CN118260625B - Charging current abnormality detection method and terminal - Google Patents

Abstract

The present invention discloses a charging current anomaly detection method and terminal, which obtains a target current sequence within a first preset time period in a charging order of a target vehicle; performs a first-order difference on the current sequence to obtain a differential sequence; calculates a preset differential statistic based on the sequence; clusters the differential statistics corresponding to all charging orders of the target vehicle according to a preset number of categories to obtain clusters; obtains the first quantile in each cluster and the second quantile symmetrical to the first quantile; and determines whether the sum between all first quantiles and second quantiles in each cluster belongs to the neighborhood of 0, and if so, marks the corresponding charging order as a normal order; otherwise, marks the corresponding charging order as an abnormal order. The present invention visualizes the fluctuations of the reaction battery state that may occur during the charging process by calculating the differential sequence and the corresponding differential statistic, further improving the accuracy of the final abnormality judgment.

Description

A charging current abnormality detection method and terminal

Technical Field

The present invention relates to the field of charging detection, and in particular to a charging current abnormality detection method and terminal.

Background Art

Due to the continuous reduction of traditional energy and the pollution to the environment caused by the use of traditional energy, the utilization and development of new energy has been raised to a new level. In the automotive industry, the popularity of electric vehicles is getting higher and higher; and batteries are the core components of electric vehicles. In order to ensure the safety of electric vehicles during use, battery safety monitoring is particularly important. In the existing technology, the collection of battery data is usually focused on the data when the vehicle is in use, and more attention is paid to the dangerous situations that may occur in the battery during the driving or charging process of the vehicle, so the amount of data collected is relatively large.

Summary of the invention

The technical problem to be solved by the present invention is to provide a charging current abnormality detection method and a terminal to improve the efficiency of judging the battery status.

In order to solve the above technical problems, a technical solution adopted by the present invention is:

A method for detecting abnormal charging current includes the following steps:

Obtaining a target current sequence within a first preset time period in a charging order for a target vehicle;

Performing a first-order difference on the current sequence to obtain a differential sequence;

Calculating a preset difference statistic according to the difference sequence;

Clustering the difference statistics corresponding to all charging orders of the target vehicle according to a preset number of categories to obtain cluster clusters;

Obtain the first quantile in each cluster and the second quantile symmetrical to the first quantile;

For each of the clusters, determine whether the sum of all the first quantiles and the second quantiles in the cluster belongs to a neighborhood of 0. If so, mark the charging order corresponding to the differential statistic in the cluster as a normal order; otherwise, mark the charging order corresponding to the differential statistic in the cluster as an abnormal order.

In order to solve the above technical problems, another technical solution adopted by the present invention is:

A charging current abnormality detection terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

Obtaining a target current sequence within a first preset time period in a charging order for a target vehicle;

Performing a first-order difference on the current sequence to obtain a differential sequence;

Calculating a preset difference statistic according to the difference sequence;

Clustering the difference statistics corresponding to all charging orders of the target vehicle according to a preset number of categories to obtain cluster clusters;

Obtain the first quantile in each cluster and the second quantile symmetrical to the first quantile;

For each of the clusters, determine whether the sum of all the first quantiles and the second quantiles in the cluster belongs to a neighborhood of 0. If so, mark the charging order corresponding to the differential statistic in the cluster as a normal order; otherwise, mark the charging order corresponding to the differential statistic in the cluster as an abnormal order.

The beneficial effects of the present invention are as follows: a target current sequence within a first preset time period in a charging order is obtained by performing differential analysis to obtain a differential sequence, and a differential statistic is obtained based on the differential sequence; an abnormality identification model trained in advance is used to obtain an abnormality determination result based on the differential statistic and vehicle information; the target current sequence within a preset time period is obtained for subsequent processing before inputting into the model, thereby reducing the amount of data to be processed; and by calculating the differential sequence and the corresponding differential statistic, the fluctuation of the reaction battery state that may occur during the charging process is visualized, thereby further improving the accuracy of the final abnormality determination; the data obtained from all orders corresponding to a single vehicle are obtained; The obtained differential statistics are clustered to obtain clusters, and the symmetrical first quantile and second quantile in each cluster are added to determine whether the sum belongs to the neighborhood of 0, that is, the quantile in the differential sequence can reflect the degree of change of the current in the target current sequence, and if the sum does not belong to the neighborhood of 0, it means that the fluctuation is not offset during the charging process, and it is likely that the battery itself is abnormal. Therefore, the corresponding orders in the entire cluster are marked as abnormal orders at this time, because there are different stages for the same battery over time. Clustering in advance and then judging can improve the efficiency of screening out abnormal orders while ensuring the accuracy.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flowchart of a method for detecting abnormal charging current according to an embodiment of the present invention;

FIG2 is a flow chart of a prediction process of a charging current abnormality detection method according to an embodiment of the present invention;

FIG3 is a flow chart of a model training process of a charging current anomaly detection method according to an embodiment of the present invention;

FIG4 is a schematic diagram of a prediction model structure of a charging current abnormality detection method according to an embodiment of the present invention;

FIG5 is a schematic diagram of the structure of a charging current abnormality detection terminal according to an embodiment of the present invention;

Description of labels:

1. A charging current abnormality detection terminal; 2. A processor; 3. A memory.

DETAILED DESCRIPTION

In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.

Referring to FIG. 1 , a method for detecting abnormal charging current includes the following steps:

Obtaining a target current sequence within a first preset time period in a charging order for a target vehicle;

Performing a first-order difference on the current sequence to obtain a differential sequence;

Calculating a preset difference statistic according to the difference sequence;

Clustering the difference statistics corresponding to all charging orders of the target vehicle according to a preset number of categories to obtain cluster clusters;

Obtain the first quantile in each cluster and the second quantile symmetrical to the first quantile;

For each of the clusters, determine whether the sum of all the first quantiles and the second quantiles in the cluster belongs to a neighborhood of 0. If so, mark the charging order corresponding to the differential statistic in the cluster as a normal order; otherwise, mark the charging order corresponding to the differential statistic in the cluster as an abnormal order.

From the above description, it can be seen that the beneficial effects of the present invention are: obtaining the target current sequence within the first preset time period in the charging order, performing differential analysis to obtain a differential sequence, and obtaining a differential statistic based on the differential sequence, obtaining an abnormality judgment result based on the differential statistic and vehicle information through a pre-trained abnormality recognition model, and obtaining the target current sequence within the preset time period for subsequent processing before inputting the model to reduce the amount of data processing, and by calculating the differential sequence and the corresponding differential statistic, the fluctuations in the reaction battery state that may occur during the charging process are visualized, further improving the accuracy of the final abnormality judgment; obtaining all the corresponding abnormality sequences for a single vehicle The differential statistics obtained from the orders are clustered to obtain clusters, and the symmetrical first quantile and second quantile in each cluster are added to determine whether the sum belongs to the neighborhood of 0, that is, the quantile in the differential sequence can reflect the degree of change of the current in the target current sequence, and if the sum does not belong to the neighborhood of 0, it means that the fluctuation is not offset during the charging process, and it is likely that the battery itself is abnormal. Therefore, the corresponding orders in the entire cluster are marked as abnormal orders at this time, because there are different stages for the same battery over time. Clustering in advance and then judging can improve the efficiency of screening out abnormal orders while ensuring the accuracy.

Further, the step of obtaining a target current sequence within a first preset time period in a target vehicle charging order includes:

Obtain the constant current value in the constant current section of the target vehicle charging order;

Acquire an initial current sequence of the target vehicle within a first preset time period before the end of the charging time of the charging order;

Dividing each sequence current value in the current sequence by the constant current value to obtain a target current value;

A target current sequence is obtained according to the target current value.

From the above description, it can be seen that as the amount of electricity stored in the battery gradually increases during charging, the battery state is usually relatively stable at this time, and the current change state before the end of the charging time, that is, at the end of the charging process, can reflect the health of the battery to a certain extent. Therefore, obtaining the current sequence in the first preset time period before the end of the charging time for processing can increase the probability of discovering battery faults, and thereby improve the accuracy of the final abnormality judgment result.

Further, the calculating a preset difference statistic according to the difference sequence includes:

The maximum value, minimum value, standard deviation, mean, maximum absolute value, difference between the maximum value and the minimum value, and a preset number of quantiles in the difference sequence are calculated.

From the above description, we can see that by obtaining differential statistics such as quantiles, maximum values, and minimum values based on the differential sequence, the characteristics of the differential sequence are further decomposed, thereby providing more in-depth and multi-dimensional training features for the model prediction, avoiding one-sidedness in the model calculation process, and thus improving the accuracy of the model prediction results.

Furthermore, after determining whether the sum of all the first quantiles and the second quantiles belongs to a neighborhood of 0, the method further includes:

Constructing a model training set according to the vehicle information, the marked charging order, and the differential statistics corresponding to each charging order;

The initial anomaly recognition model is trained according to the model training set to obtain an anomaly recognition model.

From the above description, it can be seen that by training the initial anomaly recognition model based on the marked normal orders and abnormal orders and their corresponding differential statistics, the correspondence between the order labels, differential statistics and vehicle information can be established, so that the trained anomaly recognition model has the ability to determine whether an order is abnormal.

Further, the first quantile is the eleventh percentile; and the second quantile is the eighty-eighth percentile.

From the above description, it can be seen that the 11th percentile and the 88th percentile include the start position and the end position of the difference sequence, which are representative for reflecting the fluctuation of the value within the first preset time period.

Furthermore, the preset number of quantiles includes:

Get the preset number of divisions;

Dividing the difference sequence according to the number of divisions to obtain the quantile of the first number;

A preset number of quantiles are obtained from the quantiles of the first number by equidistant sampling.

From the above description, it can be seen that the number of divisions and the preset number can be adjusted according to different scenarios, vehicle characteristics, battery characteristics or model training requirements to obtain different quantile results for model training and prediction process, thereby improving the accuracy of model training.

Furthermore, the number of divisions is 100 and the preset number is 8.

From the above description, we can see that the quantiles corresponding to the difference sequence are divided into percentiles and then 8 of them are obtained as the quantiles used in the model training and prediction process, which takes into account the amount of data and accuracy processed by the model.

Furthermore, after calculating the preset difference statistic according to the difference sequence, the method further includes:

The differential statistic and the vehicle information of the target vehicle are input into a pre-trained abnormality recognition model to obtain an abnormality determination result.

From the above description, it can be seen that after the abnormality recognition model is trained, when it is necessary to judge whether an order is abnormal, it can be judged directly through differential statistics and vehicle information. While reducing the amount of data that needs to be obtained, the accuracy of recognition is guaranteed through model training.

Furthermore, the abnormality determination result includes an abnormality probability of the charging order.

From the above description, it can be seen that the abnormal probability corresponding to the charging order is output as the abnormality judgment result, which provides a quantitative data reference for judging whether there is an abnormality in the battery.

Please refer to Figure 4, a charging current abnormality detection terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements each step of the above-mentioned charging current abnormality detection method when executing the computer program.

The above-mentioned method for detecting abnormal current of the present invention can be applied to the scenario of detecting whether a battery, especially a battery in an electric vehicle, has abnormal risk, which is described below through a specific implementation method.

Please refer to Figures 1-3, the first embodiment of the present invention is:

A method for detecting abnormal charging current includes the following steps:

S1. Obtaining a target current sequence within a first preset time period in a target vehicle charging order, including:

S10, obtaining a charging current sequence corresponding to a charging order of a target vehicle;

S11, obtaining the constant current value of the target vehicle in the constant current section in the charging order;

S12, obtaining an initial current sequence of the target vehicle within a first preset time period before the charging time of the charging order ends;

In an optional embodiment, an initial current sequence within 60 seconds before the end of the charging time is obtained;

S13, dividing each current value in the current sequence by the constant current value to obtain a target current value;

S14, obtaining a target current sequence according to the target current values, including: arranging the target current values according to the arrangement order of their corresponding sequence current values to obtain a target current sequence;

Divide each sequence current value by the constant current value to obtain a dimensionless target current value, eliminating the influence of external current when different orders are generated, thereby improving the accuracy of battery status judgment; "dimensionless" is a mathematical and physics term, which means that a certain quantity does not depend on any specific dimension in the unit selection; in some physical and engineering problems, dimensionless quantities are often introduced to simplify the problem or to more conveniently compare the relationship between different physical quantities. By choosing an appropriate unit, the physical quantity can be simplified to a pure number without involving any specific unit;

S2, performing first-order difference on the current sequence to obtain a differential sequence;

The first-order difference refers to the difference between two consecutive terms in a sequence or time series. Specifically, for a sequence {a1, a2, a3, ..., an}, its first-order difference sequence is {a2-a1, a3-a2, ..., an-an-1};

S3, calculating a preset differential statistic according to the differential sequence, including: respectively calculating the maximum value (diff_max), the minimum value (diff_min), the standard deviation (diff_std), the mean (diff_mean), the maximum absolute value (i.e., the maximum absolute value diff_length in the differential sequence), the difference between the maximum value and the minimum value (diff_max_min), and a preset number of quantiles in the differential sequence to form the differential statistic;

Wherein, calculating the preset number of quantiles includes:

S31, obtaining a preset number of divisions;

S32, dividing the difference sequence according to the number of divisions to obtain the quantile of the first number;

S33, obtaining a preset number of quantiles from the quantiles of the first number by equidistant sampling;

In an optional implementation, the number of divisions is 100, the preset number is 8, and 99 quantiles are obtained. The preset number of quantiles finally obtained is:

diff_q11: the 11th percentile of the difference sequence; diff_q22: the 22nd percentile of the difference sequence; diff_q33: the 33rd percentile of the difference sequence; diff_q44: the 44th percentile of the difference sequence; diff_q55: the 55th percentile of the difference sequence; diff_q66: the 66th percentile of the difference sequence; diff_q77: the 77th percentile of the difference sequence; diff_q88: the 88th percentile of the difference sequence;

S4, inputting the differential statistic and the vehicle information of the target vehicle into a pre-trained abnormality recognition model to obtain an abnormality determination result of the charging order;

In an optional implementation, the vehicle information includes batvin (vehicle vin identification code), battery type, battery manufacturer, etc.;

In an optional embodiment, the abnormality determination result includes the abnormality probability of the charging order; that is, the abnormal or normal determination result can be directly output according to the probability, or it can be displayed together with the probability value when outputting;

Referring to FIG. 3 , in an optional implementation, S3 further includes an abnormality recognition model training process:

S5, clustering the differential statistics corresponding to all charging orders of the target vehicle according to a preset number of categories to obtain clusters; that is, clustering all charging orders for the same target vehicle;

In an optional implementation, the clustering method is k-means, and the number of preset categories is 2;

S6, obtaining a first quantile in each cluster and a second quantile symmetrical to the first quantile;

In an optional implementation, the first quantile is the eleventh percentile; the second quantile is the eighty-eighth percentile; each differential statistic is marked with an order to facilitate distinguishing differential statistics from different orders;

S7. For each of the clusters, determine whether the sum of all the first quantiles and the second quantiles in the cluster belongs to a neighborhood of 0. If so, mark the charging order corresponding to the differential statistic in the cluster as a normal order; otherwise, mark the charging order corresponding to the differential statistic in the cluster as an abnormal order.

Among them, "neighborhood" usually refers to the local area or neighborhood around a specific point. This concept is often used in fields such as mathematics, physics, and computer science.

S8. Construct a model training set according to the vehicle information, the marked charging order, and the differential statistics corresponding to each charging order;

S9, training an initial anomaly recognition model according to the model training set to obtain an anomaly recognition model;

Please refer to FIG. 3 , the trained abnormality recognition model is used for calculating the abnormality probability value in step S4;

Referring to FIG. 4 , the anomaly recognition model includes:

(1) Sigmoid function: One of the activation functions, it is a nonlinear function with the formula:

;

Where x is the input, is the output of the Sigmoid function, with a value range between 0 and 1. This allows the Sigmoid function to map inputs to probability values in some cases;

(2) Dense layer: fully connected layer, used for feature integration;

(3) Concatenate: An operation that connects two or more objects (usually strings, lists, arrays, etc.) in sequence to form a larger object;

(4) Embedding: Embedding refers to the process of mapping high-dimensional data into a low-dimensional space. This process can retain the important features of the original data and can usually represent the data more effectively, thus facilitating subsequent processing and analysis.

Please refer to FIG. 5 , the second embodiment of the present invention is:

A charging current abnormality detection terminal 1 includes a processor 2, a memory 3, and a computer program stored in the memory 3 and executable on the processor 2. When the processor 2 executes the computer program, each step in the first embodiment is implemented.

In summary, the present invention provides a charging current anomaly detection method and terminal, which obtain the battery status by extracting the terminal current during a single charging process for analysis, thereby reducing the amount of data analysis, and analyzing the change in current during charging, so as to give early warning of quality problems before the battery is normally used, thereby improving the safety of the battery during use. At the same time, by constructing a differential sequence and calculating the corresponding differential statistics, training an abnormality recognition model and obtaining the abnormal judgment result of the order through the abnormality recognition model, the accuracy of the judgment result can be improved by combining multiple different conditions; at the same time, in the process of training the abnormality recognition model, normal orders and abnormal orders are marked by clustering the orders of the same vehicle and calculating the quantile sum, thereby comprehensively considering the continuity of the battery status in a single vehicle, eliminating the need to mark each order separately, thereby improving the processing efficiency.

The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims (6)

1. A method for detecting abnormal charging current, comprising the steps of: Obtaining a target current sequence within a first preset time period in a charging order for a target vehicle; Performing a first-order difference on the current sequence to obtain a differential sequence; Calculating a preset difference statistic according to the difference sequence, including: calculating the maximum value, minimum value, standard deviation, mean, maximum absolute value, difference between the maximum value and the minimum value, and a preset number of quantiles in the difference sequence; Clustering the difference statistics corresponding to all charging orders of the target vehicle according to a preset number of categories to obtain cluster clusters; Obtaining a first quantile in each cluster and a second quantile symmetrical to the first quantile; the first quantile and the second quantile are among a preset number of quantiles; For each of the clusters, determine whether the sum of all the first quantiles and the second quantiles in the cluster belongs to a neighborhood of 0. If so, mark the charging order corresponding to the differential statistic in the cluster as a normal order; otherwise, mark the charging order corresponding to the differential statistic in the cluster as an abnormal order; the neighborhood refers to a local area or neighborhood around 0; Constructing a model training set according to the vehicle information, the marked charging order, and the differential statistics corresponding to each charging order; Training an initial anomaly recognition model according to the model training set to obtain an anomaly recognition model; The preset number of quantiles includes: Get the preset number of divisions; Dividing the difference sequence according to the number of divisions to obtain the quantile of the first number; Obtaining a preset number of quantiles from the quantiles of the first number by equally spaced sampling; Also includes: Obtaining a target pending current sequence within a first preset time period in a pending vehicle charging order; Performing first-order difference on the undetermined current sequence to obtain an undetermined differential sequence; Calculating a preset undetermined difference statistic according to the undetermined difference sequence; The to-be-determined differential statistic and the vehicle information of the to-be-determined vehicle are input into a pre-trained anomaly recognition model to obtain an anomaly determination result. 2. A charging current anomaly detection method according to claim 1, characterized in that the step of obtaining a target current sequence within a first preset time period in a charging order of a target vehicle comprises: Obtain the constant current value in the constant current section of the target vehicle charging order; Acquire an initial current sequence of the target vehicle within a first preset time period before the end of the charging time of the charging order; Dividing each sequence current value in the current sequence by the constant current value to obtain a target current value; A target current sequence is obtained according to the target current value. 3. A charging current abnormality detection method according to claim 1, characterized in that the first quantile is the eleventh percentile; and the second quantile is the eighty-eighth percentile. 4 . The method for detecting abnormal charging current according to claim 1 , wherein the number of divisions is 100 and the preset number is 8. 5. A charging current anomaly detection method according to claim 1, characterized in that the anomaly determination result includes the anomaly probability of the charging order. 6. A charging current anomaly detection terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements each step of a charging current anomaly detection method as described in any one of claims 1 to 5 when executing the computer program.