A method for detecting and protecting abnormal temperature of a battery module

Abstract

The present invention relates to the technical field of data anomaly detection, and more specifically to a method for detecting and protecting temperature anomalies in a battery module. The method comprises obtaining a recent temperature sequence based on data changes in a temperature history sequence; obtaining a distance weight based on the position of a recent data point in the recent temperature sequence; obtaining a first weight adjustment based on the distance weight and the temperature change; obtaining a correction coefficient and a second weight adjustment based on data distribution characteristics within a neighborhood of the recent data point; and obtaining a target weight value based on the distance weight, the first weight adjustment, the second weight adjustment, and the correction coefficient. The present invention obtains a predicted temperature value using a sliding average method based on recent data points and corresponding target weight values; and performs temperature anomaly detection based on current data points and the corresponding predicted temperature value, thereby improving the accuracy of temperature prediction and anomaly detection.

Description

Technical Field

[0001] The present invention relates to the technical field of data anomaly detection, and in particular to a temperature anomaly detection and protection method for a battery module. Background Art

[0002] With the increasing popularity of electric vehicles and energy storage devices, battery modules, as their core components, are becoming increasingly important. Battery module safety is crucial. Extended use or high-load operation can lead to abnormally high internal temperatures, which can lead to performance degradation, shortened lifespan, and even safety incidents. Therefore, to ensure the proper operation of battery modules, real-time temperature monitoring is required during battery model operation.

[0003] Traditional battery module temperature anomaly detection is typically based on a threshold method. When the temperature exceeds the threshold, it is determined to be a temperature anomaly. However, because temperature change is a continuous process, a single threshold is insufficient for determining temperature anomalies. To make detection more comprehensive and accurate, anomaly detection can be performed based on a prediction method, where the difference between the predicted value and the actual value is used to determine the temperature anomaly. Traditional data prediction typically uses a sliding average method, which performs a weighted summation of recent temperature data to obtain a predicted temperature value. However, this algorithm is insensitive to outliers in recent data and has limited processing capabilities for nonlinear data changes, resulting in inaccurate temperature prediction results and low accuracy in battery module temperature anomaly detection. Summary of the Invention

[0004] In order to solve the technical problem that the sliding average method is not accurate enough in predicting temperature data, which affects the accuracy of temperature anomaly detection, the present invention aims to provide a temperature anomaly detection and protection method for a battery module. The technical solution adopted is as follows:

[0005] Acquire a temperature history sequence for detecting battery temperature; obtain a temperature change cycle length based on data change characteristics of the temperature history sequence; and obtain a recent temperature sequence of a current data point in the temperature history sequence based on the temperature change cycle length;

[0006] Obtaining a distance weight of a recent data point based on a time interval characteristic between a recent data point and the current data point in the recent temperature sequence; obtaining a temperature change based on a data difference characteristic between the recent data point and a connected data point; and obtaining a first weight adjustment based on the distance weight of the recent data point and the temperature change;

[0007] Obtaining data credibility and a correction coefficient based on data distribution characteristics within a certain time period of the recent data point; obtaining a second weight adjustment amount for the recent data point based on the data credibility, the first weight adjustment amount, and the distance weight; and obtaining a target weight value based on the distance weight, the first weight adjustment amount, the second weight adjustment amount, and the correction coefficient of the recent data point;

[0008] A predicted temperature value of a current data point is obtained by a sliding average method based on recent data points in the recent temperature sequence and corresponding target weight values; and temperature anomaly detection is performed based on the current data point and the corresponding predicted temperature value.

[0009] Furthermore, the step of obtaining the temperature change cycle length according to the data change characteristics of the temperature history sequence includes:

[0010] A preset initial number of data points is taken from the first bit of the temperature history sequence as a first subsequence; a preset initial number of data points is taken from the first bit connected to the end of the first subsequence in the temperature history sequence as a second subsequence; the Pearson correlation coefficient of the first subsequence and the second subsequence is calculated to obtain the periodic characteristic values ​​of the preset initial number, and the preset initial number corresponding to the maximum value of the periodic characteristic value is used as the temperature change period length.

[0011] Furthermore, the step of obtaining a recent temperature sequence of a current data point in the temperature history sequence according to the temperature change cycle length includes:

[0012] A temperature history sequence segment having a temperature change cycle length closest to the current moment is obtained from the temperature history sequence as a recent temperature sequence of the current data point.

[0013] Furthermore, the step of obtaining the distance weight of the recent data point according to the time interval feature between the recent data point and the current data point in the recent temperature sequence includes:

[0014] The inverse of the time interval between the recent data point and the current data point is calculated to obtain a weight coefficient; the sum of the weight coefficients of all recent data points in the recent temperature sequence is calculated to obtain a comprehensive weight coefficient; and the ratio of the weight coefficient to the comprehensive weight coefficient is calculated to obtain a distance weight of the recent data point.

[0015] Furthermore, the step of obtaining the temperature change according to the data difference characteristics between the recent data point and the connected data points includes:

[0016] The absolute value of the temperature difference between the recent data point and the connected data point at the previous moment is calculated to obtain the temperature change of the recent data point.

[0017] Furthermore, the step of obtaining a first weight adjustment amount according to the distance weight and temperature change of the recent data point includes:

[0018] When the temperature change does not exceed the preset change threshold, the first weight adjustment amount of the recent data point is a constant 0;

[0019] When the temperature change exceeds a preset change threshold, the difference between the temperature change and the preset change threshold is calculated and normalized to obtain a change amplitude; the product of the change amplitude and the distance weight is calculated to obtain a first weight adjustment amount of the recent data point.

[0020] Furthermore, the step of obtaining data credibility and correction coefficient according to the data distribution characteristics within the adjacent time of the recent data point includes:

[0021] In the temperature history sequence, a window range of a preset window length of the recent data point closest to the window center is used as a preset neighborhood window of the recent data point; for any window data point within the preset neighborhood window, the temperature difference between the arbitrary window data point and the connected data point at the previous moment is calculated to obtain the connected temperature difference value of the arbitrary window data point; the absolute value of the sum of all connected temperature difference values ​​within the preset neighborhood window is calculated to obtain the data credibility of the recent data point;

[0022] When the data credibility does not exceed the preset difference threshold, the correction coefficient of the recent data point is a constant of 0; when the data credibility exceeds the preset difference threshold, the correction coefficient of the recent data point is a constant of 1.

[0023] Furthermore, the step of obtaining the second weight adjustment amount of the recent data point according to the data credibility, the first weight adjustment amount and the distance weight includes:

[0024] When the data credibility does not exceed the preset difference threshold, the second weight adjustment amount of the recent data point is a constant 0;

[0025] When the data credibility exceeds a preset difference threshold, the difference between the data credibility and the preset difference threshold is calculated and normalized to obtain the adjustment factor of the recent data point; the difference between the distance weight of the recent data point and the first weight adjustment amount is calculated to obtain the adjustment benchmark; the product of the adjustment benchmark and the adjustment factor is calculated to obtain the second weight adjustment amount of the recent data point.

[0026] Furthermore, the step of obtaining a target weight value according to the distance weight of the recent data point, the first weight adjustment amount, the second weight adjustment amount and the correction coefficient includes:

[0027] Calculating the product of the first weight adjustment amount of the recent data point and the correction coefficient to obtain a first weight correction value; calculating the sum of the distance weight of the recent data point, the first weight correction value, and the second weight adjustment amount to obtain a comprehensive weight value of the recent data point;

[0028] Calculate the sum of the comprehensive weight values ​​of all recent data points corresponding to the current data point to obtain the recent total weight; calculate the ratio of the comprehensive weight value to the recent total weight to obtain the target weight value of the recent data point.

[0029] Furthermore, the step of performing temperature anomaly detection based on the current data point and the corresponding predicted temperature value includes:

[0030] The first condition is that the temperature value of the current data point is not within the preset temperature range; the second condition is that the predicted temperature value corresponding to the current data point is not within the preset temperature range; the third condition is that the absolute value of the difference between the temperature value of the current data point and the corresponding predicted temperature value exceeds the preset warning value; when any of the first condition, the second condition and the third condition are met, the temperature state of the current data point is abnormal.

[0031] The present invention has the following beneficial effects:

[0032] In the present invention, obtaining the temperature variation cycle can determine the length of recent data in the sliding average method, thereby determining the recent temperature sequence based on the temperature variation cycle length, initially improving the accuracy of the predicted temperature value. Obtaining the distance weight can make the temperature prediction value more consistent with recent temperature characteristics, improving prediction accuracy. Obtaining the temperature variation can characterize the temperature variation trend of the battery module at consecutive moments during operation. Obtaining the first weight adjustment value can adjust the distance weight based on the temperature variation trend, further improving the accuracy of the weight of the recent data point. Obtaining the data credibility and correction coefficient can determine the cause of temperature data fluctuation based on the temperature variation trend within the adjacent time period of the recent data point, thereby more accurately adjusting the distance weight and avoiding errors in the first weight adjustment value. Obtaining the second weight adjustment value can further adjust the distance weight based on the adjacent temperature characteristics of the recent data point, making the adjusted weight of the recent data point more accurate and reducing prediction error. Obtaining the target weight value can adaptively adjust the weight of the recent data point based on the temperature variation characteristics of the recent data point, improving the prediction accuracy of the sliding average method, reducing errors in the predicted temperature value, and ultimately improving the accuracy of temperature anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 A flow chart of a method for detecting and protecting temperature anomalies of a battery module provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0035] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a battery module temperature anomaly detection and protection method proposed by the present invention. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0036] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0037] The following describes in detail a specific solution of a temperature anomaly detection and protection method for a battery module provided by the present invention with reference to the accompanying drawings.

[0038] See also Figure 1 , which shows a flow chart of a temperature anomaly detection and protection method for a battery module provided by one embodiment of the present invention, the method comprising the following steps:

[0039] Step S1, obtaining a temperature history sequence for detecting battery temperature; obtaining a temperature change cycle length according to data change characteristics of the temperature history sequence; and obtaining a recent temperature sequence of a current data point in the temperature history sequence according to the temperature change cycle length.

[0040] In an embodiment of the present invention, the implementation scenario is to perform abnormality detection on the temperature of the battery module to improve the accuracy of abnormality detection; first, a temperature history sequence for detecting the battery temperature is obtained, and the temperature history sequence is obtained according to a preset acquisition frequency through the temperature sensor built into the battery module. In an embodiment of the present invention, the preset acquisition frequency is to collect temperature values ​​120 times per minute, and the temperature history sequence is temperature data within half an hour from the current moment. As time changes, the data in the temperature history sequence continues to change, but the data length is fixed; the implementer can determine the preset acquisition frequency and the length of the temperature history sequence according to the implementation scenario.

[0041] Using a sliding average method to predict the current temperature value, and then detecting anomalies based on the difference between the predicted and actual temperature values, this method is more accurate and comprehensive than using fixed thresholds. Existing sliding average methods assign different weights to recent data based on their proximity to the current time. Closer distances result in greater weights, and the predicted value is more similar to the values ​​of recent data points. First, determine the range of recent data within the sliding average method. Too little or too much recent data can lead to biased predictions, thus affecting detection accuracy. Battery temperature exhibits a certain cyclical nature during operation. After the temperature gradually rises, the temperature control system activates, gradually lowering the temperature. Once the temperature drops to a certain level, the temperature control system deactivates, and the temperature gradually rises again. Therefore, the battery module's temperature exhibits cyclical variations during operation. The temperature variation characteristics of each cycle are similar, and the data variation characteristics within a cycle can fully reflect the temperature trend. Using a sliding average method based on the length of the most recent cycle at the current time can initially improve prediction accuracy. Therefore, the temperature variation cycle length is determined based on the variation characteristics of the temperature history series.

[0042] Preferably, in an embodiment of the present invention, the step of obtaining the length of the temperature change cycle includes: taking a preset initial number of data points starting from the first bit of the temperature history sequence as a first subsequence; taking a preset initial number of data points starting from the first bit connected to the end of the first subsequence in the temperature history sequence as a second subsequence; for example, when the preset initial number is 10, the first subsequence is the first to tenth bits of the temperature history sequence, and the second subsequence is the eleventh to twentieth bits of the temperature history sequence; in an embodiment of the present invention, the value range of the preset initial number is 100 to 1000, and the implementer can determine it according to the change characteristics of the battery temperature in the implementation scenario. The Pearson correlation coefficient of the first subsequence and the second subsequence is calculated to obtain a preset initial number of periodic characteristic values. It should be noted that the Pearson correlation coefficient belongs to the existing technology, and the specific calculation steps will not be repeated here. The Pearson correlation coefficient can reflect the correlation characteristics between the two data sequences, and the value range is -1 to 1. When the Pearson correlation coefficient is closer to 1, it means that the change characteristics of the two data sequences are more similar. Therefore, when the periodic characteristic value is closer to 1, it means that the first subsequence and the second subsequence are more likely to represent two connected temperature change cycles, and the corresponding preset initial number is more likely to be the cycle length of the temperature change. The preset initial number corresponding to the maximum value of the periodic characteristic value is then used as the temperature change cycle length. It should be noted that since the battery working scenario is relatively fixed, the temperature change cycle length only needs to be calculated once, and there is no need to cyclically calculate the temperature change cycle length based on the data update of the temperature history sequence.

[0043] Furthermore, after obtaining the temperature variation cycle length, a recent temperature sequence for the current data point can be obtained from the temperature history sequence based on the temperature variation cycle length. Specifically, this includes obtaining the temperature history sequence segment with the temperature variation cycle length closest to the current moment in the temperature history sequence as the recent temperature sequence for the current data point. For example, if the temperature variation cycle length is 200, the 200 most recent temperature data points from the temperature history sequence are obtained as the recent temperature sequence for the current data point. Predicting the predicted value for the current data point using the sliding average method based on this recent temperature sequence can initially improve prediction accuracy.

[0044] Step S2: obtaining the distance weight of the recent data point based on the time interval characteristics between the recent data point and the current data point in the recent temperature sequence; obtaining the temperature change based on the data difference characteristics between the recent data point and the connected data points; and obtaining the first weight adjustment amount based on the distance weight and temperature change of the recent data point.

[0045] When the time interval between the recent data point and the current data point in the recent temperature series is shorter, the temperature change between the recent data point and the current data point is more correlated, and the closer the prediction result is to the value of the recent data point, the greater the prediction weight of the recent data point. Therefore, the distance weight of the recent data point is obtained according to the time interval characteristics between the recent data point and the current data point in the recent temperature series. Preferably, in one embodiment of the present invention, the step of obtaining the distance weight includes: calculating the inverse of the time interval length between the recent data point and the current data point to obtain the weight coefficient; the time interval length is the number of intervals between the timestamps corresponding to the recent data point and the current data point. When the recent data point is closer to the acquisition time of the current data point, the shorter the time interval length, the larger the weight coefficient, and the higher the contribution of the recent data point in the prediction process. Calculate the sum of the weight coefficients of all recent data points in the recent temperature series to obtain the comprehensive weight coefficient; calculate the ratio of the weight coefficient to the comprehensive weight coefficient to obtain the distance weight of the recent data point, so that the sum of the distance weights of all recent data points is 1.

[0046] Furthermore, when the temperature of the battery module is changing steadily, the distance weight obtained based only on the time interval characteristics between the recent data point and the current data point can be used to predict more accurately. However, when the working state of the battery module is abnormal, the temperature may fluctuate more significantly. In order to more accurately reflect the abnormal fluctuation during prediction, it is necessary to assign a greater weight to the temperature data with abnormal fluctuations and adjust the distance weight. Therefore, the temperature change is obtained based on the data difference characteristics between the recent data point and the connected data point. Preferably, in an embodiment of the present invention, the step of obtaining the temperature change includes: calculating the absolute value of the temperature difference between the recent data point and the connected data point at the previous moment, and obtaining the temperature change of the recent data point. When the temperature change is more obvious and the temperature change is larger, it means that the temperature characteristics of the recent data point have changed more significantly. In this case, a greater weight should be assigned to the recent data point so that the predicted value can better reflect the temperature fluctuation characteristics; and then a first weight adjustment amount is obtained based on the distance weight and temperature change of the recent data point.

[0047] Preferably, in one embodiment of the present invention, the step of obtaining a first weight adjustment includes: when the temperature change does not exceed a preset change threshold, it means that the temperature change trend is relatively stable and no obvious abnormal fluctuations have occurred. There is no need to adjust the distance weight of the recent data point, and the first weight adjustment of the recent data point is a constant of 0; in the embodiment of the present invention, the preset change threshold is 2.5, which can be determined by the implementer according to the implementation scenario. When the temperature change exceeds the preset change threshold, it means that the temperature has fluctuated significantly, and the distance weight needs to be adjusted. The difference between the temperature change and the preset change threshold is calculated and normalized to obtain the change amplitude; the greater the temperature change, the greater the degree of fluctuation. The greater the change amplitude, the greater the need for a larger adjustment to the weight of the recent data point, thereby highlighting the fluctuation characteristics of the recent data point. The product of the change amplitude and the distance weight is calculated to obtain the first weight adjustment of the recent data point; the product of the change amplitude and the distance weight can ensure that the first weight adjustment does not exceed the distance weight of the recent data point, avoiding excessive adjustment resulting in large errors in the prediction results; the greater the change amplitude, the greater the adjustment amount, and the greater the first weight adjustment amount; the formula for obtaining the first weight adjustment includes:

[0048] S=L*{1-exp[-(RN)]}

[0049] Where S represents the first weight adjustment of the recent data point, L represents the distance weight of the recent data point, R represents the temperature change of the recent data point, N represents the preset change threshold, exp() represents an exponential function with a natural constant as the base, and 1-exp[-(RN)] represents the normalization of the difference between the temperature change and the preset change threshold to obtain the change amplitude.

[0050] Step S3, obtain data credibility and correction coefficient according to the data distribution characteristics in the adjacent time of the recent data points; obtain the second weight adjustment amount of the recent data points according to the data credibility, the first weight adjustment amount and the distance weight; obtain the target weight value according to the distance weight, the first weight adjustment amount, the second weight adjustment amount and the correction coefficient of the recent data points.

[0051] Since the first weight adjustment amount is only obtained based on the difference characteristics between the recent data point and the temperature at the previous moment, noise may occur during the temperature data collection process, resulting in inaccurate temperature values. Therefore, it is necessary to analyze the temperature change trend within the local time range where the recent data point is located, so as to obtain data credibility and correction coefficient based on the data distribution characteristics in the adjacent time of the recent data point.

[0052] Preferably, in one embodiment of the present invention, the step of obtaining data credibility and correction coefficient includes: in the temperature history sequence, taking the window range of the preset window length of the recent data point closest to the window center point as the preset neighborhood window of the recent data point; in the embodiment of the present invention, the preset window length is 21, and the implementer can determine it according to the implementation scenario; if the number of data points before and after the recent data point meets 10, then the recent data point is at the center of the preset neighborhood window; if the number of data points before and after the recent data point does not meet 10, for example, there are only 5 data points at the subsequent moment of the recent data point, then the recent data point constructs a preset neighborhood window at the position closest to the window center, and there are 15 data points in front of the recent data point and 5 data points behind it in the preset neighborhood window. For any window data point within the preset neighborhood window, calculate the temperature difference between the arbitrary window data point and the connected data point at the previous moment to obtain the connected temperature difference value of the arbitrary window data point; calculate the absolute value of the sum of all connected temperature difference values ​​within the preset neighborhood window to obtain the data credibility of the recent data point; for example, there are 5 data points in the preset neighborhood window of the recent data point, and the recent data point is in the center of the window. When the corresponding connected temperature difference values ​​are 1.8, 2, 2.1, 1.6, and 2.2 respectively, the data temperature is in a state of obvious rise, the result after summing up the five values ​​is large, and the obtained data credibility is large; if the corresponding connected temperature difference values ​​are 0.2, 0.1, 4.8, -5.1, and 0.1 respectively, the temperature value of the third data point has risen sharply, but the temperature value of the fourth data point has dropped significantly, the third data point may be noise data, the result after summing up the five data tends to 0, and the data credibility tends to zero.

[0053] When the data credibility does not exceed the preset difference threshold, the correction coefficient of the recent data point is a constant 0. Noise data appears in the preset window of the recent data point, resulting in a large temperature change. In this case, there is no need to use the first weight adjustment to adjust the distance weight. When the data credibility exceeds the preset difference threshold, the correction coefficient of the recent data point is a constant 1, which means that the temperature has fluctuated significantly and the first weight adjustment needs to be used to adjust the distance weight. In the embodiment of the present invention, the preset difference threshold is 2, which can be determined by the implementer according to the implementation scenario.

[0054] The greater the data credibility, the more it can reflect the temperature change trend and the obvious temperature rise or fall trend in the adjacent time of the recent data point. Because the first weight adjustment amount only adjusts the distance weight according to the temperature change of the adjacent data points, without referring to the temperature fluctuation characteristics within a period of time, in order to further accurately correct the distance weight and highlight the abnormal temperature fluctuation characteristics during the prediction, it is necessary to further adjust the distance weight according to the data credibility obtained from the temperature fluctuation characteristics. Therefore, the second weight adjustment amount of the recent data point is obtained according to the data credibility, the first weight adjustment amount and the distance weight.

[0055] Preferably, in an embodiment of the present invention, the step of obtaining the second weight adjustment includes: when the data credibility does not exceed a preset difference threshold, the second weight adjustment for the recent data point is constant at 0. When the data credibility does not exceed the preset difference threshold, this indicates that the temperature fluctuation characteristics within the neighborhood of the recent data point are caused by noise data, and therefore, no distance weight adjustment is required. When the data credibility exceeds the preset difference threshold, the difference between the data credibility and the preset difference threshold is calculated and normalized to obtain an adjustment factor for the recent data point. The greater the data credibility of the recent data point, the more significant the temperature fluctuation within the neighborhood of the recent data point, and thus the larger the adjustment factor. The difference between the distance weight of the recent data point and the first weight adjustment is calculated to obtain an adjustment benchmark. Obtaining the adjustment benchmark can determine the adjustment range of the second weight adjustment, preventing the second weight adjustment from exceeding the difference between the distance weight and the first weight adjustment, thereby maintaining the adjustment within an appropriate range and improving prediction accuracy. The product of the adjustment benchmark and the adjustment factor is calculated to obtain the second weight adjustment for the recent data point. The greater the data credibility, the more significant the fluctuation, and thus the larger the second weight adjustment.

[0056] Furthermore, the target weight value can be obtained based on the distance weight, first weight adjustment amount, second weight adjustment amount and correction coefficient of the recent data point; preferably, in an embodiment of the present invention, the step of obtaining the target weight value includes: calculating the product of the first weight adjustment amount and the correction coefficient of the recent data point to obtain the first weight correction value; when the correction coefficient is 0, the first weight correction value is 0, which means that the first weight adjustment amount is calculated from noise data, so its value is corrected to 0; when the correction coefficient is 1, the first weight correction value is the first weight adjustment amount. Calculate the sum of the distance weight, the first weight correction value and the second weight adjustment amount of the recent data point to obtain the comprehensive weight value of the recent data point; when the fluctuation characteristics of the neighborhood data of the recent data point are more obvious, the more it is necessary to highlight the fluctuation characteristics during prediction, and the larger the comprehensive weight value. The step of obtaining the comprehensive weight value includes:

[0057] W=A+S*h+E

[0058] Where W represents the comprehensive weight value of the recent data point, A represents the distance weight of the recent data point, S represents the first weight adjustment of the recent data point, h represents the correction coefficient of the recent data point, S*h represents the first weight correction value of the recent data point, and E represents the second weight adjustment of the recent data point.

[0059] Calculate the sum of the comprehensive weights of all recent data points corresponding to the current data point to obtain the recent total weight. Calculate the ratio of the comprehensive weight to the recent total weight to obtain the target weight for the recent data point. Obtaining the corresponding target weight based on the comprehensive weight ensures that the sum of the target weights of all recent data points corresponding to the current data point is 1, thus meeting the weight setting requirements of the sliding average method.

[0060] Step S4, obtaining a predicted temperature value of the current data point by a sliding average method based on recent data points in the recent temperature sequence and corresponding target weight values; performing temperature anomaly detection based on the current data point and the corresponding predicted temperature value.

[0061] After obtaining the predicted temperature values ​​of all recent data points of the current data point, the predicted temperature value of the current data point can be obtained by using the sliding average method based on the recent data points in the recent temperature sequence and the corresponding target weight values. The temperature prediction value is obtained by weighted summing the temperature values ​​of the recent data points and the corresponding target weight values. It should be noted that the sliding average method belongs to the existing technology and the specific calculation steps are not repeated here. Compared with the traditional sliding average method, the embodiment of the present invention adaptively adjusts the weight according to the temperature change area and abnormal fluctuation characteristics, making the temperature prediction value more accurate. Furthermore, temperature anomaly detection can be performed based on the current data point and the corresponding predicted temperature value. Preferably, in the embodiment of the present invention, the step of performing temperature anomaly detection includes: taking the temperature value of the current data point not being within the preset temperature range as a first condition; taking the predicted temperature value corresponding to the current data point not being within the preset temperature range as a second condition; taking the absolute value of the difference between the temperature value of the current data point and the corresponding predicted temperature value exceeding the preset warning value as a third condition; when any of the first condition, the second condition, and the third condition are met, the temperature state of the current data point is abnormal, making the temperature anomaly detection more comprehensive and accurate. The preset temperature range and preset warning value can be determined by the implementer based on the implementation scenario and are not limited here. It should be noted that if the historical data length of the current data point does not meet the prediction requirements of the sliding average method, only the first condition will be used for temperature anomaly detection; battery maintenance will be performed based on the real-time anomaly detection results.

[0062] In summary, an embodiment of the present invention provides a method for detecting and protecting temperature anomalies in a battery module. A recent temperature sequence is obtained based on data changes in a temperature history sequence. A distance weight is obtained based on the position of a recent data point in the recent temperature sequence. A first weight adjustment is obtained based on the distance weight and the temperature change. A correction coefficient and a second weight adjustment are obtained based on data distribution characteristics within the neighborhood of the recent data point. A target weight value is obtained based on the distance weight, the first weight adjustment, the second weight adjustment, and the correction coefficient. The present invention obtains a predicted temperature value using a sliding average method based on recent data points and corresponding target weight values. Temperature anomaly detection is performed based on the current data point and the corresponding predicted temperature value, thereby improving the accuracy of temperature prediction and anomaly detection.

[0063] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0064] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for detecting and protecting temperature anomalies of a battery module, characterized in that: The method comprises the following steps: Acquire a temperature history sequence for detecting battery temperature; obtain a temperature change cycle length based on data change characteristics of the temperature history sequence; and obtain a recent temperature sequence of a current data point in the temperature history sequence based on the temperature change cycle length; Obtaining a distance weight of the recent data point based on a time interval feature between the recent data point and the current data point in the recent temperature sequence; obtaining a temperature change based on a data difference feature between the recent data point and the connected data points; Obtaining a first weight adjustment amount according to the distance weight and temperature change of the recent data point; Obtaining data credibility and a correction coefficient based on data distribution characteristics within a certain time period of the recent data point; obtaining a second weight adjustment amount for the recent data point based on the data credibility, the first weight adjustment amount, and the distance weight; and obtaining a target weight value based on the distance weight, the first weight adjustment amount, the second weight adjustment amount, and the correction coefficient of the recent data point; A predicted temperature value of a current data point is obtained by a sliding average method based on recent data points in the recent temperature sequence and corresponding target weight values; and temperature anomaly detection is performed based on the current data point and the corresponding predicted temperature value.

2. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining the temperature change period length according to the data change characteristics of the temperature history sequence includes: A preset initial number of data points is taken from the first bit of the temperature history sequence as a first subsequence; a preset initial number of data points is taken from the first bit connected to the end of the first subsequence in the temperature history sequence as a second subsequence; the Pearson correlation coefficient of the first subsequence and the second subsequence is calculated to obtain the periodic characteristic values ​​of the preset initial number, and the preset initial number corresponding to the maximum value of the periodic characteristic value is used as the temperature change period length.

3. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining a recent temperature sequence of a current data point in the temperature history sequence according to the temperature change cycle length includes: A temperature history sequence segment having a temperature change cycle length closest to the current moment is obtained from the temperature history sequence as a recent temperature sequence of the current data point.

4. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining the distance weight of the recent data point according to the time interval feature between the recent data point and the current data point in the recent temperature sequence comprises: The inverse of the time interval between the recent data point and the current data point is calculated to obtain a weight coefficient; the sum of the weight coefficients of all recent data points in the recent temperature sequence is calculated to obtain a comprehensive weight coefficient; and the ratio of the weight coefficient to the comprehensive weight coefficient is calculated to obtain a distance weight of the recent data point.

5. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining the temperature change according to the data difference characteristics between the recent data point and the connected data points includes: The absolute value of the temperature difference between the recent data point and the connected data point at the previous moment is calculated to obtain the temperature change of the recent data point.

6. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining a first weight adjustment value according to the distance weight and temperature change of the recent data point comprises: When the temperature change does not exceed the preset change threshold, the first weight adjustment amount of the recent data point is a constant 0; When the temperature change exceeds a preset change threshold, the difference between the temperature change and the preset change threshold is calculated and normalized to obtain a change amplitude; the product of the change amplitude and the distance weight is calculated to obtain a first weight adjustment amount of the recent data point.

7. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining data credibility and correction coefficient according to the data distribution characteristics within the adjacent time of the recent data point includes: In the temperature history sequence, a window range of a preset window length of the recent data point closest to the window center is used as a preset neighborhood window of the recent data point; for any window data point within the preset neighborhood window, the temperature difference between the arbitrary window data point and the connected data point at the previous moment is calculated to obtain the connected temperature difference value of the arbitrary window data point; the absolute value of the sum of all connected temperature difference values ​​within the preset neighborhood window is calculated to obtain the data credibility of the recent data point; When the data credibility does not exceed the preset difference threshold, the correction coefficient of the recent data point is a constant of 0; when the data credibility exceeds the preset difference threshold, the correction coefficient of the recent data point is a constant of 1.

8. The method for detecting and protecting a battery module from temperature anomaly according to claim 7, wherein: The step of obtaining the second weight adjustment amount of the recent data point according to the data credibility, the first weight adjustment amount and the distance weight includes: When the data credibility does not exceed the preset difference threshold, the second weight adjustment amount of the recent data point is a constant 0; When the data credibility exceeds a preset difference threshold, the difference between the data credibility and the preset difference threshold is calculated and normalized to obtain the adjustment factor of the recent data point; the difference between the distance weight of the recent data point and the first weight adjustment amount is calculated to obtain the adjustment benchmark; the product of the adjustment benchmark and the adjustment factor is calculated to obtain the second weight adjustment amount of the recent data point.

9. The method for detecting and protecting a battery module from temperature anomaly according to claim 1, wherein: The step of obtaining a target weight value according to the distance weight of the recent data point, the first weight adjustment amount, the second weight adjustment amount and the correction coefficient comprises: Calculating the product of the first weight adjustment amount of the recent data point and the correction coefficient to obtain a first weight correction value; calculating the sum of the distance weight of the recent data point, the first weight correction value, and the second weight adjustment amount to obtain a comprehensive weight value of the recent data point; Calculate the sum of the comprehensive weight values ​​of all recent data points corresponding to the current data point to obtain the recent total weight; calculate the ratio of the comprehensive weight value to the recent total weight to obtain the target weight value of the recent data point.

10. The method for detecting and protecting battery module temperature anomaly according to claim 1, wherein: The step of performing temperature anomaly detection based on the current data point and the corresponding predicted temperature value includes: The first condition is that the temperature value of the current data point is not within the preset temperature range; the second condition is that the predicted temperature value corresponding to the current data point is not within the preset temperature range; the third condition is that the absolute value of the difference between the temperature value of the current data point and the corresponding predicted temperature value exceeds the preset warning value; when any of the first condition, the second condition and the third condition are met, the temperature state of the current data point is abnormal.

Images