Electric vehicle charging detection method

Abstract

The invention discloses an electric vehicle charging detection method, which comprises the following steps: taking the change of an active power transient waveform in a socket as a transient event, and monitoring the transient event based on a time sliding window; when the transient event is ended, calculating the variable quantity of the active power before and after the transient event; comparing the variable quantity of the active power with an active power threshold value of the electric vehicle to determine that the electric vehicle is inserted into the socket, and acquiring an active power sampling sequence in the socket in the time sliding window; processing the active power sampling sequence based on an electric vehicle charging detection model, and detecting the charging state of the electric vehicle; the detection accuracy of the charging state of the electric bicycle is improved, and the charging misjudgment event of the electric bicycle is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicles, and in particular to a charging detection method for an electric vehicle. Background Art

[0002] With the rapid development of the electric vehicle industry, the problem of electric vehicle charging has become increasingly prominent. Accurately detecting whether electric vehicle charging is connected to the distribution network is of great significance for smart grid management, energy scheduling, and user electricity consumption behavior analysis.

[0003] In the related art, a charging safety warning method and system based on convolutional neural network is proposed in the patent application document with application publication number CN115453370A. In this scheme, convolutional neural network is used to predict the battery status, and safety detection and danger warning are performed according to the predicted battery status. The purpose of this scheme is to predict the battery status, so the charging data includes battery type, charging current data, charging voltage data, battery temperature data and battery SOC data, and the charging data mainly focuses on the state change during the battery charging process. In the patent application with application publication number CN117471157A, a non-intrusive electric bicycle load charging event detection and extraction method is disclosed. This scheme uses the mutation of the effective value of the current to judge the access of the electric bicycle. However, the access or disconnection of other loads may also cause the same mutation, so that the load access of non-electric bicycles is misjudged as an electric bicycle charging event, resulting in false detection. Although the sliding window technology is used in the document "Improved algorithm for sliding window bilateral CUSUM event detection based on small current electrical appliances, Huang Wei, Hydropower Energy Science", the purpose is to identify small current electrical appliances. Summary of the invention

[0004] The technical problem to be solved by the present invention is how to accurately determine the charging status of an electric vehicle.

[0005] The present invention solves the above technical problems by the following technical means:

[0006] A method for detecting charging of an electric vehicle is proposed, the method comprising:

[0007] Taking the change of the transient waveform of active power in the socket as a transient event, and monitoring the transient event based on a time sliding window;

[0008] When the transient event ends, calculating the change in active power before and after the transient event occurs;

[0009] When comparing the change in active power with the electric vehicle active power threshold to determine that an electric vehicle is plugged into the socket, an active power sampling sequence in the socket within the time sliding window is collected;

[0010] The active power sampling sequence is processed based on the electric vehicle charging detection model to detect whether the electric vehicle is in a charging state.

[0011] Further, the method of taking the active power transient waveform in the socket as a transient event and monitoring the transient event based on a time sliding window includes:

[0012] The transient event is monitored using the time sliding window N of length 2D, wherein the time sliding window includes time sub-windows N1 and N2, where N1 = [P1, ..., P D ],N2=[P D+1 ,…,P 2D ], P D represents the active power at the Dth sampling point in the time sliding window N;

[0013] Based on the active power in the time sliding window N, calculate the variance F1 of the active power sampled in the time subwindow N1 and the variance F2' of the active power in the time subwindow N2, wherein the variance F2' is calculated based on the average value of the active power sampled in the time subwindow N1;

[0014] When F2'>F1β FF When , it is determined that a transient event has occurred, where β FF is the threshold of the active power variance multiple of time sub-window N1 and time sub-window N2.

[0015] Further, when the transient event ends, calculating the change in active power before and after the transient event occurs includes:

[0016] Based on the active power in the time sliding window N, calculate the variance F2 of the active power sampled in the time subwindow N2 and the variance F1' of the active power in the time subwindow N1, wherein the variance F1' is calculated based on the average value of the active power sampled in the time subwindow N2;

[0017] In meeting F1' <F2β FF When , it is determined that the transient event ends;

[0018] The change in active power before and after the transient event is calculated based on the active power at the starting point and the active power at the end point of the transient event.

[0019] Furthermore, the calculation formula of the variance F2' of the active power in the time sub-window N2 is:

[0020]

[0021] The calculation formula of the variance F1' of the active power in the time sub-window N1 is:

[0022]

[0023] In the formula, Represents the average value of the active power sampled in the time sub-window N1.

[0024] Furthermore, the method further comprises:

[0025] From the first sampling point of the time sliding window N, all sampling points are traversed to the right, and the first traversal to the point that satisfies The sampling point is the starting point of the transient event;

[0026] From the 2Dth sampling point of the time sliding window N, traverse all sampling points to the left, and traverse to the first point that satisfies The sampling point is the end point of the transient event;

[0027] in, represents the average value of active power sampled in time sub-window N1, represents the average value of active power sampled in time sub-window N2, P n Represents the active power at the nth sampling point.

[0028] Furthermore, the calculation formula for the change in active power before and after the transient event occurs is:

[0029]

[0030] Where ΔP represents the change in active power before and after the transient event occurs, P start Represents the active power at the starting point of the transient event, P over Represents the active power at the end point of the transient event, P over+n represents the active power at n sampling points after the end of the transient event, P start-n Represents the active power of n sampling points before the starting point of the transient event.

[0031] Further, when comparing the change in active power with the electric vehicle active power threshold to determine that an electric vehicle is plugged into the socket, collecting an active power sampling sequence in the socket within the time sliding window includes:

[0032] When the change in active power is greater than the electric vehicle active power threshold, it is preliminarily determined that an electric vehicle is plugged into the socket;

[0033] The active power of the socket at each sampling time in the time sliding window is collected, each active power sampling value is traversed and preprocessed, and a processed active power sampling sequence is obtained.

[0034] Furthermore, the active power of the socket at each sampling time in the time sliding window is collected, each active power sampling value is traversed and preprocessed to obtain a processed active power sampling sequence, including:

[0035] Collect the active power [P1,…P D ,…,P D+l …,P 2D ];

[0036] Determine the active power P of the D+lth sampling point D+l Whether it is an outlier or missing value;

[0037] If so, then the active power P D+l is an abnormal value, then the active power P D+l Corrected to:

[0038]

[0039] Where P represents P1 to P 2D The active power of all sampling points. Average means taking the active power from P1 to P 2D The average value of Std is taken from P1 to P 2D The standard deviation of

[0040] In the active power P D+l If it is a missing value, then for the active power P D+l Corrected to:

[0041]

[0042] In the formula, None means that the sampling point is a set of missing values;

[0043] If not, assign l+1 to l and re-judge the active power P of the D+lth sampling point. D+l Whether it is an abnormal value or a missing value, until l>D, the processed active power sampling sequence is obtained.

[0044] Furthermore, the electric vehicle charging detection model adopts a GRU neural network model, which includes three GRU layers, each GRU layer includes K GRU blocks, and each GRU block includes a reset gate R t and an update gate Z t ;

[0045] The output of the third GRU layer is input into the fully connected layer, which is followed by a sigmoid activation layer. The sigmoid activation layer outputs a two-dimensional state vector, which respectively indicates that the electric vehicle is charging and the non-electric vehicle is charging.

[0046] Furthermore, before processing the active power sampling sequence based on the electric vehicle charging detection model to detect the charging state of the electric vehicle, the method further includes:

[0047] Based on the constructed data set, the GRU neural network model is trained by back propagation and gradient descent algorithms, and the cross entropy loss function L is calculated to update the network parameters until the maximum number of iterations is reached or the loss function value converges, thereby obtaining the electric vehicle charging detection model.

[0048] The advantages of the present invention are:

[0049] (1) The present invention first determines whether a transient event occurs through a time sliding window, then determines whether an electric vehicle is plugged into a socket through a power threshold, and then inputs the sampling points of the active power waveform into an electric vehicle charging detection model to further determine whether the electric vehicle is charging, thereby avoiding the occurrence of misjudgment events of electric bicycle charging.

[0050] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of an electric vehicle charging detection method proposed in one embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, an embodiment of the present invention provides an electric vehicle charging detection method, the method comprising the following steps:

[0054] S10, taking the change of the transient waveform of active power in the socket as a transient event, and monitoring the transient event based on a time sliding window;

[0055] S20, when the transient event ends, calculating the change in active power before and after the transient event occurs;

[0056] S30, when comparing the change in active power with the electric vehicle active power threshold to determine that an electric vehicle is plugged into the socket, collecting an active power sampling sequence in the socket within the time sliding window;

[0057] S40: Process the active power sampling sequence based on an electric vehicle charging detection model to detect the charging state of the electric vehicle.

[0058] It should be noted that the data required for this embodiment is active power. The active power waveform focuses on the changes in power demand in the power system. By treating the process of plugging an electrical appliance into a socket as a transient event, it is determined whether an electrical appliance is plugged into the socket based on the changes in the active power transient waveform in the socket. The difference between the active power before and after is compared with the power threshold to preliminarily determine whether an electric vehicle is plugged into the socket. The sampling points of the active power waveform are then input into the electric vehicle charging detection model to further determine whether the electric vehicle is being charged, thereby avoiding the occurrence of misjudgment of electric bicycle charging events.

[0059] As a further preferred technical solution, the step S10: taking the active power transient waveform in the socket as a transient event, and monitoring the transient event based on a time sliding window, includes:

[0060] S11, using the time sliding window N of length 2D to monitor the transient event, the time sliding window includes time sub-windows N1 and N2, where N1 = [P1, ..., P D ],N2=[P D+1 ,…,P 2D ], P D represents the active power at the Dth sampling point in the time sliding window N;

[0061] S12, based on the active power in the time sliding window N, calculate the variance F1 of the active power sampled in the time subwindow N1 and the variance F2' of the active power in the time subwindow N2, wherein the variance F2' is calculated based on the average value of the active power sampled in the time subwindow N1;

[0062] S13, when F2'>F1β FF When , it is determined that a transient event has occurred, where β FF is the threshold of the active power variance multiple of time sub-window N1 and time sub-window N2.

[0063] It should be noted that the variance is used in this embodiment to measure the volatility of active power in a time sub-window. The calculation is based on the average value of the active power in the time sub-window. Through the time sliding window, the change of active power can be dynamically tracked and abnormal power fluctuations can be discovered in time.

[0064] Specifically, the following formula is used to establish the data information related to the transition from the steady state region to the transient event region as the time sliding window N passes:

[0065]

[0066] In the formula, represents the average value of the active power sampled in the time sub-window N1; F1 represents the variance of the active power sampled in the time sub-window N1; F2' represents the variance of the active power in the time sub-window N2 based on the average value of the active power in the time sub-window N1.

[0067] Specifically, if F2'>F1β FF , it is determined that a transient event has occurred, where β FF is the threshold of the active power variance multiple of the time sub-windows N1 and N2; otherwise, it is considered that the power change does not reach the threshold and it is determined that no transient event occurs.

[0068] As a further preferred technical solution, step S20: when the transient event ends, calculating the change in active power before and after the transient event occurs, includes:

[0069] S21, based on the active power in the time sliding window N, calculate the variance F2 of the active power sampled in the time subwindow N2 and the variance F1' of the active power in the time subwindow N1, wherein the variance F1' is calculated based on the average value of the active power sampled in the time subwindow N2;

[0070] S22, in the case of meeting F1' <F2β FF When , it is determined that the transient event ends;

[0071] S23. Calculate the change in active power before and after the transient event occurs based on the active power at the starting point and the active power at the end point of the transient event.

[0072] Specifically, the following formula is used to establish the data information related to the transition from the transient event region to the steady state region as the time sliding window N passes:

[0073]

[0074] In the formula, represents the average value of the active power sampled in the time sub-window N2; F2 represents the variance of the active power sampled in the time sub-window N2; F1' represents the variance of the active power in the time sub-window N1 based on the average value of the active power in the time sub-window N2.

[0075] Specifically, if F1' <F2β FF, it is determined that the transient event is over; otherwise, it is determined that the transient event is in the process of occurring.

[0076] As a further preferred technical solution, the method further comprises the following steps:

[0077] From the first sampling point of the time sliding window N, all sampling points are traversed to the right, and the first traversal to the point that satisfies The sampling point is the starting point of the transient event;

[0078] From the 2Dth sampling point of the time sliding window N, traverse all sampling points to the left, and traverse to the first point that satisfies The sampling point is the end point of the transient event;

[0079] in, represents the average value of active power sampled in time sub-window N1, represents the average value of active power sampled in time sub-window N2, P n Represents the active power at the nth sampling point.

[0080] This embodiment assumes that an electrical appliance is plugged into a socket between the starting point and the end point of the transient event.

[0081] Furthermore, the calculation formula for the change in active power before and after the transient event occurs is:

[0082]

[0083] Where ΔP represents the change in active power before and after the transient event occurs, P start Represents the active power at the starting point of the transient event, P over Represents the active power at the end point of the transient event, P over+n represents the active power at n sampling points after the end of the transient event, P start-n Represents the active power of n sampling points before the starting point of the transient event.

[0084] It should be noted that if the value of n is too small or too large, the calculation of the change in active power will result in a large error. In this embodiment, n is 8, and the change in active power can be better reflected through 8 sampling points.

[0085] As a further preferred technical solution, the step S30: when comparing the change in the active power with the electric vehicle active power threshold to determine that an electric vehicle is plugged into the socket, collecting the active power sampling sequence in the socket within the time sliding window includes:

[0086] S31, when the change in active power is greater than the electric vehicle active power threshold, preliminarily determining that an electric vehicle is plugged into the socket;

[0087] Specifically, the change in active power ΔP before and after the transient event occurs is compared with the set electric vehicle active power threshold For comparison, if If the power is higher than 10V, it is determined that an electric vehicle is plugged into the socket. Otherwise, it is considered that the power change during charging of the electric vehicle has not been reached, and the electric vehicle may not be charging.

[0088] S32, collecting the active power of the socket at each sampling time in the time sliding window, traversing each active power sampling value and performing preprocessing to obtain a processed active power sampling sequence.

[0089] As a further preferred technical solution, the step S32: collecting the active power of the socket at each sampling time in the time sliding window, traversing each active power sampling value and performing preprocessing to obtain a processed active power sampling sequence, includes the following steps:

[0090] S321, collecting the active power [P1, ...P D ,…,P D+l …,P 2D ];

[0091] S322, determine the active power P of the D+1th sampling point D+l Whether it is an abnormal value or a missing value, if yes, execute step S323 or S324, if no, execute step S325;

[0092] S323, in active power P D+l is an abnormal value, then the active power P D+l Corrected to:

[0093]

[0094] Where P represents P1 to P 2D The active power of all sampling points. Average means taking the active power from P1 to P 2D The average value of Std is taken from P1 to P 2D The standard deviation of

[0095] S324, in active power P D+l If it is a missing value, then for the active power P D+l Corrected to:

[0096]

[0097] In the formula, None means that the sampling point is a set of missing values;

[0098] S325, after assigning l+1 to l, re-determine the active power P of the D+lth sampling pointD+l Whether it is an abnormal value or a missing value, until l>D, the processed active power sampling sequence is recorded as

[0099] This embodiment can fill in data gaps by correcting missing values ​​and outliers, reduce analysis errors caused by missing or anomaly data, and improve data integrity and consistency.

[0100] As a further preferred technical solution, the electric vehicle charging detection model adopts a GRU neural network model, which includes three GRU layers, each GRU layer includes K GRU blocks, and each GRU block includes a reset gate R t and an update gate Z t ;

[0101] The output of the third GRU layer is input into the fully connected layer, which is followed by a sigmoid activation layer. The sigmoid activation layer outputs a two-dimensional state vector, which respectively indicates that the electric vehicle is charging and the non-electric vehicle is charging.

[0102] This embodiment determines whether the electric vehicle is being charged and again determines whether the electric vehicle is being charged or other electrical appliances other than the electric vehicle are being charged, uses a sliding window technique and then uses a neural network to further perform charging discrimination, thereby reducing misjudgment through two judgments.

[0103] Among them, the following formula is used to establish the GRU block update gate Z at time t t , Reset Gate R t And output H t :

[0104]

[0105] Where W Z , U Z Both represent the update of gate Z at time t t The weight of B Z Indicates that gate Z is updated at time t t Bias of W R , U R Both indicate that the gate R is reset at time t t The weight of B R Reset gate R at time t t Bias of Represents the candidate hidden state at time t, output H at time t-1 t-1 and input X at time t t Obtained by weighted linear combination; W H and U H Both represent the hidden state at time t The weight of BH Indicates the hidden state at time t Bias of represents the sigmoid activation function; tanh represents the hyperbolic tangent activation function; Denotes the Hadamard product.

[0106] Then the output of the third GRU layer is input into the fully connected layer, and the sigmoid layer outputs the classified state vector. Each component of the state vector represents that the electric vehicle is charging and the non-electric vehicle is charging. The formula is expressed as:

[0107]

[0108] Where h represents the input of the sigmoid layer; S(h) represents the output of the sigmoid layer, which is the state vector after classification.

[0109] As a further preferred technical solution, before processing the active power sampling sequence based on the electric vehicle charging detection model to detect the charging state of the electric vehicle, the method further includes:

[0110] Based on the constructed data set, the GRU neural network model is trained by back propagation and gradient descent algorithms, and the cross entropy loss function L is calculated to update the network parameters until the maximum number of iterations is reached or the loss function value converges, thereby obtaining the electric vehicle charging detection model.

[0111] Specifically, the cross entropy loss function L used in the back propagation process of the GRU neural network model is established:

[0112]

[0113] In the formula, is the predicted value output by the GRU neural network model; y i is the true label value of the data; M is the number of samples.

[0114] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0115] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0116] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for detecting charging of an electric vehicle, characterized in that: The method comprises: Taking the change of the transient waveform of active power in the socket as a transient event, and monitoring the transient event based on a time sliding window; When the transient event ends, calculating the change in active power before and after the transient event occurs; When comparing the change in active power with the electric vehicle active power threshold to determine that an electric vehicle is plugged into the socket, an active power sampling sequence in the socket within the time sliding window is collected; The active power sampling sequence is processed based on the electric vehicle charging detection model to detect the charging state of the electric vehicle.

2. The electric vehicle charging detection method according to claim 1, characterized in that: The method of taking the active power transient waveform in the socket as a transient event and monitoring the transient event based on a time sliding window includes: The transient event is monitored using the time sliding window N of length 2D, wherein the time sliding window includes time sub-windows N1 and N2, where N1 = [P1, ..., P D ],N2=[P D+1 ,…,P 2D ], P D represents the active power at the Dth sampling point in the time sliding window N; Based on the active power in the time sliding window N, calculate the variance F1 of the active power sampled in the time subwindow N1 and the variance F2' of the active power in the time subwindow N2, wherein the variance F2' is calculated based on the average value of the active power sampled in the time subwindow N1; When F2'>F1β FF When , it is determined that a transient event has occurred, where β FF is the threshold of the active power variance multiple of time sub-window N1 and time sub-window N2.

3. The electric vehicle charging detection method according to claim 2, characterized in that: When the transient event ends, calculating the change in active power before and after the transient event occurs includes: Based on the active power in the time sliding window N, calculate the variance F2 of the active power sampled in the time subwindow N2 and the variance F1' of the active power in the time subwindow N1, wherein the variance F1' is calculated based on the average value of the active power sampled in the time subwindow N2; In meeting F1' <F2β FF When , it is determined that the transient event ends; The change in active power before and after the transient event is calculated based on the active power at the starting point and the active power at the end point of the transient event.

4. The electric vehicle charging detection method according to claim 3, characterized in that: The calculation formula of the variance F2' of the active power in the time sub-window N2 is: The calculation formula of the variance F1' of the active power in the time sub-window N1 is: In the formula, Represents the average value of the active power sampled in the time sub-window N1.

5. The electric vehicle charging detection method according to claim 3, characterized in that: The method further comprises: From the first sampling point of the time sliding window N, all sampling points are traversed to the right, and the first traversal to the point that satisfies The sampling point is the starting point of the transient event; From the 2Dth sampling point of the time sliding window N, traverse all sampling points to the left, and traverse to the first point that satisfies The sampling point is the end point of the transient event; in, represents the average value of active power sampled in time sub-window N1, represents the average value of active power sampled in time sub-window N2, P n Represents the active power at the nth sampling point.

6. The electric vehicle charging detection method according to claim 1, characterized in that: The calculation formula for the change in active power before and after the transient event is: Where ΔP represents the change in active power before and after the transient event occurs, P start Represents the active power at the starting point of the transient event, P over Represents the active power at the end point of the transient event, P over+n represents the active power at n sampling points after the end of the transient event, P start-n Represents the active power of n sampling points before the starting point of the transient event.

7. The electric vehicle charging detection method according to claim 1, characterized in that: The step of comparing the change in active power with the electric vehicle active power threshold to determine that an electric vehicle is plugged into the socket, and collecting an active power sampling sequence in the socket within the time sliding window, comprises: When the change in active power is greater than the electric vehicle active power threshold, it is preliminarily determined that an electric vehicle is plugged into the socket; The active power of the socket at each sampling time in the time sliding window is collected, each active power sampling value is traversed and preprocessed, and a processed active power sampling sequence is obtained.

8. The electric vehicle charging detection method according to claim 7, characterized in that: The collecting of the active power of the socket at each sampling time in the time sliding window, traversing each active power sampling value and performing preprocessing to obtain a processed active power sampling sequence, includes: Collect the active power [P1,…P D ,…,P D+l …,P 2D ]; Determine the active power P of the D+lth sampling point D+l Whether it is an outlier or missing value; If so, then the active power P D+l is an abnormal value, then the active power P D+l Corrected to: Where P represents P1 to P 2D The active power of all sampling points. Average means taking the active power from P1 to P 2D The average value of Std is taken from P1 to P 2D The standard deviation of In the active power P D+l If it is a missing value, then for the active power P D+l Corrected to: In the formula, None means that the sampling point is a set of missing values; If not, assign l+1 to l and re-judge the active power P of the D+lth sampling point. D+l Whether it is an abnormal value or a missing value, until l>D, the processed active power sampling sequence is obtained.

9. The electric vehicle charging detection method according to claim 1, characterized in that: The electric vehicle charging detection model adopts a GRU neural network model, which includes three GRU layers, each GRU layer includes K GRU blocks, and each GRU block includes a reset gate R t and an update gate Z t ; The output of the third GRU layer is input into the fully connected layer, which is followed by a sigmoid activation layer. The sigmoid activation layer outputs a two-dimensional state vector, which respectively indicates that the electric vehicle is charging and the non-electric vehicle is charging.

10. The electric vehicle charging detection method according to any one of claims 1 to 9, characterized in that: Before processing the active power sampling sequence based on the electric vehicle charging detection model to detect the charging state of the electric vehicle, the method further includes: Based on the constructed data set, the GRU neural network model is trained by back propagation and gradient descent algorithms, and the cross entropy loss function L is calculated to update the network parameters until the maximum number of iterations is reached or the loss function value converges, thereby obtaining the electric vehicle charging detection model.

Images