Insulation anomaly detection method for new energy vehicle power battery system

Abstract

The application relates to the technical field of power batteries, in particular to a new energy automobile power battery system insulation abnormality detection method, which comprises the following steps: obtaining original messages, performing analysis, extracting battery signal data, and performing data cleaning and normalization processing; the battery signal data is subjected to feature selection, insulation resistance, voltage, current and temperature related to the insulation state are selected as feature data; the first k modes of the insulation resistance, the voltage, the current and the temperature are calculated as initial values; the feature data is classified based on a k-means clustering algorithm to obtain k-class classification features; the clustering effect of the clustering results obtained by different k is evaluated by using a contour coefficient; the best k of the clustering effect is selected, and the abnormality degree of different categories is analyzed; and insulation abnormality is judged according to the difference of the k-class classification features. The scheme has strong anti-interference capability, is favorable for stably obtaining detection results with higher precision, and is convenient for taking measures for processing in a more timely manner.

Description

Technical Field

[0001] This invention relates to the field of power battery technology, specifically to a method for detecting insulation abnormalities in a power battery system for new energy vehicles. Background Technology

[0002] The insulation anomaly detection of the power battery system in new energy vehicles aims to monitor and detect the insulation status of the power battery system to ensure the safety and reliability of the battery system. It is an important safety measure designed to prevent short circuits between the high voltage of the battery and the vehicle chassis, thereby avoiding potential fires or other safety accidents.

[0003] Generally, methods for detecting abnormal battery insulation include measuring the resistance between the battery and the vehicle chassis. Under normal conditions, this resistance value will be very high, indicating good insulation. If this resistance value drops, it may indicate a short circuit between the battery and the vehicle chassis. In this case, the vehicle should be taken out of service immediately and repairs should be contacted.

[0004] However, existing technologies have some problems or shortcomings:

[0005] 1. Some insulation anomaly detection methods may be sensitive to environmental changes and interference, which may lead to false alarms or missed alarms. They have low anti-interference ability, which affects the accuracy of subsequent analysis results.

[0006] 2. Insulation status is affected by a variety of factors, such as temperature, humidity, current, and voltage. Relying on a single parameter may make it difficult to accurately determine the insulation status, resulting in low accuracy of the analysis results.

[0007] 3. Insulation anomaly detection methods require setting thresholds to determine whether the insulation condition is abnormal. However, setting an appropriate threshold remains a challenge and requires comprehensive consideration of multiple factors. Existing technologies rely on experience and historical data, and the thresholds set may not be the most suitable for the current situation, leading to a decrease in the accuracy of the analysis results.

[0008] Therefore, there is an urgent need for a method to detect insulation abnormalities in the power battery system of new energy vehicles, which has strong anti-interference capabilities and is conducive to obtaining more accurate detection results in a stable manner, so as to take more timely measures to deal with the problem. Summary of the Invention

[0009] The present invention aims to provide a method for detecting insulation abnormalities in the power battery system of new energy vehicles. It has strong anti-interference ability and is conducive to obtaining more accurate detection results in a stable manner, so as to take more timely measures to deal with the problem.

[0010] This invention provides the following basic solution: a method for detecting insulation anomalies in a power battery system for new energy vehicles, comprising the following contents:

[0011] S1. Obtain the original message, parse the message, and extract the battery signal data, which includes: insulation resistance, voltage, current, and temperature.

[0012] S2. Perform data cleaning and normalization on the battery signal data;

[0013] S3. Perform feature selection on the battery signal data, selecting insulation resistance, voltage, current and temperature related to the insulation state as feature data;

[0014] S4. Find the first k modes of insulation resistance, voltage, current and temperature as initial values;

[0015] S5. Classify the feature data based on the k-means clustering algorithm to obtain k-class classification features as the clustering result;

[0016] S6. Use the silhouette coefficient to evaluate the clustering effect of the clustering results obtained at different k.

[0017] S7. Select the clustering result with the best k and analyze the degree of anomaly in different categories;

[0018] S8. Determine the insulation anomaly based on the differences in the classification characteristics of class k.

[0019] Furthermore, S2 includes:

[0020] Delete invalid data;

[0021] Outliers are handled by using a sliding median to smooth out the battery signal data to some extent.

[0022] Delete invalid battery signal data;

[0023] The battery signal data is normalized.

[0024] Further, step S4 includes: for each insulation resistance, first calculating the first k modes of each insulation resistance, then calculating the modes of voltage, current, and temperature under that insulation resistance value, thereby obtaining k points as the initial values ​​for the k-means clustering algorithm, denoted as:

[0025] Furthermore, S5 includes:

[0026] S501, the initial k center points are the k initial values ​​selected in S4;

[0027] S502. Define the loss function as the sum of squared errors between each sample point and the center point of its class:

[0028]

[0029] Where x i For the i-th sample, c i It is x i Category to which it belongs The center point is the center point corresponding to the category to which it belongs, and M is the total number of samples;

[0030] S503. Calculate the distance between each sample point and the sample center of each category, and classify it into the category with the closest sample center.

[0031] S504. For all the sample points that have been classified, calculate new cluster centers according to the categories and use them as new sample centers.

[0032] S505. Determine whether the loss function has converged. If yes, proceed to S506; otherwise, proceed to S503.

[0033] S506, Stop the iteration and obtain the clustering results.

[0034] Furthermore, the contour coefficient is:

[0035]

[0036] in represents the intra-category dissimilarity, where n is the number of sample points within the current category;

[0037] b(i)=min{b1(i),b2(i),…,b m (i)}, where The dissimilarity between categories is represented by n, where n is the number of sample points within category l.

[0038] The average silhouette coefficient of each sample point is the silhouette coefficient S of the clustering result. The silhouette coefficient ranges from -1 to 1. The closer it is to 1, the better the clustering and separation. The silhouette coefficient is:

[0039]

[0040] in represents the intra-category dissimilarity, where n is the number of sample points within the current category;

[0041] b(i)=min{b1(i),b2(i),…,b m (i)}, where The dissimilarity between categories is represented by n, where n is the number of sample points within category l.

[0042] The average silhouette coefficient of each sample point is the silhouette coefficient S of the clustering result. The silhouette coefficient ranges from -1 to 1. The closer it is to 1, the better the clustering degree and separation degree are, and the better the clustering effect is.

[0043] The beneficial effects of this solution: Insulation abnormalities may manifest in many aspects of the power battery system. Some common manifestations include: decreased insulation resistance, increased capacitance, increased voltage drop, increased temperature, abnormal current, changes in insulation impedance, frequent fault alarms, decreased battery pack performance, and abnormal charging. Different insulation abnormalities may lead to different reactions.

[0044] Therefore, relying on a single parameter may be insufficient to accurately determine the insulation status. Compared to existing technologies that rely on a single parameter for detection, this solution comprehensively analyzes multiple parameters (insulation resistance, voltage, current, and temperature) and extracts them from the original message. This eliminates the need for additional data acquisition and monitoring systems and detection technologies, reducing costs. Furthermore, since the power battery system consists of numerous battery modules, the mutual influence between modules can increase the complexity of insulation anomalies, making it difficult to accurately identify the source of the anomaly. Multi-parameter analysis of the original message can pinpoint the source of the anomaly. After cleaning and normalizing the battery signal data to reduce the amount of data processing required for subsequent analysis, the insulation anomaly detection is performed based on the k-means clustering algorithm, which can more accurately determine whether the insulation status is abnormal. At the same time, the multi-parameter indicators have stronger anti-interference capabilities, which is conducive to obtaining more accurate detection results and enabling more timely intervention.

[0045] Furthermore, this solution can be validated under different environments and operating conditions to ensure its reliability and accuracy under various circumstances. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating an embodiment of the method for detecting insulation abnormalities in a new energy vehicle power battery system according to the present invention. Detailed Implementation

[0047] The following detailed description illustrates the specific implementation method:

[0048] The basic implementation examples are as follows: Figure 1 The following is a method for detecting insulation abnormalities in the power battery system of new energy vehicles:

[0049] S1. Obtain the original message, parse the message, and extract the battery signal data. The original message conforms to the GB32960 national standard. The battery signal data includes: insulation resistance, voltage, current, and temperature.

[0050] S2. Perform data cleaning and normalization on the battery signal data. Specifically, to ensure the accuracy of subsequent analysis, invalid data, such as NAN and spaces, are deleted. Outliers are handled by using the moving median filling method to smooth the data to some extent, thereby reducing the impact of outliers. Additionally, unreasonable data is deleted; for example, for voltage data, there may be some unreasonable values, such as voltages greater than 6V or less than 1V. These data need to be deleted to ensure the reliability of the analysis. To eliminate the influence of dimensions between indicators, data normalization is performed. After data normalization, all indicators are on the same order of magnitude, thus eliminating the adverse effects caused by outlier samples.

[0051] S3. Perform feature selection on the battery signal data, selecting insulation resistance, voltage, current, and temperature related to the insulation state as feature data; specifically, for the battery signal data after data cleaning and normalization, perform feature selection, selecting insulation resistance INSULATION_RESISTANCE, voltage SUM_VOLT, current SUM_CURRENT, and temperature SUM_TEMP related to the insulation state, and perform insulation anomaly detection using the selected battery signal data.

[0052] S4. Find the first k modes of insulation resistance, voltage, current and temperature as initial values;

[0053] Specifically, for each insulation resistance, the first k modes of each insulation resistance are calculated. Then, the modes of voltage, current, and temperature at that insulation resistance value are calculated respectively, thus obtaining k points as the initial values ​​for the k-means clustering algorithm, denoted as: Traditional k-means clustering algorithms randomly select k centers, but this can lead to centers being too close together, requiring more iterations to converge. Selecting initial centers that are as far apart as possible often yields better results. Furthermore, the concepts of classification and mode are somewhat similar; therefore, this approach sequentially calculates the first k modes of insulation resistance, voltage sum, and current / temperature sum after observing the feature data. This facilitates faster convergence in subsequent iterations and yields results more quickly.

[0054] S5. Classify the feature data based on the k-means clustering algorithm to obtain k-class classification features;

[0055] Specifically, S5 includes:

[0056] S501, the initial k center points are the k initial values ​​selected in S4;

[0057] S502. Define the loss function as the sum of squared errors between each sample point and the center point of its class:

[0058]

[0059] Where x i For the i-th sample, c i It is x i Category to which it belongs The center point is the center point corresponding to the category to which it belongs, and M is the total number of samples;

[0060] S503. Calculate the distance between each sample point and the sample center (center point) of each category, and classify it into the category with the closest sample center.

[0061] S504. For all the sample points that have been classified, calculate new cluster centers according to the categories and use them as new sample centers.

[0062] S505. Determine whether the loss function has converged. If yes, proceed to S506; otherwise, proceed to S503.

[0063] S506, Stop the iteration and obtain the clustering results.

[0064] S6. Use the silhouette coefficient to evaluate the clustering effect of the clustering results obtained at different k.

[0065] Specifically, the contour coefficient is:

[0066]

[0067] in represents the intra-category dissimilarity, where n is the number of sample points within the current category;

[0068] b(i)=min{b1(i),b2(i),…,b m (i)}, where denoted as the dissimilarity between categories, and n is the number of sample points within category l.

[0069] The average silhouette coefficient of each sample point is the silhouette coefficient S of the clustering result. The value of the silhouette coefficient is between [-1, 1]. The closer it is to 1, the better the clustering degree and separation degree are, that is, the better the clustering effect.

[0070] S7. Select the clustering result with the best k and analyze the degree of abnormality of different categories; for example, if k=2, it means that the sample points are divided into two categories: abnormal and normal.

[0071] S8. Determine insulation anomalies based on the differences in classification characteristics of k categories; for example, classify a test point by calculating the distance between the test point and the center of k categories and combining the different classification categories given in S7.

[0072] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for detecting insulation abnormalities in a power battery system for new energy vehicles, characterized in that, Includes the following: S1. Obtain the original message, parse the message, and extract the battery signal data, which includes: insulation resistance, voltage, current, and temperature. S2. Perform data cleaning and normalization on the battery signal data; S3. Perform feature selection on the battery signal data, selecting insulation resistance, voltage, current and temperature related to the insulation state as feature data; S4. Find the first k modes of insulation resistance, voltage, current and temperature as initial values; S5. Classify the feature data based on the k-means clustering algorithm to obtain k-class classification features as the clustering result; S6. Use the silhouette coefficient to evaluate the clustering effect of the clustering results obtained at different k. S7. Select the clustering result with the best k and analyze the degree of anomaly in different categories; S8. Determine the insulation anomaly based on the differences in the classification characteristics of class k.

2. The method for detecting insulation abnormalities in a new energy vehicle power battery system according to claim 1, characterized in that, S2 includes: Delete invalid data; Outliers are handled by using a sliding median to smooth out the battery signal data to some extent. Delete invalid battery signal data; The battery signal data is normalized.

3. The method for detecting insulation abnormalities in a new energy vehicle power battery system according to claim 1, characterized in that, S4 includes: for each insulation resistance, first calculating the first k modes of each insulation resistance, then calculating the modes of voltage, current, and temperature under that insulation resistance value, thereby obtaining k points as the initial values ​​for the k-means clustering algorithm, denoted as:

4. The method for detecting insulation abnormalities in a new energy vehicle power battery system according to claim 3, characterized in that, The S5 includes: S501, the initial k center points are the k initial values ​​selected in S4; S502. Define the loss function as the sum of squared errors between each sample point and the center point of its class: Where x i For the i-th sample, c i It is x i Category to which it belongs The center point is the center point corresponding to the category to which it belongs, and M is the total number of samples; S503. Calculate the distance between each sample point and the sample center of each category, and classify it into the category with the closest sample center. S504. For all the sample points that have been classified, calculate new cluster centers according to the categories and use them as new sample centers. S505. Determine whether the loss function has converged. If yes, proceed to S506; otherwise, proceed to S503. S506, Stop the iteration and obtain the clustering results.

5. The method for detecting insulation abnormalities in a new energy vehicle power battery system according to claim 4, characterized in that, The contour coefficient is: in This represents the intra-category dissimilarity, where n is the number of sample points within the current category. b(i)=min{b1(i),b2(i),…,b m (i)}, where The dissimilarity between categories is represented by n, where n is the number of sample points within category l. The average silhouette coefficient of each sample point is the silhouette coefficient S of the clustering result. The silhouette coefficient ranges from -1 to 1. The closer it is to 1, the better the clustering and separation. The silhouette coefficient is: in This represents the intra-category dissimilarity, where n is the number of sample points within the current category. b(i)=min{b1(i),b2(i),…,b m (i)}, where The dissimilarity between categories is represented by n, where n is the number of sample points within category l. The average silhouette coefficient of each sample point is the silhouette coefficient S of the clustering result. The silhouette coefficient ranges from -1 to 1. The closer it is to 1, the better the clustering degree and separation degree are, and the better the clustering effect is.

Images