile data; construct a multi-classification model according to the negative excess pile data, negative deviation standard pile data, standard pile data, positive deviation standard pile data and positive excess pile data, perform multi-classification model training, classify and judge the charging data of the new charging pile according to the trained multi-classification model, and judge the corresponding charging pile category according to the classification judgment result.

[0021] The present invention is further configured such that when training a multi-classification model, the loss function of the multi-classification model is: Among them, Loss multi is the loss function of the multi-classification model, M is the number of samples, which means the number of data in the model input data set used for multi-classification model training, L is the number of categories, which is used to calculate the multi-category loss, and y ij is the actual value, is the predicted value, κ is the regularization parameter, which is used to control the weight of the regularization term, W l is the weight matrix of the lth layer, ∥W l ∥ q is the q-norm regularization term of the weight matrix.

[0022] The present invention provides a remote metering method for a charging pile. The method collects charging data of a charging pile according to a preset collection frequency, performs stability analysis on the charging data, screens out charging data that meets the standard, and removes abnormal charging data; samples the charging data after the stability analysis, intercepts constant current stage data in the sampled charging data, analyzes the correlation between each parameter in the constant current stage data and the charging electric energy, and sets a preset number of parameters as charging process characteristics according to the correlation; selects the interval range value of the battery SOC in the constant current stage data as the starting point and end point of the model input, performs one-hot encoding on the charging pile manufacturer and the electric meter manufacturer, splices the charging pile commissioning years with the charging process characteristics, constructs a feature sequence according to the value of the spliced ​​charging process characteristics, and sets the feature sequence as a model input data set; according to the model input data set, performs binary classification model and multi-classification model training based on a deep neural network, the binary classification model is used to judge the charging data category of the charging pile, and the multi-classification model is used to further subdivide the type of the charging data category; the charging data of the new charging pile is classified and judged according to the trained binary classification model and / or multi-classification model, and the beneficial effects generated include:

[0023] 1. Effectively utilize charging-related data to avoid a large amount of data waste: Through stability analysis and data sampling of the charging data of the charging pile, all data in the charging process can be effectively utilized. By screening out valid data that meets the standards, the waste of a large amount of redundant data can be avoided, thereby improving data utilization and the accuracy of charging pile performance evaluation.

[0024] 2. Comprehensively collect all valid data types to fully reflect the charging status: The collected data covers all valid data types involved in the charging process of the charging pile, including charging pile data, transaction data, meter data, vehicle data and charging process data, etc., which fully reflects the charging status of the charging pile under different usage conditions, and provides comprehensive data support for accurately evaluating the performance of the charging pile.

[0025] 3. Efficient data processing to improve data utilization efficiency: By sampling and feature screening the original data, the data types that may have a negative impact on the charging pile metering are removed. In the feature screening process, the correlation between each parameter and the charging power is analyzed to select the most representative feature data, thereby improving the utilization efficiency of the original data and ensuring the accuracy and reliability of the measurement.

[0026] 4. Realize remote, high-frequency and wide-coverage charging pile performance evaluation: By adopting the binary and multi-classification models of deep neural networks, the charging data of charging piles can be analyzed and judged remotely and frequently, realizing real-time monitoring and remote evaluation of charging pile performance, with a wide coverage range, not restricted by location and time, greatly improving the efficiency of charging pile performance evaluation.

[0027] 5. Reduce the cost of charging pile calibration and improve calibration efficiency: The present invention does not need to rely on traditional on-site calibration equipment and operations, reducing the labor cost and equipment cost in the calibration process. At the same time, the method of the present invention can automatically perform data analysis and classification, reducing manual participation, improving the efficiency of charging pile calibration, and ensuring the accuracy and stability of charging pile measurement.

[0028] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0030] In the attached picture:

[0031] Figure 1 The present invention is a flowchart of a remote metering method for a charging pile according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0033] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0034] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0035] A charging pile remote metering method, such as Figure 1 As shown, including:

[0036] Collect charging data of the charging pile according to a preset collection frequency, perform stability analysis on the charging data, screen out charging data that meets the standards, and eliminate abnormal charging data;

[0037] Sampling the charging data after stability analysis, intercepting constant current stage data from the sampled charging data, analyzing the correlation between various parameters in the constant current stage data and the charging electric energy, and setting a preset number of parameters as charging process characteristics according to the correlation;

[0038] The interval range value of the battery SOC in the constant current stage data is selected as the starting point and end point of the model input, the charging pile manufacturer and the meter manufacturer are one-hot encoded, and spliced ​​with the charging pile operation years to the charging process characteristics, and a feature sequence is constructed according to the spliced ​​value of the charging process characteristics, and the feature sequence is set as the model input data set;

[0039] According to the model input data set, a binary classification model and a multi-classification model are trained based on a deep neural network, respectively. The binary classification model is used to determine the category of charging data of the charging pile, and the multi-classification model is used to further subdivide the type of charging data category;

[0040] The charging data of the new charging pile is classified and judged according to the trained binary classification model and / or multi-classification model.

[0041] Specifically, in a feasible embodiment of the present invention, the sampling frequency is set to 250ms, that is, one data point is collected every 250ms, and a charging process includes multiple sampling data points, and each sampling data point includes all parameter types; the present invention is further configured that the charging data of the charging pile includes: charging pile data, transaction data, meter data, charging vehicle data, and charging process data; wherein the parameters of the charging pile data include: charging pile manufacturer, charging pile number, charging pile calibration error, and charging pile commissioning life; the parameters of the transaction data include: transaction order number, charging start time, and charging end time; the parameters of the meter data include: meter manufacturer data; the parameters of the charging vehicle data include: vehicle model, vehicle service life, and vehicle driving history; the parameters of the charging process data include: battery SOC, charging pile output voltage, charging pile output current, charging pile output power, ambient temperature, battery demand voltage, battery demand current, battery charging mode, single cell maximum voltage, single cell maximum current, battery pack maximum temperature, battery pack minimum temperature, charging power, and timestamp; specifically, the battery SOC (State SOC refers to the state of charge of the battery, expressed as a percentage. SOC is an important indicator for measuring the current remaining power of the battery, indicating the ratio between the current storage energy of the battery and its total available capacity. Furthermore, charging data is collected. In one embodiment, the charging data of the last order of the charging pile is shown in the following table:

[0042]

[0043] The present invention is further configured to perform stability analysis on the charging data, including: for each charging pile, classifying the charging data of connected charging vehicles with the same vehicle model, vehicle age and mileage belonging to the same range into one category, analyzing the variation range of the charging power of the corresponding charging vehicles under the same variation range of the battery SOC, screening the charging data according to the variation range, and screening out the charging data outside the variation range.

[0044] Specifically, the main purpose of stability analysis of charging data is to ensure the accuracy and reliability of charging data, so as to improve the accuracy of charging pile metering and the credibility of evaluation results. The present invention is further configured to analyze the variation range of charging power of the corresponding charging vehicle under the same variation range of battery SOC, and quantify the variation range of charging power by calculating the change rate of charging power of the corresponding charging vehicle when the battery SOC changes at time t. The calculation logic of the change rate of charging power is: ΔE i (t) is the rate of change of the charging power of the i-th charging vehicle, n is the total number of timestamps during the charging process, E ijFor the i-th charging vehicle at time stamp t j Charging energy, SOC ij For the i-th charging vehicle at time stamp t j The battery SOC at the time of charging, α is the adjustment index, which is used to adjust the sensitivity of the battery SOC when it changes slightly, sign() is the sign function, which is used to determine the direction of the battery SOC change, and the function result is +1 or -1; specifically, by analyzing the stability of the charging data, abnormal changes and noise data in the data can be identified and eliminated, thereby improving the metering accuracy of the charging pile. Only stable and high-quality data will be used for further analysis to ensure the accuracy of the results; the purpose of introducing the adjustment index α and the sign function sign() in the formula is to flexibly handle data changes in different situations. By adjusting the α value, the sensitivity to small SOC changes can be adaptively enhanced or weakened, and the applicability of the formula in various charging situations can be improved. The α value is usually between 1 and 2. The sensitivity is adjusted according to the actual situation. When the SOC changes slightly but its impact needs to be amplified, the value is higher (close to 2); when the SOC changes significantly, the value is lower (close to 1). No specific restrictions are made here.

[0045] The present invention is further configured to sample the charging data after stability analysis, and the sampling logic is: extract the charging data corresponding to the initial value of the battery SOC and the first charging data corresponding to each change in the battery SOC from the charging data after stability analysis; specifically, whenever the battery SOC value changes, extract the first charging data point after the change, and record the corresponding voltage, current, power and other information to reflect the state changes of the battery at different stages of the charging process; by only extracting the initial value of the battery SOC and the first data point after each change in the battery SOC, it is ensured that the extracted data can represent the key state changes in the charging process, avoiding the interference of irrelevant or redundant data on the model training, and improving the representativeness and effectiveness of the data; the form of the sampled charging data is shown in the following table:

[0046]

[0047] The present invention is further configured such that the logic of intercepting the constant current stage data is: according to the data points of the sampled charging data, extracting a voltage value of a subsequent data point that is greater than or equal to the voltage value of a previous data point, and a difference between a current value of the subsequent data point and a current value of the previous data point does not exceed a preset current threshold;

[0048] Analyze the correlation between each parameter in the constant current stage data and the charging electric energy, and set a preset number of parameters as the charging process characteristics according to the correlation, specifically including: judging whether each parameter changes with time during the charging process, removing the parameters that do not change with time, selecting according to whether there is a positive or negative correlation between the remaining parameters in the charging data and the charging electric energy, and selecting a preset number of parameters in descending order according to the correlation as the charging process characteristics. The present invention is further configured to select according to whether there is a positive or negative correlation between the remaining parameters in the charging data and the charging electric energy, and select a preset number of parameters in descending order according to the correlation as the charging process characteristics, and the calculation logic is: Among them, F i is the charging process feature set, X i The remaining parameters in the charging data after removing the parameters that do not change with time. Top_N is N parameters selected in descending order. Indicates parameter X i Sensitivity to the charging power change rate, δ() is a weighted function used to measure the correlation between the characteristic change rate and the charging power change rate; Specifically, calculate each residual parameter X i Sensitivity to the rate of change of charging energy Sensitivity measures the value of feature X when i When a small change occurs, the degree of influence on the charging power change rate is weighted using the weighting function δ() to weight the sensitivity of each feature, further emphasizing the features that are most significantly correlated with the parameter change rate and the charging power change rate, in order to balance the influence of different features.

[0049] In another feasible embodiment of the present invention, among the various parameters, the ambient temperature T and the charging power P remain unchanged and do not change with time, so there is no need to perform correlation analysis on the ambient temperature T and the charging power P. The correlation analysis is performed on the remaining parameters, and the correlation coefficient R2 table is as follows:

[0050] SOC U(t) I(t) E(t) <![CDATA[ T max ]]> <![CDATA[ T min ]]> SOC 1 0.9844 -0.7935 0.9999 0.9839 0.9769 U(t) 0.9844 1 -0.6791 0.9818 0.9631 0.9609 I(t) -0.7935 -0.6791 1 -0.8016 -0.8133 -0.7924 ... ... ... ... ... ... ... E(t) 0.9999 0.9818 -0.8016 1 0.9844 0.9767 <![CDATA[ T max ]]> 0.9839 0.9631 -0.8133 0.9844 1 0.9888 <![CDATA[ T min ]]> 0.9769 0.9609 -0.7924 0.9767 0.9888 1

[0051] According to the selection principle of the starting point and the end point, the starting point is selected as the characteristic data point corresponding to the battery SOC = 30, and the end point is selected as the characteristic data point corresponding to the battery SOC = 82. According to the selected starting point and end point, the characteristic values ​​in the charging process data are taken as the sequence of each variable. Such as U(t1)~U(t5), SOC(t1)~SOC(t5),..........., E(t1)~E(t5), as shown in the following table:

[0052] Data No. Article 1 Data <![CDATA[30,31,32.33.34.35,U1,U2,U3,U4,U5,......,E1,E2,E3,E4,E5]]> Article 2 Data <![CDATA[31,32,33,34,35,36,U2,U3,U4,U5,U6,......,E2,E3,E4,E5,E6]]> ...... ...... Article 48 Data <![CDATA[77,78,79,80,81,U 48 ,U 49 ,U 50 ,U 51 ,U 52， ......,AND 48 ,AND 49 ,AND 50 ,AND 51 ,AND 52 ]]> Article 49 Data <![CDATA[78,79,80,81,82,U 49 ,U 50 ,U 51 ,U 52 ,U 53 ,......,AND 49 ,AND 50 ,AND 51 ,AND 52 ,AND 53 ]]>

[0053] The charging pile manufacturer and the meter manufacturer are one-hot coded; specifically, the one-hot coding result of the charging pile manufacturer is as follows:

[0054] Charging pile manufacturer One-hot encoded features (the last digit is reserved) Charging pile manufacturer 1 1000 Charging pile manufacturer 2 0100 Charging pile manufacturer 3 0010

[0055] The one-hot encoding results of the meter manufacturer are as follows:

[0056] Electric meter manufacturers One-hot encoded features (the last digit is reserved) Electric meter manufacturer 1 1000 Electric meter manufacturer 2 0100 Electric meter manufacturer 3 0010

[0057] The one-hot encoded charging pile manufacturer, electric meter manufacturer and charging pile operation years are spliced ​​to the charging process feature, and a feature sequence is constructed according to the spliced ​​values ​​of the charging process feature.

[0058] The present invention is further configured to divide the charging data of the charging pile into two categories according to the absolute value of the error between the model input data set and the predicted value, record the absolute value of the charging pile error as error1, record the charging data corresponding to the charging pile whose error1 is less than or equal to the first error threshold as standard pile data, and record the charging data corresponding to the charging pile whose absolute value of error1 is greater than the first error threshold as out-of-tolerance pile data, construct a binary classification model based on the standard pile data and the out-of-tolerance pile data, perform binary classification model training, classify and judge the charging data of the new charging pile according to the trained binary classification model, and judge the corresponding charging pile category according to the classification judgment result.

[0059] Specifically, the corresponding labels are assigned according to the errors of the corresponding charging piles, and then sent to the binary classification model for model training. The trained model is then used to judge the new unlabeled data and identify its data type.

[0060] According to the identification results of multiple data on a charging pile, the type of charging pile is judged according to the charging pile type judgment method.

[0061] For example, there are 2000 data in the feature data set of pile A, 1500 of which are judged as out-of-tolerance pile data and 500 as standard pile data. Then pile A is an out-of-tolerance pile, and the corresponding probability is:

[0062] The present invention is further configured such that when training a binary classification model, the loss function of the binary classification model is: Among them, Loss binary is the loss function of the binary classification model, M is the number of samples, which means the number of data in the model input data set used for binary classification model training, y i is the actual value, is the predicted value, η is the regularization parameter, which is used to control the weight of the regularization term, W is the weight matrix, which represents the connection weight of each layer in the binary classification model, ||W|| p is the p-norm regularization term of the weight matrix; specifically, the loss function is the core part of machine learning model training, which is used to measure the difference between the model's predicted results and the actual results. By minimizing the value of the loss function, the model parameters can be optimized, so that the model's predictive ability is continuously improved. The cross entropy error between the predicted value and the actual value of each sample is calculated. Cross entropy is a commonly used loss function, especially for classification tasks. It measures the difference between the actual category and the predicted probability distribution. η·||W|| p The regularization term prevents the model from being overly complicated by controlling the size of the weight matrix. Smaller weights can make the model simpler and more generalizable. Common regularization methods include L1 regularization (Lasso) and L2 regularization (Ridge); the regularization parameter η is selected to a smaller value through cross-validation or experience to balance the error term and regularization term in the loss function, and the value is between 0 and 1 (including 0.001, 0.01, 0.1), and there is no restriction here.

[0063] The present invention is further configured to divide the charging data of the charging pile into multiple categories according to the model input data set and the predicted value error value, record the charging pile error value as error2, record the charging data corresponding to the charging pile whose error2 is less than the second error threshold as negative excess pile data, record the charging data corresponding to the charging pile whose error2 is greater than or equal to the second error threshold and less than 0 as negative deviation standard pile data, record the charging data corresponding to the charging pile whose error2 is equal to 0 as standard pile data, record the charging data corresponding to the charging pile whose error2 is greater than 0 and less than or equal to the first error threshold as positive deviation standard pile data, and record the charging data corresponding to the charging pile whose error2 is greater than the first error threshold as positive excess pile data; construct a multi-classification model according to the negative excess pile data, negative deviation standard pile data, standard pile data, positive deviation standard pile data and positive excess pile data, perform multi-classification model training, classify and judge the charging data of the new charging pile according to the trained multi-classification model, and judge the corresponding charging pile category according to the classification judgment result.

[0064] Specifically, the corresponding labels are assigned according to the errors of the corresponding charging piles, and then sent to the multi-classification model for model training. The trained model is then used to judge the new unlabeled data and identify its data type.

[0065] According to the identification results of multiple data on a charging pile, the type of charging pile is judged according to the charging pile type judgment method.

[0066] For example, there are 10,000 data in the feature data set of pile B. After the multi-classification model is used to judge, there are 8,000 positive and negative pile data and 2,000 other types of data. Then pile B is a positive and negative pile, and the corresponding probability is:

[0067] The present invention is further configured such that when training a multi-classification model, the loss function of the multi-classification model is: Among them, Loss multi is the loss function of the multi-classification model, M is the number of samples, which means the number of data in the model input data set used for multi-classification model training, L is the number of categories, which is used to calculate the multi-category loss, and y ij is the actual value, is the predicted value, κ is the regularization parameter, which is used to control the weight of the regularization term, W l is the weight matrix of the lth layer, ∥W l ∥ q is the q-norm regularization term of the weight matrix; specifically, the loss function of the multi-classification model is used to measure the prediction accuracy of the model in the multi-class classification task. By minimizing the loss function, the model parameters can be optimized so that the model can show high classification accuracy in multiple categories; The cross entropy error of each sample on all categories is calculated. The regularization term prevents the model from being overly complicated by controlling the size of the weight matrix. Through the loss function of the multi-classification model, it can effectively guide the multi-classification model to continuously optimize the weight parameters during the training process, thereby improving the performance and stability of the multi-classification model.

[0068] Furthermore, the method also includes: using a binary classification model to perform binary classification discrimination on data in a short-term range; using a binary classification and a multi-classification model for fusion in a long-term range, and specifically, the specific method of model fusion is: according to the data category classification table in the multi-classification model, when the prediction type in the binary classification model is consistent with the prediction type of the multi-classification model, the final result is given according to the multi-classification data category, and the corresponding probability is the product of the binary classification prediction probability and the multi-classification prediction result. When the prediction type in the binary classification model is inconsistent with the prediction type of the multi-classification model, the charging data is not judged.

[0069] According to the model result fusion method, the binary classification and multi-classification model results are fused. The fusion forms are divided into three cases:

[0070] Case 1: If the number of judgment data is less than the preset number, only the binary classification model is used for judgment. In a feasible embodiment of the present invention, the preset number of data is 2000, which is not limited here and can be set according to needs.

[0071] Case 2: If the number of data is greater than or equal to the preset number, and if the type of charging pile determined by the binary classification model is inconsistent with the multi-classification model, the charging pile is listed as a pending pile, and the specific type of the charging pile is not determined.

[0072] Case 3: If the number of data is greater than or equal to the preset number, and if the charging pile type determined by the binary classification model is consistent with the multi-classification model, the specific charging pile category is given according to the multi-classification model, and the corresponding probability is: P 融合 =P 二分类 *P 多分类 Furthermore, the corresponding relationship between the charging pile type determined by the binary classification model and the multi-classification model is shown in the following table:

[0073]

[0074] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0075] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0076] In this application, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0077] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0078] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0079] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0080] In the several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0081] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0082] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0083] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0084] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A charging pile remote metering method, characterized in that: include: Collect charging data of the charging pile according to a preset collection frequency, perform stability analysis on the charging data, screen out charging data that meets the standards, and eliminate abnormal charging data; Sampling the charging data after stability analysis, intercepting constant current stage data from the sampled charging data, analyzing the correlation between various parameters in the constant current stage data and the charging electric energy, and setting a preset number of parameters as charging process characteristics according to the correlation; The interval range value of the battery SOC in the constant current stage data is selected as the starting point and end point of the model input, the charging pile manufacturer and the meter manufacturer are one-hot encoded, and spliced ​​with the charging pile operation years to the charging process characteristics, and a feature sequence is constructed according to the spliced ​​value of the charging process characteristics, and the feature sequence is set as the model input data set; According to the model input data set, a binary classification model and a multi-classification model are trained based on a deep neural network, respectively. The binary classification model is used to determine the category of charging data of the charging pile, and the multi-classification model is used to further subdivide the type of charging data category; The charging data of the new charging pile is classified and judged according to the trained binary classification model and / or multi-classification model.

2. A charging pile remote metering method according to claim 1, characterized in that: The charging data of the charging pile include: charging pile data, transaction data, electric meter data, charging vehicle data, and charging process data; among them, the parameters of the charging pile data include: charging pile manufacturer, charging pile number, charging pile calibration error, and charging pile commissioning life; the parameters of the transaction data include: transaction order number, charging start time, and charging end time; the parameters of the electric meter data include: electric meter manufacturer data; the parameters of the charging vehicle data include: vehicle model, vehicle age, and vehicle driving history; the parameters of the charging process data include: battery SOC, charging pile output voltage, charging pile output current, charging pile output power, ambient temperature, battery demand voltage, battery demand current, battery charging mode, single cell maximum voltage, single cell maximum current, battery pack maximum temperature, battery pack minimum temperature, charging power, and timestamp.

3. A charging pile remote metering method according to claim 2, characterized in that: The charging data is subjected to stability analysis, including: for each charging pile, the charging data of connected charging vehicles with the same vehicle model, vehicle age and mileage belonging to the same range are classified into one category, and the variation range of the charging power of the corresponding charging vehicles under the condition that the battery SOC has the same variation range is analyzed, and the charging data is screened according to the variation range to filter out the charging data outside the variation range.

4. A charging pile remote metering method according to claim 3, characterized in that: Analyze the variation range of the charging power of the corresponding charging vehicle under the same variation range of the battery SOC. Quantify the variation range of the charging power by calculating the change rate of the charging power of the corresponding charging vehicle when the battery SOC changes at time t. The calculation logic of the change rate of the charging power is: ΔE i (t) is the rate of change of the charging power of the i-th charging vehicle, n is the total number of timestamps during the charging process, E ij For the i-th charging vehicle at time stamp t j Charging energy, SOC ij For the i-th charging vehicle at time stamp t j The battery SOC at that time, α is the adjustment index, which is used to adjust the sensitivity of the battery SOC to a small change, sign() is the sign function, which is used to determine the direction of the battery SOC change, and the function result is +1 or -1.

5. A charging pile remote metering method according to claim 2, characterized in that: The charging data after stability analysis is sampled, and the sampling logic is: extracting the charging data corresponding to the initial value of the battery SOC and the first charging data corresponding to each change in the battery SOC from the charging data after stability analysis; The logic of intercepting the constant current stage data is: according to the data points of the sampled charging data, extract the voltage value of the next data point greater than or equal to the voltage value of the previous data point, and the difference between the current value of the next data point and the current value of the previous data point does not exceed the preset current threshold; Analyze the correlation between each parameter in the constant current stage data and the charging electric energy, and set a preset number of parameters as charging process characteristics according to the correlation, specifically including: judging whether each parameter changes with time during the charging process, removing parameters that do not change with time, selecting according to whether the remaining parameters in the charging data have a positive or negative correlation with the charging electric energy, and selecting a preset number of parameters as charging process characteristics in descending order according to the correlation.

6. A charging pile remote metering method according to claim 5, characterized in that: The selection is made based on whether there is a positive or negative correlation between the remaining parameters in the charging data and the charging electric energy, and a preset number of parameters are selected in descending order according to the correlation as the charging process characteristics. The calculation logic is: Among them, F i is the charging process feature set, X i The remaining parameters in the charging data after removing the parameters that do not change with time. Top_N is N parameters selected in descending order. Indicates parameter X i Sensitivity to the charging power change rate, δ() is a weighted function used to measure the correlation between the characteristic change rate and the charging power change rate.

7. A charging pile remote metering method according to claim 1, characterized in that: The charging data of the charging pile are divided into two categories according to the absolute value of the error between the model input data set and the predicted value. The absolute value of the charging pile error is recorded as error1, and the charging data corresponding to the charging piles whose error1 is less than or equal to the first error threshold is recorded as standard pile data. The charging data corresponding to the charging piles whose absolute value of error1 is greater than the first error threshold is recorded as out-of-tolerance pile data. Based on the standard pile data and the out-of-tolerance pile data, a binary classification model is constructed and trained. The charging data of the new charging pile is classified and judged according to the trained binary classification model, and the corresponding charging pile category is judged according to the classification judgment result.

8. A charging pile remote metering method according to claim 7, characterized in that: When training a binary classification model, the loss function of the binary classification model is: Among them, Loss binary is the loss function of the binary classification model, M is the number of samples, which means the number of data in the model input data set used for binary classification model training, y i is the actual value, is the predicted value, η is the regularization parameter, which is used to control the weight of the regularization term, and W is the weight matrix, which represents the connection weight of each layer in the binary classification model, ∥W∥ p is the p-norm regularization term of the weight matrix.

9. A charging pile remote metering method according to claim 7, characterized in that: The charging data of the charging pile is divided into multiple categories according to the model input data set and the error value of the predicted value, and the error value of the charging pile is recorded as error2. The charging data corresponding to the charging pile whose error2 is less than the second error threshold is recorded as negative excess pile data, the charging data corresponding to the charging pile whose error2 is greater than or equal to the second error threshold and less than 0 is recorded as negative deviation standard pile data, the charging data corresponding to the charging pile whose error2 is equal to 0 is recorded as standard pile data, the charging data corresponding to the charging pile whose error2 is greater than 0 and less than or equal to the first error threshold is recorded as positive deviation standard pile data, and the charging data corresponding to the charging pile whose error2 is greater than the first error threshold is recorded as positive excess pile data. According to the negative excess pile data, negative deviation standard pile data, standard pile data, positive deviation standard pile data and positive excess pile data, a multi-classification model is constructed, and the multi-classification model is trained. According to the trained multi-classification model, the charging data of the new charging pile is classified and judged, and the corresponding charging pile category is judged according to the classification judgment result.

10. A charging pile remote metering method according to claim 9, characterized in that: When training a multi-classification model, the loss function of the multi-classification model is: Among them, Loss multi is the loss function of the multi-classification model, M is the number of samples, which means the number of data in the model input data set used for multi-classification model training, L is the number of categories, which is used to calculate the multi-category loss, and y ij is the actual value, is the predicted value, κ is the regularization parameter, which is used to control the weight of the regularization term, W l is the weight matrix of the lth layer, ∥W l ∥ q is the q-norm regularization term of the weight matrix.

Images