Lithium battery envelope feature extraction method for random use data

Abstract

The invention belongs to the technical field of lithium battery health management and state evaluation, and particularly relates to a lithium battery envelope feature extraction method for random use data, which comprises the following steps: S1, acquiring an original sampling sequence of each key feature of a target battery; the key features comprise an accumulated usage amount feature, a health state feature, a working condition state feature and a time sequence feature; the original sampling sequence comprises a cumulative usage amount time sequence, a health state time sequence, a working condition state time sequence and a corresponding time index sequence, and the time index sequence is used for identifying or recording the occurrence sequence or moment of the original sampling sequence in time and recording the time point of data acquisition; and S2, constructing a multi-dimensional feature vector F by using the obtained original sampling sequence. According to the method, the accuracy of residual life prediction and state evaluation of the lithium battery can be effectively improved through collection, analysis and processing of multi-dimensional data such as the accumulative usage amount, the health state and the working condition state of the lithium battery.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium battery health management and status assessment, and in particular relates to a lithium battery envelope feature extraction method for random usage data. Background Art

[0002] With the widespread application of lithium batteries in electric vehicles, energy storage systems and other fields, how to accurately evaluate the battery's state of health (State of Health, SOH) and predict its remaining useful life (Remaining Useful Life, RUL) has become a key issue. Most traditional battery state assessment methods rely on regular sampling data or data collection under ideal working conditions, which is difficult to cope with complex and irregular actual use environments. That is, existing envelope methods are usually mainly used to process time-frequency signals, such as speech signals, vibration signals, and biological signals. These methods help analyze the frequency components and change trends of signals by extracting the envelope characteristics of signals. However, in application scenarios such as electric vehicles and energy storage systems, lithium batteries are often in random working conditions, which makes the assessment method based on fixed conditions unable to accurately reflect the true state of the battery. Therefore, it is of great significance to develop a feature extraction method that can adapt to complex working conditions and random usage data.

[0003] In particular, as the most commonly used energy storage solution, the performance of lithium batteries directly affects the mileage of electric vehicles, the operating efficiency and safety of energy storage equipment. Therefore, effective methods for health management and life prediction of lithium batteries are particularly important. A technology is needed to randomly sample data under random working conditions to ensure accurate evaluation of the status of lithium batteries, thereby improving their safety and economy. Summary of the invention

[0004] The present invention provides a lithium battery envelope feature extraction method for random usage data. The method can effectively improve the accuracy of lithium battery remaining life prediction and status assessment by collecting, analyzing and processing multi-dimensional data such as the cumulative usage, health status and operating status of the lithium battery.

[0005] In order to solve the above technical problems, this application provides the following technical solutions:

[0006] A lithium battery envelope feature extraction method for random usage data comprises the following steps:

[0007] S1: Obtaining the original sampling sequence of each key feature of the target battery; the key features include cumulative usage features, health status features, operating status features, and time series features; the original sampling sequence includes a cumulative usage time series, a health status time series, an operating status time series, and a corresponding time index sequence, the time index sequence is used to identify or record the order or moment of the original sampling sequence in time, and record the time point of data collection;

[0008] S2: construct a multidimensional feature vector F using the acquired original sampling sequence; calculate the point distance between the corresponding value of the multidimensional feature vector F at each moment and the coordinate origin to generate a numerical distance feature vector D;

[0009] S3: Perform envelope processing on the obtained distance feature vector D to obtain a discrete envelope distance value and determine the corresponding discrete envelope moment;

[0010] S4: obtaining the order of discrete envelope moments in the time index sequence, sampling from the original sampling sequence of each key feature, and constructing the envelope sampling sequence of each key feature;

[0011] S5: construct an interpolation function according to the relationship between the envelope sampling sequence of each key feature and the discrete envelope moment; calculate the characteristic value of each key feature at the original time index through the interpolation function, and generate a continuous interpolation repair sequence of each key feature.

[0012] The basic principles and beneficial effects of this solution are as follows: First, key feature data is collected from the operation of the lithium battery, including cumulative usage, health status, operating status and timing characteristics, to form time series data. The collected data is used to construct a multidimensional feature vector, and then the point distance between the value of the vector at each moment and the coordinate origin is calculated to generate a distance feature vector. The distance feature vector is enveloped to extract the discrete envelope distance value and the corresponding discrete envelope moment. According to the order of the discrete envelope moments in the time index sequence, key feature values ​​are extracted from the original sampling sequence to construct an envelope sampling sequence. The relationship between the envelope sampling sequence and the discrete envelope moments is used to construct an interpolation function, which is used to calculate the feature values ​​on the original time index to generate a continuous interpolation repair sequence.

[0013] This solution can improve accuracy: through envelope processing and interpolation functions, the key features of lithium batteries can be extracted and repaired more accurately, improving the accuracy of feature extraction.

[0014] Enhanced robustness: This method can handle random usage data, so that feature extraction is not restricted by specific working conditions or usage patterns, which enhances the robustness of the method. Real-time monitoring and prediction: Through continuous interpolation repair sequences, real-time monitoring and prediction of the health status of lithium batteries can be achieved, which is of great significance for battery maintenance and life prediction. Improved data utilization: This method can effectively utilize the collected data, and improves the utilization of data and the integrity of information extraction by constructing multi-dimensional feature vectors and envelope sampling sequences. Improved safety: Real-time monitoring of the health status of the battery is very necessary for the safe operation of the battery system. This method helps to promptly detect potential problems with the battery and prevent the occurrence of safety accidents. Reduced maintenance costs: Through accurate assessment of the battery health status, battery maintenance and replacement can be arranged more reasonably, reducing maintenance costs and avoiding unnecessary replacements.

[0015] The first difference between this solution and the prior art is the ability to process random data. Existing lithium battery life prediction methods are mainly based on regular sampling or fixed operating conditions, and cannot effectively process irregular or random data. In contrast, this method innovatively targets random operating data, and can accurately extract multi-dimensional features of lithium batteries (such as cumulative usage, health status, and operating status) in complex usage environments, and construct feature vectors, which is not possible with prior art. The improvement in random data processing capabilities means that battery status assessment is no longer limited to ideal conditions, thereby enhancing the practical applicability of battery management. This is crucial for application areas such as electric vehicles and energy storage systems.

[0016] The second difference between this case and the prior art is the combination of multidimensional feature vectors and digital distance features. Traditional methods usually only focus on a key feature of the battery, while ignoring its complex, multidimensional characteristics. This patent provides a comprehensive description of the state of lithium batteries through the construction of multidimensional feature vectors and the generation of digital distance feature vectors. This enables the method to more accurately assess the overall health of the battery when faced with complex data. This multidimensional analysis method provides a deeper understanding of the overall evaluation of the battery and helps to promptly detect potential performance problems in a changing usage environment.

[0017] The third difference between this case and the prior art lies in the application of envelope processing technology. The prior art lacks effective means of processing irregular data, and this patent introduces envelope processing technology (such as Hilbert transform, extreme point interpolation method, etc.) into the field of lithium battery data processing for the first time, enabling it to extract key information from complex data and improve evaluation accuracy. The application of envelope processing technology has opened up a new direction for the analysis of battery data, allowing effective information to be extracted under irregular data conditions. This innovation has improved the reliability of battery evaluation and promoted the development of related fields.

[0018] The fourth difference between this case and the prior art lies in the innovative application of interpolation repair. The data repair methods in the prior art mostly use simple linear interpolation and cannot process complex discrete data. This method innovatively generates a continuous feature sequence from discrete data through interpolation repair technology after envelope sampling, overcomes the problem of data discontinuity, and improves the accuracy of evaluation and the flexibility of data processing. The application of interpolation repair technology enhances the continuity and integrity of the data, providing a more reliable data basis for subsequent analysis. This innovation ensures that the evaluation results are still highly credible when faced with incomplete data.

[0019] The fifth difference between this case and the prior art lies in the flexible selection of multiple distance calculation methods. Traditional methods mostly use a single distance calculation method, while this method introduces multiple distance calculation methods including Euclidean distance and Mahalanobis distance, providing flexible choices for feature extraction, thereby achieving the best evaluation effect in different application scenarios. The flexible use of multiple distance calculation methods enables this method to select the most appropriate analysis method for specific data features. This innovation provides a guarantee for improving the accuracy of battery status assessment and ensuring the accuracy of the assessment results.

[0020] The fifth difference between this case and the prior art is the life prediction that adapts to complex application scenarios. Existing battery life prediction models usually assume stable operating conditions and cannot adapt to complex and changeable actual usage environments. This method can provide more accurate prediction results under uneven usage conditions by dynamically analyzing multi-dimensional features and combining historical data to predict life. The ability to adapt to complex application scenarios enables this method to better meet the needs of battery management, thereby improving the safety and reliability of battery use. This adaptability enhances the flexibility and practicality of batteries in different applications.

[0021] This solution can adapt to complex working conditions. This innovation can extract key features of lithium batteries from irregular and random data, adapt to complex usage scenarios, and effectively improve the accuracy of battery status assessment. Through adaptive processing of complex working conditions, this method effectively improves the battery status assessment capability and ensures that reliable data can still be obtained under different usage environments. This is crucial to ensure the safe operation of electric vehicles and energy storage systems.

[0022] This solution realizes the fusion of multi-dimensional data. Through the construction and analysis of multi-dimensional feature vectors, it can comprehensively evaluate the cumulative battery usage, health status, working condition and other aspects, providing a more comprehensive basis for subsequent life prediction. The fusion of multi-dimensional data not only improves the comprehensiveness of the evaluation, but also provides a more scientific basis for subsequent maintenance and management decisions. Through comprehensive analysis, potential problems can be discovered in time and corresponding measures can be taken.

[0023] This solution achieves a seamless integration of envelope processing and interpolation repair. Envelope processing technology is used to extract key information of the data, and the data sequence is reconstructed through interpolation repair methods. It can generate a continuous feature sequence when the original data is incomplete, improving the robustness and accuracy of data processing. The innovation lies in the combination of envelope processing and interpolation repair to enhance the continuity and integrity of the data. This lays a more solid foundation for subsequent data analysis and battery status assessment, making the assessment results more credible.

[0024] This solution can perform flexible distance calculation. Combining multiple distance calculation methods, such as Euclidean distance, Manhattan distance, etc., it can flexibly adapt to different data features and improve the accuracy of feature extraction. The flexible distance calculation method enables the present invention to select the most appropriate calculation method according to the data conditions of different features. This flexibility provides a guarantee for the accuracy of feature extraction and helps to further improve the effect of battery evaluation. Further, it also includes step S6: using at least one of the envelope sampling sequence of each key feature, the discrete sampling sequence of each key feature, and the sampling repair sequence of each key feature; using historical data as input and the remaining life as output to train the model.

[0025] Furthermore, the health status time series includes changes in battery capacity, changes in battery power capacity, changes in internal resistance, and changes in charge and discharge efficiency.

[0026] Further, the cumulative usage characteristics include the cumulative charging work, the cumulative discharging work, the total cumulative absolute value of charging and discharging work, the cumulative charging time, the cumulative discharging time, the total cumulative charging and discharging time, the cumulative charging amount, the cumulative discharging amount, the total cumulative absolute value of charging and discharging amount, the cumulative number of charging times, the cumulative number of discharging times, the total cumulative number of charging and discharging times, the cumulative number of shelving times, the cumulative shelving time, the cumulative calendar service time, the cumulative charging ratio, the cumulative discharge ratio, the total cumulative absolute value of the charging ratio and the discharging ratio, the cumulative charging power ratio, the cumulative discharge power ratio, the total cumulative absolute value of the charging power ratio and the discharge power ratio, the cumulative actual workload generated by the battery module for power-consuming equipment, the cumulative actual work generated by the battery module for power-consuming equipment, and the cumulative actual mileage generated by the battery module for vehicle driving.

[0027] Further, the health status characteristics include: actual maximum power storage capacity, attenuation of actual maximum power storage capacity, actual maximum power storage capacity, attenuation of actual maximum power storage capacity, relative power storage capacity, attenuation of relative power storage capacity, relative power storage capacity, attenuation of relative power storage capacity, actual internal resistance, attenuation of actual internal resistance, single actual discharge time, attenuation of single actual discharge time, relative discharge time, single actual charging time, attenuation of single actual charging time, relative charging time, actual work generated by the actual maximum power storage capacity of the battery for the operation of power-consuming equipment, and mileage generated by the actual maximum power storage capacity of the battery for automobile driving.

[0028] Furthermore, the operating state characteristics include: various operating conditions when the battery is running; in the actual operation of the battery, the setting of different operating conditions will have an impact on the performance of the battery. The optional types of operating states include: the specific changes or averages of the output current, output voltage, and output power of the battery during battery operation, air humidity, heat dissipation conditions, air pressure conditions, equipment operating power, equipment production efficiency, and vehicle speed.

[0029] Furthermore, the calculation type of the point distance in step S2 includes any one of Manhattan distance, Euclidean distance, Hamming distance, standardized Euclidean distance, Mahalanobis distance, Chebyshev distance, and Minkowski distance.

[0030] Further, the sampling of discrete envelope moments in step S3 is performed by uniform sampling or non-uniform sampling;

[0031] Envelope processing methods include extreme point interpolation method, Hilbert transform, filtering method, and moving average method.

[0032] Further, the step of sampling the interpolation repair sequence in step S5 is: firstly, the envelope sampling sequence of each key feature is subjected to spline interpolation, and the interpolation function of the cumulative usage is obtained as S x (·) to realize the interpolation of cumulative usage and time series; the interpolation function of health status is S y (·) realizes the interpolation of health status and time series; the interpolation function of working condition is S c (·) to realize the interpolation of working state and time series.

[0033] The present invention solves scientific problems such as random working condition data processing, comprehensive analysis of multi-dimensional features, data envelopment processing and interpolation repair.

[0034] This solution solves the problem of how to realize random operating condition data processing, that is, how to extract representative features from irregular or random usage data to evaluate the health status of lithium batteries and predict their remaining life. The processing of random operating condition data is a major challenge in the current field of lithium battery evaluation. Through an effective feature extraction method, the present invention provides a solution that enables the true status information of lithium batteries to be obtained even when the data is irregular. Emphasizing the particularity and complexity of lithium batteries, the present invention has higher flexibility and adaptability when processing such data.

[0035] This case aims to solve the problem of how to achieve comprehensive analysis of multi-dimensional features, that is, how to combine multi-dimensional information such as the cumulative usage, health status and working condition of lithium batteries to generate a feature vector that can reflect the comprehensive condition of the battery. Comprehensive analysis of multi-dimensional features is the key to improving the accuracy of battery status assessment. By integrating data from different dimensions, we can have a more comprehensive understanding of the working condition of lithium batteries and their potential failure risks, thereby optimizing maintenance strategies. This process emphasizes the performance of lithium batteries in a changing environment and provides a comprehensive perspective for their health management.

[0036] This case aims to solve the problem of how to implement data envelopment processing and interpolation repair, that is, how to extract key information from irregular data through envelope processing technology, and generate continuous feature sequences through interpolation repair technology to make up for the incompleteness of the data. The combination of data envelopment processing and interpolation repair helps to improve the robustness of the evaluation model. In the case of missing data, the accuracy of the model can be ensured through interpolation repair technology, thereby providing a reliable data basis for lithium battery life prediction. This process reflects the importance of lithium battery data characteristics, making the feature extraction process more detailed.

[0037] This solution realizes the construction of multi-dimensional feature vectors. By extracting multi-dimensional data such as the cumulative usage, health status and working condition of lithium batteries, a multi-dimensional feature vector is generated. This comprehensive feature extraction method surpasses the single feature analysis method in the existing technology and significantly improves the comprehensiveness and accuracy of battery status assessment. The construction of multi-dimensional feature vectors makes the battery evaluation not limited to a single performance indicator, but integrates information from multiple aspects. This innovation provides important support for a comprehensive understanding of the battery status.

[0038] This solution realizes the innovative application of envelope processing. Introducing envelope processing technology (such as Hilbert transform, extreme point interpolation method, etc.) into the field of lithium battery status assessment can extract eigenvalues ​​from random data and generate envelope sequences, which is an innovative application that has not been used in the prior art. The innovative application of envelope processing provides a new perspective for the analysis of battery data, making it possible to extract key information even when the data is irregular. This innovation improves the flexibility and applicability of the assessment.

[0039] This solution realizes the combination of interpolation and repair technology. After extracting discrete eigenvalues, this method generates a continuous feature sequence through interpolation and repair technology, and can combine multiple interpolation methods (such as spline interpolation) to enhance the continuity and accuracy of the eigenvalues. The combination of interpolation and repair technology not only makes up for the problem of missing data, but also provides more continuity for the analysis of feature sequences. This innovation makes the evaluation results more reliable and helps to better reflect the true state of the battery.

[0040] This solution can flexibly adapt to random data. Most existing technologies are aimed at regular data or specific working conditions, while this method can flexibly cope with complex usage scenarios through adaptive processing of random data and multi-dimensional feature extraction. The ability to flexibly adapt to random data enables this method to maintain a high evaluation accuracy when facing complex and changeable usage conditions in the real world. This feature makes the present invention have a wide range of application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a line chart of original data;

[0042] Figure 2 is a scatter plot of the original data;

[0043] Figure 3 This is a schematic diagram of the result after the envelope feature extraction method in this case is applied to the original data;

[0044] Figure 4 This is a schematic diagram of the final envelope curve result;

[0045] Figure 5 Schematic diagram of the steps of a lithium battery envelope feature extraction method for random usage data. DETAILED DESCRIPTION

[0046] The following is further described in detail through specific implementation methods:

[0047] Embodiment 1 is as attached Figure 5 As shown,

[0048] S1: Obtain the original sampling sequence of each key feature of the rechargeable battery. The key features of the lithium battery include cumulative usage features, health status features, operating status features, and timing features;

[0049] The original sampling sequence of the key characteristics of the lithium battery includes the cumulative usage time series, the health status time series, the operating status time series, and the corresponding time index sequence.

[0050] Time index sequences can identify or record the order or moment of occurrence of certain events or data in time. They are usually used to record the time point of data collection so as to facilitate subsequent analysis and comparison of data changes in different time periods. k} refers to the time index sequence, each time point t1, t2, t3, ... t k The spacing between them can be unequal;

[0051] The operating state time series includes the characteristic values ​​of various operating states recorded at different time points, which can be used to describe the state changes of the system or equipment under different operating conditions; k} refers to {t1,t2,t3,…t k}The corresponding working state time series;

[0052] The health status time series describes a set of data about the changes in the health status of a rechargeable battery at different time points. These characteristics may include changes in battery capacity, changes in battery power capacity, changes in internal resistance, changes in charge and discharge efficiency, etc. By recording and analyzing these time series data, the health status of the battery can be evaluated, its future performance can be predicted, and appropriate maintenance or operation strategies can be formulated; k} refers to {t1,t2,t3,…t k}corresponding health status time series;

[0053] The health status time series describes a set of data about the changes in the health status of a rechargeable battery at different time points. These characteristics may include changes in battery capacity, changes in battery power capacity, changes in internal resistance, changes in charge and discharge efficiency, etc. By recording and analyzing these time series data, the health status of the battery can be evaluated, its future performance can be predicted, and appropriate maintenance or operation strategies can be formulated; k} refers to {t1,t2,t3,…t k}corresponding health status time series;

[0054] The cumulative usage characteristics include: the cumulative charging work, the cumulative discharging work, the total cumulative absolute value of the charging and discharging work, the cumulative charging time, the cumulative discharging time, the total cumulative charging and discharging time, the cumulative charging amount, the cumulative discharging amount, the total cumulative absolute value of the charging and discharging amount, the cumulative number of charging times, the cumulative number of discharging times, the total cumulative number of charging and discharging times, the cumulative number of shelving times, the cumulative shelving time, the cumulative calendar service time, the cumulative charging ratio, the cumulative discharge ratio, the total cumulative absolute value of the charging ratio and the discharging ratio, the cumulative charging power ratio, the cumulative discharge power ratio, the total cumulative absolute value of the charging power ratio and the discharge power ratio, the cumulative actual workload generated by the battery module for power-consuming equipment, the cumulative actual work generated by the battery module for power-consuming equipment, and the cumulative actual mileage generated by the battery module for vehicle driving.

[0055] The health status characteristics include: actual maximum power storage capacity, attenuation of actual maximum power storage capacity, actual maximum power storage capacity, attenuation of actual maximum power storage capacity, relative power storage capacity, attenuation of relative power storage capacity, relative power storage capacity, attenuation of relative power storage capacity, actual internal resistance, attenuation of actual internal resistance, single actual discharge time, attenuation of single actual discharge time, relative discharge time, single actual charging time, attenuation of single actual charging time, relative charging time, actual work generated by the operation of power-consuming equipment with the actual maximum power storage capacity of the battery, and mileage generated by the actual maximum power storage capacity of the battery for vehicle driving.

[0056] The operating state characteristics include: various operating conditions when the battery is running; in the actual operation of the battery, the setting of different operating conditions will affect the performance of the battery. The optional types of operating states include: the specific changes or average values ​​of the battery's output current, output voltage, and output power during battery operation, air humidity, heat dissipation conditions, air pressure conditions, equipment operating power, equipment production efficiency, vehicle speed, etc.

[0057] In this step and any subsequent steps, the original sampling sequence can be normalized or denormalized in a timely manner. Normalization and denormalization are two key operations in data processing, especially when processing the characteristic data of rechargeable batteries. Normalization maps the data to a fixed range, usually [0,1] or [-1,1], so that the scale of each feature is consistent, which is conducive to model learning and prediction. Denormalization, on the other hand, restores the normalized data to the original scale in order to interpret and apply the actual values. The following is a detailed explanation of these two processes. The purpose of normalization is to convert data features of different scales into the same range for subsequent processing. Usually, different features in the original data (such as time, cumulative usage, health status, etc.) have large differences in magnitude, which will affect the effect of model training. Commonly used normalization methods are minimum-maximum normalization and Z-score normalization.

[0058] Minimum-maximum normalization: Minimum-maximum normalization maps the data to the range of [0,1]. For a feature sequence, such as the cumulative usage sequence X = {x1, x2, x3, ..., x k}, the normalized value x' i The calculation formula is as follows:

[0059] x' i =(x i -x min ) / (x max -x min )

[0060] in:

[0061] ·x i is a value in the original data,

[0062] ·x min and x max are the minimum and maximum values ​​in the sequence X respectively.

[0063] After min-max normalization, all data will be scaled between 0 and 1, making the data have a uniform scale. This method can be applied to cumulative usage, health status, and time series data.

[0064] Z-score standardization: Z-score standardization adjusts the mean of the data to 0 and the variance to 1. The calculation formula is:

[0065] x' i =(x i -μ) / σ

[0066] in:

[0067] μ is the mean value of the sequence X,

[0068] σ is the standard deviation.

[0069] Z-score standardization has low requirements on data distribution and is suitable for situations where the original data is not necessarily in a fixed range. This method makes the data more consistent with the normal distribution and is more friendly to some machine learning models.

[0070] Denormalization is the process of restoring the data to its original scale after model processing. It can restore the normalized predicted data to its original value with practical significance, which is convenient for interpretation and application. For example, if the health status data is normalized to between 0 and 1 during the minimum-maximum normalization process, denormalization can restore the predicted health status value to the original range.

[0071] Minimum-maximum denormalization: For the data x' that has been normalized by minimum-maximum i , can be restored to its original scale by the following formula:

[0072] x i =x' i *(x max -x min )+x min

[0073] In this way, the normalized value x'i can be restored to its original value range.

[0074] Z-score denormalization: For data that has been normalized by Z-score, it can be restored to the original data using the following formula:

[0075] x i =x' i *σ+μ

[0076] This method can also restore the data from the standardized distribution to the original distribution state, so that the predicted data can truly reflect the physical meaning of the original data.

[0077] The original data sequence obtained in the feature collection step, such as the cumulative usage feature X = {x1, x2, x3, ..., x k}、Health status characteristics Y={y1,y2,y3,…y k}, etc., there may be differences in value range and distribution. These sequences can be normalized when appropriate so that the features can play an equivalent role in subsequent processing or models.

[0078] After the model calculation is completed, if the results are to be meaningful, the processed data can be denormalized. For example, if the predicted values ​​of health status or cumulative usage have been normalized, they need to be converted back to the original scale through denormalization to facilitate subsequent interpretation and application.

[0079] Through normalization and denormalization, data with different characteristics can be processed and applied more flexibly, ensuring consistency in data calculation and interpretation.

[0080] S2: construct a multidimensional feature vector F using the acquired original sampling sequence; calculate the point distance between the corresponding value of the multidimensional feature vector F at each moment and the coordinate origin to generate a numerical distance feature vector D;

[0081] The distance types that can be adopted in the point distance calculation process include any one of Manhattan distance, Euclidean distance, Hamming distance, standardized Euclidean distance, Mahalanobis distance, Chebyshev distance, and Minkowski distance.

[0082] For example, the time index series, cumulative usage time series, and health status time series are represented as T = {t1, t2, t3, ... t k}, X={x1,x2,x3,…,x k} and Y={y1,y2,y3,…y k Then, these three time series can be combined into a multidimensional feature vector F = {(t1,x1,y1),(t2,x2,y2),(t3,x3,y3),…(t k ,x k ,y k )}. Alternatively, only the cumulative usage and health status time series are combined to obtain a multidimensional feature vector F = {(x1,y1),(x2,y2),(x3,y3),…(x k ,y k )}.

[0083] To quantify the multidimensional data in the feature vector, a distance feature vector D can be constructed to represent the distance between each feature point and the origin. k} represents the distance feature vector corresponding to the multidimensional feature vector F = {(x1,y1), (x2,y2), (x3,y3), ... (x4,y4)}, where d1 = distance((x1,y1), (0,0)), d2 = distance((x2,y2), (0,0)), i.e., the one-dimensional data points d1, d2, d3, ... d k The value of is equal to each multidimensional data point in F (x1, y1), (x2, y2), (x3, y3), … (x k ,y k ) and the multidimensional origin (0,0). In other words, for each d i (where i = 1, 2, ..., k), define d i =distance((x i ,yi ),(0,0)), that is, calculate (x i ,y i ) from the multidimensional origin (0,0). In some cases, other distances can also be used. Thus, each element d in the distance feature vector D i Corresponds to the distance between each data point in the feature vector F and the origin.

[0084] S3: performing envelope processing on the obtained distance feature vector to obtain a discrete envelope distance value and determine the corresponding discrete envelope moment;

[0085] According to the specific distribution of the discrete envelope distance values, the discrete envelope moments can be uniformly sampled or non-uniformly sampled.

[0086] For example, the discrete envelope distance values ​​after envelope processing are d2, d5, d 10 ,d 23 ,…d k-2 , whose rank in the distance eigenvector is 2 nd ,5 th ,10 th ,twenty three th ,…k-2 th , so the corresponding discrete envelope moments are t2, t5, t 10 ,t 23 ,…t k-2 is non-uniform;

[0087] Envelope processing is usually used in signal processing to extract the envelope information of the signal, that is, the shape of the signal. The optional envelope processing methods include: extreme point interpolation method, Hilbert transform, filtering method, and moving average method.

[0088] Hilbert transform is a commonly used method in signal processing, which is used to calculate the analytical signal of a signal and extract the envelope of the signal from it; the specific steps are to calculate the Hilbert transform of the signal to obtain the analytical signal; then the modulus of the analytical signal is the envelope.

[0089] The extreme point interpolation method first finds all the local maximum points in the signal, and then uses these maximum points to perform interpolation (such as spline interpolation) to obtain the envelope.

[0090] The filtering method uses a low-pass filter to remove high-frequency components in the signal to obtain an envelope; first, a low-pass filter is generated, and then the signal is passed through the low-pass filter to obtain an envelope.

[0091] The moving average method smoothes the signal by calculating the moving average of the signal to obtain the envelope; first select a suitable window size, and then calculate the moving average of the signal.

[0092] S4: obtaining the order of discrete envelope moments in the time index sequence, sampling from the original sampling sequence of each key feature, and constructing the envelope sampling sequence of each key feature;

[0093] S5: According to the relationship between the envelope sampling sequence of each key feature and the discrete envelope moment, an interpolation function is constructed to describe this correlation; the eigenvalue of each key feature at the original time index is calculated by the interpolation function to generate a continuous interpolation repair sequence of each key feature.

[0094] For example, after envelope processing, the discrete envelope distance values ​​obtained are d2, d5, d 10 ,d 23 ,…d k-2 , whose rank in the distance eigenvector is 2 nd ,5 th ,10 th ,twenty three th ,…k-2 th , then the original sampling sequence of each key feature is sampled in the same order to obtain the envelope sampling sequence of the cumulative usage {x2,x5,x 10 ,x 23 ,…x k-2}、Health state envelope sampling sequence {y2,y5,y 10 ,y 23 ,…y k-2}、Working state envelope sampling sequence {c2,c5,c 10 ,c 23 ,…c k-2};

[0095] When obtaining the sampling repair sequence, first perform spline interpolation on the envelope sampling sequence of each key feature to obtain the interpolation function. For example, the interpolation function of the cumulative usage is S x (·), which can realize the interpolation of cumulative usage and time series; the interpolation function of health status is S y (·), which can realize the interpolation of health status and time series; the interpolation function of working condition is S c (·), which can realize the interpolation of working state and time series;

[0096] After obtaining the interpolation function, the original time index sequence {t1, t2, t3, …t k} is the time base, and the interpolation function is used to generate the cumulative usage sampling repair sequence {X1,X2,X3,…X k}、Healthy state sampling repair sequence {Y1,Y2,Y3,…Y k}、Working condition sampling repair sequence {C1,C2,C3,…C k};

[0097] S6: using at least one of an envelope sampling sequence of each key feature, a discrete sampling sequence of each key feature, and a sampling repair sequence of each key feature;

[0098] The model is trained using historical data as input and remaining life as output.

[0099] Figure 1 The original data line chart is used to show the changing trend of continuous data. The data points are connected by lines, which is suitable for showing the change of variables over time or sequence. Figure 1 The distribution of the original capacity data of three different lithium batteries A, B, and C relative to the cumulative capacity is shown. The curve fluctuates greatly, indicating that there is a lot of noise and irregular fluctuations in the original data, which cannot be directly used for accurate status assessment or life prediction.

[0100] Figure 2 The original data scatter plot is used to show the relationship between two variables. Each point represents a data pair independently, and there is no connection between the points, which is suitable for observing the correlation between variables. Figure 2 Converting the original data into a scatter plot shows the discrete distribution of capacity and cumulative computation. Figure 1 The visualization is much clearer than before, but the data still has obvious discreteness and discontinuity, which is difficult to directly process using traditional methods. The charts at this stage emphasize the challenges brought by data discontinuity.

[0101] Figure 3It is the result after applying the envelope feature extraction method in this case. After the original capacity data is processed, three envelope curves A, B, and C are obtained. By enveloping the data, the noise is significantly suppressed, the data fluctuations are smoothed, and the overall trend between the capacity and the accumulated amount is shown. This step is the key. Through the envelope feature extraction technology, a smooth feature curve is generated from discrete data, which provides a more stable basis for subsequent analysis. The difference between the new method in this case and the existing methods is that in traditional methods, such as Hilbert transform, extreme point interpolation method and filtering method, they are mainly used to process periodic and regular time-frequency signals. These methods extract the envelope characteristics of the signal by capturing the frequency and time components of the signal, and perform well in time-frequency fields such as vibration signals and voice signals. However, these methods face huge challenges in processing lithium battery data because the working environment and usage of lithium batteries are often random, and the data does not have obvious periodic laws. In contrast, the new envelope feature extraction method is more suitable for complex random usage data environments such as lithium batteries. By combining multi-dimensional feature analysis and advanced envelope algorithms, this method can effectively eliminate noise and extract the core health characteristics and usage patterns of lithium batteries. This breakthrough extraction technology breaks through the limitations of traditional methods, making it not only suitable for periodic signals, but also able to cope with random and nonlinear usage conditions in battery systems. Figure 2 to Figure 3 Compared with traditional methods, it can more effectively process random usage data of lithium batteries, reduce noise interference, and improve the accuracy and stability of feature extraction through multi-dimensional data analysis, making it suitable for more complex usage scenarios and life prediction tasks.

[0102] Figure 4 The final envelope curve result is shown, which further simplifies and removes unnecessary fluctuations, presenting the continuous change of capacity with the accumulated calculation amount. This step demonstrates the powerful ability of the new envelope extraction method in processing complex data, which can remove noise while retaining the core features of the data, generate highly accurate battery characteristic curves, and provide strong support for battery life prediction and status monitoring.

[0103] The above are only embodiments of the present invention. The invention is not limited to the fields involved in this implementation case. The common sense such as the known specific structures and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to explain the content of the claims.

Claims

1. A lithium battery envelope feature extraction method for random usage data, characterized in that: The following steps are involved: S1: Obtaining the original sampling sequence of each key feature of the target battery; the key features include cumulative usage features, health status features, operating status features, and time series features; the original sampling sequence includes a cumulative usage time series, a health status time series, an operating status time series, and a corresponding time index sequence, the time index sequence is used to identify or record the order or moment of the original sampling sequence in time, and record the time point of data collection; S2: construct a multidimensional feature vector F using the acquired original sampling sequence; calculate the point distance between the corresponding value of the multidimensional feature vector F at each moment and the coordinate origin to generate a numerical distance feature vector D; S3: Perform envelope processing on the obtained distance feature vector D to obtain a discrete envelope distance value and determine the corresponding discrete envelope moment; S4: obtaining the order of discrete envelope moments in the time index sequence, sampling from the original sampling sequence of each key feature, and constructing the envelope sampling sequence of each key feature; S5: construct an interpolation function according to the relationship between the envelope sampling sequence of each key feature and the discrete envelope moment; calculate the characteristic value of each key feature at the original time index through the interpolation function, and generate a continuous interpolation repair sequence of each key feature.

2. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: It also includes step S6: using at least one of the envelope sampling sequence of each key feature, the discrete sampling sequence of each key feature, and the sampling repair sequence of each key feature; using historical data as input and the remaining life as output to train the model.

3. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: The health status time series includes changes in battery capacity, changes in battery power capacity, changes in internal resistance, and changes in charge and discharge efficiency.

4. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: The cumulative usage characteristics include the cumulative charging work, the cumulative discharging work, the total cumulative absolute value of charging and discharging work, the cumulative charging time, the cumulative discharging time, the total cumulative charging and discharging time, the cumulative charging amount, the cumulative discharging amount, the total cumulative absolute value of charging and discharging amount, the cumulative number of charging times, the cumulative number of discharging times, the total cumulative number of charging and discharging times, the cumulative number of shelving times, the cumulative shelving time, the cumulative calendar service time, the cumulative charging ratio, the cumulative discharge ratio, the total cumulative absolute value of the charging ratio and the discharging ratio, the cumulative charging power ratio, the cumulative discharge power ratio, the total cumulative absolute value of the charging power ratio and the discharge power ratio, the cumulative actual workload generated by the battery module for power-consuming equipment, the cumulative actual work generated by the battery module for power-consuming equipment, and the cumulative actual mileage generated by the battery module for automobile driving.

5. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: The health status characteristics include: actual maximum power storage capacity, attenuation of actual maximum power storage capacity, actual maximum power storage capacity, attenuation of actual maximum power storage capacity, relative power storage capacity, attenuation of relative power storage capacity, relative power storage capacity, attenuation of relative power storage capacity, actual internal resistance, attenuation of actual internal resistance, single actual discharge time, attenuation of single actual discharge time, relative discharge time, single actual charging time, attenuation of single actual charging time, relative charging time, actual work generated by the actual maximum power storage capacity of the battery for the operation of power-consuming equipment, and mileage generated by the actual maximum power storage capacity of the battery for vehicle driving.

6. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: The operating state characteristics include: various operating conditions when the battery is running; in the actual operation of the battery, the setting of different operating conditions will have an impact on the performance of the battery. The optional types of operating states include: the specific changes or averages of the output current, output voltage, and output power of the battery during battery operation, air humidity, heat dissipation conditions, air pressure conditions, equipment operating power, equipment production efficiency, and vehicle speed.

7. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: The calculation type of the point distance in step S2 includes any one of Manhattan distance, Euclidean distance, Hamming distance, standardized Euclidean distance, Mahalanobis distance, Chebyshev distance, and Minkowski distance.

8. The method for extracting lithium battery envelope features for random usage data according to claim 1, characterized in that: The sampling of discrete envelope moments in step S3 is performed by using either uniform sampling or non-uniform sampling; Envelope processing methods include extreme point interpolation, Hilbert transform, filtering, and moving average.

9. The method for extracting lithium battery envelope features for random usage data according to claim 8, characterized in that: The steps of sampling the interpolation repair sequence in step S5 are: firstly, the envelope sampling sequence of each key feature is subjected to spline interpolation, and the interpolation function of the cumulative usage is obtained as S x (·) to realize the interpolation of cumulative usage and time series; the interpolation function of health status is S y (·) realizes the interpolation of health status and time series; the interpolation function of working condition is S c (·) to realize the interpolation of working state and time series.

Images