CN-122017719-A - Charging pile metering error detection method and device based on KDE-EnKF

CN122017719ACN 122017719 ACN122017719 ACN 122017719ACN-122017719-A

Abstract

The invention provides a charging pile metering error detection method and device based on KDE-EnKF, the method comprises the steps of S101 fitting a probability distribution function of an electric energy loss item in a charging station by using a nuclear density estimation method, S102 initializing a set Kalman filter model, S103 continuously correcting a charging pile error coefficient at a plurality of moments in a continuous time period through the set Kalman filter model, wherein the continuous correction is that total output electric energy and sub-meter output electric energy at each moment are used as input of the set Kalman filter model, and the set Kalman filter model is used for correcting the charging pile error coefficient at the last moment and using the corrected value as output of the filter model in combination with the probability distribution function of the electric energy loss item. According to the invention, the probability characteristics are integrated into the state estimation process based on the energy conservation model, so that the error coefficient of the charging pile can be continuously and accurately estimated, and the verification efficiency of the charging pile is remarkably improved.

Inventors

ZHANG HUANGHUI
ZHENG DI
DONG XIAOLONG
JIANG KAI
LIU HONGHUI
DONG QIQI
CHEN QIRONG
FANG JIE

Assignees

福建省计量科学研究院（福建省眼镜质量检验站）
中国计量大学

Dates

Publication Date: 20260512
Application Date: 20251210

Claims (8)

1. The invention provides a charging pile metering error detection method and device based on KDE-EnKF, the method comprises the steps of S101 fitting a probability distribution function of an electric energy loss item in a charging station by using a nuclear density estimation method, S102 initializing a set Kalman filter model, S103 continuously correcting a charging pile error coefficient at a plurality of moments in a continuous time period through the set Kalman filter model, wherein the continuous correction is that total output electric energy and sub-meter output electric energy at each moment are used as input of the set Kalman filter model, and the set Kalman filter model is used for correcting the charging pile error coefficient at the last moment and using the corrected value as output of the filter model in combination with the probability distribution function of the electric energy loss item. According to the invention, the probability characteristics are integrated into the state estimation process based on the energy conservation model, so that the error coefficient of the charging pile can be continuously and accurately estimated, and the verification efficiency of the charging pile is remarkably improved.
2. The method according to claim 1, wherein the step S101 comprises the steps of: S11, for a charging station with N direct current charging piles, based on an energy conservation law, an energy conservation equation of total output electric energy, output electric energy of each sub-meter and electric energy loss terms is established, wherein the equation is as follows: ; In the formula, Outputting electric energy for a specific period of time for the charging station summary table; outputting electric energy in the specific time period for the ith charging pile sub-meter; S is an electric energy loss item in the same period of time in the charging station; The total table of the charging stations and the output electric energy of each charging pile sub-table in the specific time period are calculated by the difference between the accumulated readings of the corresponding electric energy meter at the beginning of the specific time period and the end of the specific time period; Secondly, according to the prior data sample, calculating the electric energy loss items in the charging stations at t different moments through the established energy conservation equation And calculate the power loss term average And standard deviation of sample The calculation formulas are respectively as follows: ; ; ; Wherein n is the sample size, For the total output power of the charging station at time k, The output electric energy of the ith charging pile sub-meter at the k moment; s12, calculating the bandwidth h, wherein the calculation formula is as follows: ; s13, defining a kernel function The formula is as follows: ; Wherein u is a normalized distance variable of an estimated point s and an electric energy loss term; s14, for each estimated point S and each power loss term The normalized distance variable u is calculated as follows: ; s15, calculating probability density function of electric energy loss item in charging station The calculation formula is as follows: 。
3. The method according to claim 2, wherein the step S102 comprises the steps of: s21, for the N direct current charging piles, determining initial error coefficients according to verification results or charging data Constitute the initial state variable mean value ; S22, determining covariance matrix of state variable according to priori knowledge The mean value is Variance is Is a normal distribution of (2); s23, sampling M variables from the normal distribution Forming initial state variable set of error coefficient of charging pile The number M of the variables in the set takes larger values of 2*n and 100, and if N is larger, the value of M is correspondingly increased to be 100 to 1000; Wherein the method comprises the steps of And the error coefficient of the nth charging pile in the ith initial state variable.
4. The method according to claim 3, wherein the step S103 comprises the steps of: s31, under the set Kalman filter model, calculating a prior estimation value of a state variable set of the error coefficient of the charging pile at the current moment, wherein the calculation formula is as follows: ; In the formula, For the set of state variable estimates at time k, For the i-th state variable a priori estimate at time k, Error coefficients for the nth charge pile therein; A state variable correction value set at the moment k-1; a process noise set at the moment k; Obeying the N (0, Q) distribution, Q is A covariance matrix; The system performs a first iteration when k=1, based on the initial state variable set of the charging pile error coefficient For the state variable at the first moment The set is estimated a priori, and each iteration thereafter is based on the set of state variable correction values at the previous time ; S32, distributing probability density functions according to electric energy loss Sampling to form an estimated set of power losses at the current time ; The sampling method is aimed at state variables Each element in the set firstly selects a data point pi from the original data set with uniform probability, secondly extracts a random offset from a kernel probability distribution taking the data point pi as a center and taking the bandwidth h as a scale parameter, and finally, the estimated value of the electric energy loss at the current moment is the sum of pi and the extracted random offset; s33, calculating an estimated value of total output electric energy of the charging station at the current moment based on the energy conservation equation established in the S11 The calculation formula is as follows: ; In the formula, The output electric energy value obtained for the nth charging pile sub-meter at the k moment; wherein Is a collection Elements of (a) And matrix The matrix is formed by inverting each element of the matrix obtained by addition, A full one matrix of 1 ́ N; Outputting an electric energy estimated value set for the total table at the moment k; outputting an electric energy estimated value for the ith summary table at the k moment; A measurement noise set at the moment k; Obeying N (0, R) distribution, wherein R is covariance matrix; S34, calculating the gain of the Kalman filter of the current moment set The calculation formula is as follows: ; In the formula, ; S35, calculating a current charging pile error coefficient state variable correction value set And an estimated value The calculation formulas are respectively as follows: ; ; In the formula, The output electric energy value obtained from the charging station summary table at the moment k; Is that Is a full matrix of (a); , the final estimated value of the error coefficient of the nth charging pile at the k moment; For a set of state variable correction values at time k I-th element of (a) in the list.
5. The charging pile metering error detection device based on KDE-EnKF is characterized by comprising: the fitting module is used for fitting a probability distribution function of an electric energy loss item in the charging station by using a nuclear density estimation method; the initialization module is used for initializing the set Kalman filter model; The correction module is used for continuously correcting the charging pile error coefficient at a plurality of moments in a continuous time period through the set Kalman filter model, wherein the continuous correction is that total output electric energy and sub-meter output electric energy at each moment are used as input of the set Kalman filter model, the probability distribution function of an electric energy loss term is combined, the set Kalman filter model is used for correcting the charging pile error coefficient at the last moment, and the correction value is used as output of the filter model.
6. The apparatus of claim 5, wherein the fitting module is configured to perform the following: S11, for a charging station with N direct current charging piles, based on an energy conservation law, an energy conservation equation of total output electric energy, output electric energy of each sub-meter and electric energy loss terms is established, wherein the equation is as follows: ; In the formula, Outputting electric energy for a specific period of time for the charging station summary table; outputting electric energy in the specific time period for the ith charging pile sub-meter; S is an electric energy loss item in the same period of time in the charging station; The total table of the charging stations and the output electric energy of each charging pile sub-table in the specific time period are calculated by the difference between the accumulated readings of the corresponding electric energy meter at the beginning of the specific time period and the end of the specific time period; Secondly, according to the prior data sample, calculating the electric energy loss items in the charging stations at t different moments through the established energy conservation equation And calculate the power loss term average And standard deviation of sample The calculation formulas are respectively as follows: ; ; ; Wherein n is the sample size, For the total output power of the charging station at time k, The output electric energy of the ith charging pile sub-meter at the k moment; s12, calculating the bandwidth h, wherein the calculation formula is as follows: ; s13, defining a kernel function The formula is as follows: ; Wherein u is a normalized distance variable of an estimated point s and an electric energy loss term; s14, for each estimated point S and each power loss term The normalized distance variable u is calculated as follows: ; s15, calculating probability density function of electric energy loss item in charging station The calculation formula is as follows: 。
7. the apparatus of claim 6, wherein the initialization module is configured to perform the following: s21, for the N direct current charging piles, determining initial error coefficients according to verification results or charging data Constitute the initial state variable mean value ; S22, determining covariance matrix of state variable according to priori knowledge The mean value is Variance is Is a normal distribution of (2); s23, sampling M variables from the normal distribution Forming initial state variable set of error coefficient of charging pile The number M of the variables in the set takes larger values of 2*n and 100, and if N is larger, the value of M is correspondingly increased to be 100 to 1000; Wherein the method comprises the steps of And the error coefficient of the nth charging pile in the ith initial state variable.
8. The apparatus of claim 7, wherein the correction module is configured to perform the following: s31, under the set Kalman filter model, calculating a prior estimation value of a state variable set of the error coefficient of the charging pile at the current moment, wherein the calculation formula is as follows: ; In the formula, For the set of state variable estimates at time k, For the i-th state variable a priori estimate at time k, Error coefficients for the nth charge pile therein; A state variable correction value set at the moment k-1; a process noise set at the moment k; Obeying the N (0, Q) distribution, Q is A covariance matrix; The system performs a first iteration when k=1, based on the initial state variable set of the charging pile error coefficient For the state variable at the first moment The set is estimated a priori, and each iteration thereafter is based on the set of state variable correction values at the previous time ; S32, distributing probability density functions according to electric energy loss Sampling to form an estimated set of power losses at the current time ; The sampling method is aimed at state variables Each element in the set firstly selects a data point pi from the original data set with uniform probability, secondly extracts a random offset from a kernel probability distribution taking the data point pi as a center and taking the bandwidth h as a scale parameter, and finally, the estimated value of the electric energy loss at the current moment is the sum of pi and the extracted random offset; s33, calculating an estimated value of total output electric energy of the charging station at the current moment based on the energy conservation equation established in the S11 The calculation formula is as follows: ; In the formula, The output electric energy value obtained for the nth charging pile sub-meter at the k moment; wherein Is a collection Elements of (a) And matrix The matrix is formed by inverting each element of the matrix obtained by addition, Is that Is a full matrix of (a); Outputting an electric energy estimated value set for the total table at the moment k; outputting an electric energy estimated value for the ith summary table at the k moment; A measurement noise set at the moment k; Obeying N (0, R) distribution, wherein R is covariance matrix; S34, calculating the gain of the Kalman filter of the current moment set The calculation formula is as follows: ; In the formula, ; S35, calculating a current charging pile error coefficient state variable correction value set And an estimated value The calculation formulas are respectively as follows: ; ; In the formula, The output electric energy value obtained from the charging station summary table at the moment k; Is that Is a full matrix of (a); , the final estimated value of the error coefficient of the nth charging pile at the k moment; For a set of state variable correction values at time k I-th element of (a) in the list.

Description

Charging pile metering error detection method and device based on KDE-EnKF Technical Field The invention relates to the technical field of electric automobile charging pile verification, in particular to a charging pile metering error detection method and device. Background The charging pile is used as a key infrastructure for supplying energy to the electric automobile, and is widely applied along with the rapid expansion of the new energy automobile market in recent years. According to different charging technologies and power levels, the charging piles can be mainly divided into two types, namely alternating current charging piles and direct current charging piles. The AC charging pile has relatively simple structure, but the limited charging power results in longer charging time, and the DC charging pile has high power output capability, so that the charging time can be obviously shortened, and the equipment and the installation cost are correspondingly higher. With the continuous development of the electric automobile industry and continuous breakthrough of the charging technology, the coverage density and the intelligent level of the charging infrastructure are gradually improved, the large-scale application of the electric automobile is expected to be further promoted, and important support is provided for constructing a clean low-carbon energy system. However, in the charging process, various internal or external factors affect the accuracy of electric energy metering of the charging pile, and the benefits of the user and the service provider are directly related. Because the charging pile relates to trade settlement of electric energy, the charging pile belongs to a forced verification working metering device according to the regulations in the metering method of the people's republic of China. Traditional field manual identification methods are high in cost and low in efficiency, and cannot meet the requirements of a large number of existing verification. Most of the existing researches utilize big data to detect the charging piles, and the existing researches can be divided into two methods of vehicle pile interaction and energy conservation according to different principles. The modeling requirements of the energy conservation model on the electric energy loss are high, the accuracy of the model is a key factor influencing the final result, most of the existing models are based on simple assumptions or specific distribution characteristics, and the actual distribution situation of the electric energy loss is difficult to accurately embody. For example: The invention in China disclosed in 20250506 and with publication number CN119936723A discloses a method, a device, equipment and a medium for analyzing the error characteristics of a charging pile group, which are used for constructing an initial state variable set, calculating a state variable estimated value set of the error coefficient of the charging pile group at the moment k, calculating an output variable estimated value set at the moment k, calculating a state variable corrected value set at the moment k and covariance thereof, and outputting a state variable estimated value of the error coefficient vector of the charging pile group corrected at the moment k and uncertainty thereof. The invention only provides a method for solving error coefficients by a set Kalman filter, does not process electric energy loss items (such as line loss, equipment heating loss and the like), causes incomplete energy conservation models, cannot interpret the difference value between total output electric energy and branch meter output electric energy through the loss items, directly attributes the difference value to the error coefficients, can lead the error coefficient estimation result to contain the interference of the loss items, leads to the increase of estimation deviation, lacks a loss item dynamic correction mechanism, is difficult to adapt to complex working conditions, cannot distinguish 'real error coefficient change' and 'loss item change', and leads to the poor robustness of the complex working conditions. Ultimately affecting the accuracy and stability of the error coefficient estimation. The invention of China, publication No. CN113346579A, published in 20210903 discloses a method for monitoring the operation error of metering equipment in a DC charging station. The method comprises the following steps of establishing equations about power supply electric quantity of an alternating current side, electric quantity of each charging gun, inherent loss of a charging station, energy conversion efficiency of an AC-DC conversion module of each charging pile, equivalent error parameters and heat loss of each charging gun, obtaining electric quantity data of metering points of each charging gun of the alternating current side and the direct current side in a plurality of metering periods and energy conversion efficiency of each charging pile, substitu