Battery state monitoring analysis method and system based on machine learning

Abstract

The invention provides a battery state monitoring analysis method and system based on machine learning, and the method comprises the steps: firstly obtaining real-time operation data, covering voltage, current, temperature sequences and timestamp sequences in the charge-discharge circulation of a target battery, then carrying out the multi-dimensional feature extraction to generate an original feature set, obtaining a standardized feature matrix through the dynamic time sequence processing, and carrying out the dynamic time sequence processing; the method comprises the following steps of: acquiring a battery state monitoring model, inputting the battery state monitoring model into a pre-trained battery state monitoring model, generating degradation, health and residual life prediction parameters, generating an abnormal detection result and a maintenance decision suggestion, and finally adaptively adjusting the weight of the model parameter based on the result, and updating the model prediction precision. The method can comprehensively and accurately monitor and analyze the battery state, improves the prediction precision, and provides a reliable basis for battery maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and in particular to a battery status monitoring and analysis method and system based on machine learning. Background Art

[0002] In today's era, batteries, as a key energy storage device, are widely used in many fields such as electric vehicles, portable electronic devices, energy storage power stations, etc. As the requirements for battery performance and reliability of various devices continue to increase, accurate monitoring and analysis of battery status has become crucial.

[0003] Existing technologies attempt to evaluate battery status through some conventional data analysis methods, but these methods often lack in-depth mining of implicit features in battery operation data. For example, they usually only perform surface statistical analysis on the data, ignoring the deep correlation and time series features between the data. For example, when analyzing battery charging and discharging data, the dynamic changes of multiple variables such as current and temperature over time and their interactions are not fully considered, resulting in low accuracy in predicting battery health status and remaining life, making it difficult to meet the demand for accurate grasp of battery status in practical applications.

[0004] In terms of battery maintenance decisions, most existing methods are based on fixed maintenance cycles or general maintenance standards, and do not fully consider the actual operating status and characteristics of each battery. This may not only lead to excessive maintenance, increase unnecessary costs and waste resources, but also fail to deal with individual abnormal conditions of the battery in a timely manner, causing premature damage or performance degradation of the battery, affecting the normal operation of the equipment. Summary of the invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a battery status monitoring and analysis method based on machine learning, the method comprising: Acquire real-time operation data of the target battery in the charge and discharge cycle, wherein the real-time operation data includes a voltage sequence, a current sequence, a temperature sequence, and a timestamp sequence; Performing multi-dimensional feature extraction on the real-time operation data to generate an original feature set, and performing dynamic time series processing on the original feature set to obtain a standardized feature matrix; Inputting the standardized feature matrix into a pre-trained battery state monitoring model to generate degradation state parameters, health state parameters and remaining life prediction parameters of the target battery; Generate an abnormality detection result and maintenance decision suggestion for the target battery according to the degradation state parameter, health state parameter and remaining life prediction parameter; Based on the abnormality detection results and maintenance decision suggestions, the parameter weights of the battery status monitoring model are adaptively adjusted to update the prediction accuracy of the battery status monitoring model.

[0006] On the other hand, an embodiment of the present invention also provides a battery status monitoring and analysis system based on machine learning, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0007] Based on the above aspects, the embodiment of the present application first collects real-time operating data such as voltage, current, temperature and timestamp sequence in the charge and discharge cycle of the target battery in the data acquisition stage, which can fully and dynamically reflect the actual operating status of the battery under different working conditions. Secondly, the multi-dimensional feature extraction and dynamic timing processing link, by extracting multi-dimensional features of real-time operation data, generates an original feature set, and can deeply mine the key information hidden in the data. The dynamic timing processing further integrates these original features into a standardized feature matrix, which not only fully considers the time series characteristics of the battery operation data, but also adapts to the dynamic changes of battery performance over time. Compared with the traditional fixed mode feature processing method, it can better capture the subtle changes in the battery state, greatly improving the effectiveness and representativeness of the features. Furthermore, the standardized feature matrix is ​​input into the pre-trained battery state monitoring model to accurately generate degradation state parameters, health state parameters and remaining life prediction parameters. The battery state monitoring model is trained based on a large amount of data, has a high degree of generalization and accuracy, can effectively parse complex battery state information, and provides a quantitative and reliable basis for the full life cycle management of the battery. In terms of generating abnormal detection results and maintenance decision recommendations, based on comprehensive and accurate parameter analysis, it can quickly and accurately identify abnormal conditions in battery operation, and generate practical maintenance decision recommendations based on different battery states and potential risks, realizing personalized and intelligent maintenance management based on the actual state of the battery, greatly improving maintenance efficiency, reducing maintenance costs, and extending the service life of the battery. Finally, based on the abnormal detection results and maintenance decision recommendations, the parameter weights of the battery state monitoring model are adaptively adjusted, which can continuously optimize the performance of the battery state monitoring model according to the actual monitoring results and update the prediction accuracy in real time. Compared with the traditional fixed parameter model, this method can dynamically adapt to the complex changes in the battery state, continuously maintain high-precision monitoring and prediction capabilities, effectively avoid monitoring errors caused by model aging or gradual changes in battery performance, and provide a strong guarantee for long-term stable monitoring of the battery state. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1It is a schematic diagram of the execution flow of the battery status monitoring and analysis method based on machine learning provided in an embodiment of the present invention.

[0009] Figure 2 It is a schematic diagram of the hardware architecture of a battery status monitoring and analysis system based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a battery status monitoring and analysis method based on machine learning provided by an embodiment of the present invention. The battery status monitoring and analysis method based on machine learning is introduced in detail below.

[0011] Step S110, acquiring real-time operation data of the target battery in the charge and discharge cycle, wherein the real-time operation data includes a voltage sequence, a current sequence, a temperature sequence and a timestamp sequence.

[0012] In this embodiment, taking the power battery of an electric vehicle as an example, during the operation of the electric vehicle, the battery is continuously charged and discharged.

[0013] In order to obtain the real-time operating data of the target battery, a distributed sensor node network can be constructed. Among them, the voltage acquisition node collects the battery voltage at a higher first sampling frequency (such as every 5 milliseconds). During the acceleration, deceleration or charging of the vehicle, the voltage will change differently, and these collected voltage values ​​constitute a voltage sequence. The current acquisition node collects the current sequence at a second sampling frequency of every 3 milliseconds. When the vehicle is in different driving states (such as climbing, driving at a constant speed, and rapid acceleration), the battery's charge and discharge current will change significantly. The temperature acquisition node collects the temperature sequence at a third sampling frequency of every 10 milliseconds. Since the battery generates heat during high-power charging and discharging, the temperature acquisition point can capture the temperature changes at different locations of the battery.

[0014] Furthermore, the time synchronization node aligns the sampling moments of the above three acquisition nodes at the microsecond level. For example, when the vehicle accelerates sharply at a certain moment, the voltage acquisition node collects the voltage while the current acquisition node collects the corresponding large current value. The temperature acquisition node also collects the change in battery temperature at the same microsecond moment to ensure that the timestamps of the three are consistent. The collected data is sent to the edge computing gateway through an encrypted transmission protocol for temporary caching and integrity verification. After the verification is passed, the real-time operation data containing voltage series, current series, temperature series and timestamp series can be read in batches.

[0015] Step S120, performing multi-dimensional feature extraction on the real-time operation data to generate an original feature set, and performing dynamic time series processing on the original feature set to obtain a standardized feature matrix.

[0016] For voltage sequence feature extraction, during a charging process, the local maximum and minimum in the voltage sequence are found through the extreme point detection algorithm. Assuming that during a certain period of charging, the local maximum is 4.2 volts and the minimum is 4.0 volts, the absolute difference is calculated to obtain the voltage fluctuation extreme value of 0.2 volts. A first-order difference operation is performed to obtain the differential voltage sequence, and the arithmetic mean is calculated within a 5-second sliding time window to obtain the mean voltage change rate. At the same time, the duration of the voltage platform is detected. If the maximum continuous time period for which the voltage change rate is continuously lower than the platform threshold of 0.01 volts / second is 20 seconds, then 20 seconds is the duration of the voltage platform.

[0017] From the current sequence, if the charging input is 1000 watt-hours and the discharging output is 850 watt-hours, then the current charging and discharging efficiency is 85%. Identify the current peaks. If the first peak is at the 8th second and the second at the 20th second, the current peak interval is 12 seconds. At the end of the discharge, calculate the current decay slope based on the current change trend from 30 amps to 5 amps.

[0018] In terms of temperature sequence, assuming that the temperature at the center of the battery surface rises faster and slower at the edge, the temperature gradient distribution is obtained through data from multiple collection points. If the temperature rises from 30 degrees Celsius to 35 degrees Celsius within 10 minutes, the temperature rise rate is 0.5 degrees Celsius / minute. If the temperature rise rate threshold is 0.4 degrees Celsius / minute, it can be determined whether it is exceeded. The temperature standard deviation of each position is calculated to obtain the temperature balance index.

[0019] From the timestamp sequence, the interval from a complete charge to the next charge is 2.5 hours, which is the length of the charge and discharge cycle. The 15 minutes of rest from the end of discharge to the start of the next charge is the length of the rest stage. The cycle interval consistency coefficient is obtained by counting the time intervals of multiple charge and discharge cycles.

[0020] These features are normalized and spliced ​​to generate the original feature set. Then dynamic time series processing is performed to fill missing values ​​and filter noise for each feature dimension of the original feature set. For example, the current charge and discharge efficiency may have missing values ​​due to collection interference, which are filled and filtered using interpolation. According to the statistical distribution characteristics of the features, relatively stable features such as the charge and discharge cycle length are classified as static feature subsets, and dynamically changing features such as the current charge and discharge efficiency are classified as dynamic feature subsets.

[0021] The static feature subset is subjected to mean smoothing with a sliding window of 3 periods. For the dynamic feature subset, taking the temperature gradient distribution as an example, a first-order seasonal difference operation is performed to eliminate the trend component and retain the fluctuation component, and then a fast Fourier transform is performed to obtain the dominant frequency component to construct a reference phase signal. After calculating the phase offset, time domain translation compensation is performed to make the phase difference less than the preset value. Finally, amplitude normalization and timestamp resampling are performed to obtain the dynamic alignment feature subset. The static smooth feature subset and the dynamic alignment feature subset are fused across dimensions according to the timestamp sequence to obtain a standardized feature matrix.

[0022] Step S130, inputting the standardized feature matrix into a pre-trained battery state monitoring model to generate degradation state parameters, health state parameters and remaining life prediction parameters of the target battery.

[0023] In this embodiment, the standardized feature matrix is ​​input into the pre-trained battery state monitoring model. The convolutional temporal coding layer captures the local pattern to generate the primary spatiotemporal feature map. Taking the electric vehicle battery as an example, this layer can identify the local change pattern of the battery characteristics under different driving modes.

[0024] The attention weight allocation layer strengthens the key time points of the primary spatiotemporal feature map. For example, the time point when the battery is deeply discharged is critical to judging the battery status. This layer will increase the weight of the feature corresponding to this time point and generate a weighted spatiotemporal feature map.

[0025] The bidirectional recurrent prediction layer divides the weighted spatiotemporal feature map into time steps. Assuming the time step is 5 seconds, the forward input sequence is the process from charging to discharging, and the reverse is the process from discharging to charging. The degradation increment of each time step is calculated by the forward recurrent neural network unit and accumulated to obtain the degradation accumulation. The reverse recurrent neural network unit calculates the life residual component and weightedly aggregates it to obtain the life residual coefficient. At the same time, cross-layer connection and nonlinear transformation are performed to output the correction factor.

[0026] In the multi-task output layer, the degradation accumulation and the life residual coefficient are linearly weighted and spliced, and the cross-channel feature is scaled to obtain the standardized feature. The standardized degradation feature is input into the first fully connected branch, and the initial degradation state parameters are generated through the degradation state activation function. The standardized residual feature is input into the second fully connected branch to generate the initial life prediction parameters. The cross-attention interaction between the two generates the health state interaction weight matrix, and the standardized degradation feature is weighted and aggregated to obtain the health state intermediate vector, which is input into the third fully connected branch and the initial health state parameters are generated through the health state activation function. Perform operations such as timing alignment compensation, residual coupling, nonlinear interpolation fusion, and correction module adjustment, and finally perform output consistency constraints to obtain the final degradation state parameters, health state parameters, and remaining life prediction parameters.

[0027] Step S140, generating an abnormality detection result and a maintenance decision suggestion for the target battery according to the degradation state parameter, health state parameter and remaining life prediction parameter.

[0028] In this embodiment, the degradation state parameter is compared with the preset degradation threshold interval, for example, the preset interval is [0, 0.2]. If the calculated value is 0.25, the first abnormal mark is generated if it exceeds the interval. The trend deviation of the health state parameter and the historical health state baseline is calculated. If the dynamic deviation threshold is set to 0.15 and the calculated value is 0.2, the second abnormal mark is generated. The remaining life prediction parameter is compared with the median of the battery life distribution of the same batch. If the median of the same batch is 3 years, the life warning quantile is 0.4, and the predicted remaining life is 1 year, the third abnormal mark is generated.

[0029] According to the maintenance strategy knowledge base matched with the abnormal mark combination pattern, if the first and third abnormal marks are present at the same time, a high maintenance priority may be given, the maintenance operation type is to replace the battery cell, and the maintenance time window is a maintenance decision recommendation within 72 hours.

[0030] Step S150, based on the abnormality detection result and the maintenance decision suggestion, adaptively adjust the parameter weights of the battery status monitoring model to update the prediction accuracy of the battery status monitoring model.

[0031] For example, according to the type of abnormal mark, the weight update rule is selected from the model parameter adjustment strategy library. If the first abnormal mark appears, the weight of the convolutional temporal coding layer related to the degradation state is adjusted. The learning rate and regularization strength are determined according to the maintenance priority. If the maintenance priority is medium, the learning rate is adjusted from 0.001 to 0.003, and the regularization strength is adjusted from 0.05 to 0.1.

[0032] A virtual feedback dataset is constructed based on the maintenance operation type and maintenance time window. For example, if the maintenance operation is balanced charging and the time window is 48 hours, a virtual feedback dataset for this period is constructed and mixed with the historical training data in a certain ratio (such as 1:4). Based on the mixed sampling data, the updated learning rate and regularization strength, the parameter weights of the convolutional temporal encoding layer, the attention weight allocation layer, and the bidirectional recurrent prediction layer are back-propagated and optimized. At the same time, the parameter weights of the multi-task output layer are frozen and iteratively updated until the prediction error converges.

[0033] Based on the above steps, the embodiment of the present application first collects real-time operating data such as voltage, current, temperature and timestamp sequence in the charge and discharge cycle of the target battery in the data acquisition stage, which can fully and dynamically reflect the actual operating status of the battery under different working conditions. Secondly, the multi-dimensional feature extraction and dynamic timing processing link, by extracting multi-dimensional features of real-time operation data, generates an original feature set, and can deeply mine the key information hidden in the data. The dynamic timing processing further integrates these original features into a standardized feature matrix, which not only fully considers the time series characteristics of the battery operation data, but also can adapt to the dynamic changes of battery performance over time. Compared with the traditional fixed mode feature processing method, it can better capture the subtle changes in the battery state, greatly improving the effectiveness and representativeness of the features. Furthermore, the standardized feature matrix is ​​input into the pre-trained battery state monitoring model to accurately generate degradation state parameters, health state parameters and remaining life prediction parameters. The battery state monitoring model is trained based on a large amount of data, has a high degree of generalization ability and accuracy, can effectively parse complex battery state information, and provides a quantitative and reliable basis for the full life cycle management of the battery. In terms of generating abnormal detection results and maintenance decision recommendations, based on comprehensive and accurate parameter analysis, it can quickly and accurately identify abnormal conditions in battery operation, and generate practical maintenance decision recommendations based on different battery states and potential risks, realizing personalized and intelligent maintenance management based on the actual state of the battery, greatly improving maintenance efficiency, reducing maintenance costs, and extending the service life of the battery. Finally, based on the abnormal detection results and maintenance decision recommendations, the parameter weights of the battery state monitoring model are adaptively adjusted, which can continuously optimize the performance of the battery state monitoring model according to the actual monitoring results and update the prediction accuracy in real time. Compared with the traditional fixed parameter model, this method can dynamically adapt to the complex changes in the battery state, continuously maintain high-precision monitoring and prediction capabilities, effectively avoid monitoring errors caused by model aging or gradual changes in battery performance, and provide a strong guarantee for long-term stable monitoring of the battery state.

[0034] In a possible implementation, step S120 includes: Step S121, extracting the voltage fluctuation extreme value, the voltage change rate mean value and the voltage platform duration from the voltage sequence.

[0035] In this embodiment, during the driving and charging process of the electric vehicle, the voltage sequence of the battery carries important battery status information. When the voltage fluctuation extreme value is extracted from the voltage sequence, when the car is driving, such as during acceleration and deceleration, the battery voltage will fluctuate. In a specific driving process, by analyzing the collected voltage sequence, the maximum voltage value detected is 4.2 volts, the minimum voltage value is 3.8 volts, and the difference between the two, 0.4 volts, is the voltage fluctuation extreme value. When calculating the mean value of the voltage change rate, the voltage sequence is firstly subjected to a first-order difference operation to obtain a differential voltage sequence. In a specific 5-second sliding time window, for example, during the stable driving of the car, the values ​​of the differential voltage sequence are 0.01 volts, 0.02 volts, -0.01 volts, etc., and the arithmetic mean of these values, 0.01 volts, is determined as the mean value of the voltage change rate. When detecting the duration of the voltage platform, when the car is in the parked charging state, at a certain stage of charging, the voltage change rate continues to be lower than the platform threshold of 0.01 volts / second, and the maximum continuous time period is 20 seconds. This 20 seconds is the duration of the voltage platform.

[0036] Step S122, extracting the current charge and discharge efficiency, the current peak interval period and the current decay slope from the current sequence.

[0037] When extracting features from the current sequence, for the determination of the current charge and discharge efficiency, in a complete charge-discharge cycle, assuming that the input power during charging is 1000 watt-hours and the output power during discharge is 850 watt-hours, the current charge and discharge efficiency is calculated to be 850÷1000×100%=85%. For the current peak interval period, during the driving process of the car, the current will change with the power demand of the vehicle, and the peak points in the current sequence are identified. For example, the first current peak appears in the 8th second after the car starts, and the second current peak appears in the 20th second when the car accelerates sharply, then the current peak interval period is 20-8=12 seconds. At the end of discharge, as the battery power decreases, the current will gradually decay. According to the current change trend from 30 amperes to 5 amperes, the current decay slope is obtained through relevant mathematical calculations.

[0038] Step S123, extracting the temperature gradient distribution, the temperature rise rate threshold and the temperature balance index from the temperature sequence.

[0039] During the battery charging and discharging process, the temperature changes at different positions of the battery are different. Because the temperature at the center of the battery rises faster and the temperature at the edge rises slower when the battery is discharging at high power (such as when the car is driving at high speed or accelerating suddenly) or charging quickly, the temperature gradient distribution is constructed through data from multiple temperature collection points. During the charging process, if the temperature rises from 30 degrees Celsius to 35 degrees Celsius within 10 minutes, the calculated temperature rise rate is (35-30) ÷ 10 = 0.5 degrees Celsius / minute. If the pre-set temperature rise rate threshold is 0.4 degrees Celsius / minute, it can be determined whether the threshold is exceeded. Calculate the standard deviation of the temperature at each position of the battery. If the standard deviation is small, for example 0.2 degrees Celsius, it means that the temperature balance is good. This 0.2 degrees Celsius represents the temperature balance index.

[0040] Step S124, extracting the charge and discharge cycle length, the rest phase duration and the cycle interval consistency coefficient from the timestamp sequence.

[0041] The calculation of the length of the charge and discharge cycle is based on the time interval from the start of a complete charge to the start of the next charge. For example, the time interval from the completion of a charge to the start of the next charge is 2.5 hours, which is the length of the charge and discharge cycle. The length of the rest phase refers to the rest time of the battery between charge and discharge. For example, after the car is discharged, it is not charged immediately for some reason, and it is left to rest for 15 minutes before charging. This 15 minutes is the length of the rest phase. The cycle interval consistency coefficient reflects the consistency of the time interval between each charge and discharge cycle. It is obtained by counting the time intervals of multiple charge and discharge cycles and calculating the ratio of the standard deviation to the mean. It is assumed that the coefficient is 0.1 after multiple statistical calculations.

[0042] Step S125, normalize and splice the voltage fluctuation extreme value, voltage change rate mean, voltage platform duration, current charge and discharge efficiency, current peak interval period, current decay slope, temperature gradient distribution, temperature rise rate threshold, temperature balance index, charge and discharge cycle length, static stage length and cycle interval consistency coefficient to generate the original feature set.

[0043] Step S126, performing missing value filling and noise filtering processing on each feature dimension in the original feature set to generate a denoised feature set.

[0044] Each feature dimension in the original feature set may have incomplete data or be affected by noise. For each feature dimension, missing value filling and noise filtering are performed to generate a denoised feature set. For example, in the feature dimension of current charge and discharge efficiency, the data collection at certain times may be inaccurate due to external electromagnetic interference or short-term failure of the acquisition equipment, resulting in missing values. Through interpolation methods, such as linear interpolation, the missing values ​​are filled according to the values ​​of the surrounding valid data points. At the same time, for possible noise, such as small fluctuations in current data caused by electromagnetic interference from the internal electrical system of the car, a filtering algorithm, such as a low-pass filtering algorithm, is used to remove high-frequency noise, thereby obtaining the denoised current charge and discharge efficiency feature. For other feature dimensions, such as the mean value of the voltage change rate, the temperature balance index, etc., the same missing value filling and noise filtering processing methods are also used to finally generate a denoised feature set.

[0045] Step S127: dividing the denoising feature set into a static feature subset and a dynamic feature subset according to the statistical distribution characteristics of each feature dimension in the denoising feature set.

[0046] For example, features such as the length of the charge and discharge cycle and the length of the static phase are relatively stable during the normal use of electric vehicles, and their values ​​will not change dramatically with each charge and discharge process. These features are divided into static feature subsets. Features such as current charge and discharge efficiency and temperature gradient distribution will have obvious dynamic changes under different driving conditions (such as constant speed driving, acceleration, deceleration) and charge and discharge conditions of the car. These features are divided into dynamic feature subsets.

[0047] Step S128, performing sliding window mean smoothing processing on the static feature subset to generate a static smooth feature subset.

[0048] Taking the length of the charge and discharge cycle as an example, set the sliding window to 3 cycles and calculate the mean in each window. Assuming that the lengths of the charge and discharge cycles in the first window are 2.5 hours, 2.4 hours, and 2.6 hours, respectively, the mean is (2.5+2.4+2.6)÷3=2.5 hours. Calculate the mean in each window in this way to obtain the smoothed charge and discharge cycle length value. Perform the same processing on other static features to finally generate a static smooth feature subset.

[0049] Step S129, performing time series difference transformation and phase alignment processing on the dynamic feature subset to generate a dynamic alignment feature subset, cross-dimensionally fusing the static smoothing feature subset with the dynamic alignment feature subset according to the timestamp sequence to generate the standardized feature matrix.

[0050] Taking the dynamic feature of temperature gradient distribution as an example, firstly, a first-order seasonal difference operation is performed to eliminate the trend component and retain the periodic fluctuation component. During the driving process of the car, the battery temperature gradient distribution may be affected by various factors such as ambient temperature and driving conditions, and there are certain trend changes. This trend is removed by differential operation. Then a fast Fourier transform is performed to extract the dominant frequency component. Assuming that the frequency corresponding to the dominant frequency component is 0.1 Hz, a reference phase signal is constructed based on this dominant frequency component. The cyclic phase difference between the phase component of the characteristic dimension of temperature gradient distribution and the reference phase signal is calculated to obtain the phase offset. Due to various complex situations during the driving process of the car, the phase of the temperature gradient distribution collected at different times may be offset, and this offset is quantified by calculating the phase offset. According to this phase offset, the temperature gradient distribution is compensated for time domain shift so that the phase difference of its dominant frequency component is less than the preset phase tolerance, for example, the preset phase tolerance is 0.1 radians. Finally, the temperature gradient distribution after phase compensation is amplitude normalized and timestamp resampled to generate a dynamic alignment feature subset. Other dynamic features are also processed according to the same steps.

[0051] Finally, the static smooth feature subset and the dynamic alignment feature subset are cross-dimensionally fused in the timestamp sequence to generate a standardized feature matrix. During the fusion process, the eigenvalues ​​in the static smooth feature subset are combined with the corresponding eigenvalues ​​in the dynamic alignment feature subset in the order of timestamps to construct a new matrix. This matrix is ​​the standardized feature matrix, which integrates the characteristic information of the battery in different aspects, and after processing, it has better consistency and analyzability, preparing for the subsequent input into the battery status monitoring model.

[0052] In a possible implementation, step S130 includes: Step S131, capturing local patterns of the standardized feature matrix through the convolutional temporal coding layer in the battery status monitoring model to generate a primary spatiotemporal feature map.

[0053] In this embodiment, when the standardized feature matrix is ​​input into the convolutional time-series coding layer in the battery state monitoring model, the convolutional time-series coding layer is intended to capture the local pattern of the standardized feature matrix to generate a primary spatiotemporal feature map. In the actual operation of electric vehicles, the standardized feature matrix contains a lot of information related to the battery state, such as voltage fluctuation extremes, current charge and discharge efficiency, and other data at different times and driving conditions. The convolution kernel in the convolutional time-series coding layer is like a specific filter, sliding on the matrix according to a certain step size to analyze different local areas. Taking an electric vehicle in a long-distance driving process as an example, the driving process includes different road conditions such as flat roads, climbing, downhill, and different driving behaviors such as uniform driving, acceleration, deceleration, etc. In this process, the various characteristics of the battery will present specific local patterns at different times and working conditions. The convolutional time-series coding layer can identify the relationship patterns between characteristics such as voltage, current, and temperature under specific working conditions, such as when driving at high speed and the battery is in high-power discharge, a specific combination pattern of characteristics such as voltage, current, and temperature. In this way, local spatiotemporal features related to the battery state are extracted from the standardized feature matrix, and then a primary spatiotemporal feature map is generated, which can preliminarily reflect the combination of battery state features at different times and spaces.

[0054] Step S132, through the attention weight allocation layer in the battery status monitoring model, the key time points of the primary spatiotemporal feature graph are enhanced to generate a weighted spatiotemporal feature graph.

[0055] In the entire operation cycle of electric vehicles, there are some key time points that are of great significance to the judgment of battery status. For example, in the deep discharge stage of the battery, the changes in the battery's voltage, temperature and other characteristics have a great impact on the battery's health status and remaining life; or in the fast charging stage, the inflow of large currents may cause the chemical reactions inside the battery to intensify. The features corresponding to these time points are crucial for accurately judging the battery status. The attention weight allocation layer strengthens the weights of the features corresponding to these key time points in the primary spatiotemporal feature map based on the weights learned in the pre-trained model. For example, in the deep discharge stage, the weights of the voltage, current, temperature and other features of the corresponding primary spatiotemporal feature map will be increased, making these features more prominent in subsequent analysis, thereby generating a weighted spatiotemporal feature map. Such a weighted spatiotemporal feature map can focus more on feature information that is important for judging the battery status.

[0056] Step S133, performing forward cumulative degradation analysis and reverse life residual fitting on the weighted spatiotemporal feature graph through the bidirectional cyclic prediction layer in the battery state monitoring model to generate degradation accumulation and life residual coefficient.

[0057] For example, the weighted spatiotemporal feature graph is segmented according to the time step. Taking a complete charge and discharge cycle of an electric vehicle as an example, assuming that the time step is set to 5 seconds, the forward input sequence is a feature sequence in chronological order from the start of charging to the end of discharging, and the reverse input sequence is a feature sequence in reverse order from the end of discharging to the start of charging. The forward input sequence is recursively hidden in the forward recurrent neural network unit, and the degradation increment is calculated according to the various characteristics of the battery in each time step. For example, during the normal driving and discharging process of the car, the capacity of the battery will gradually decrease over time. According to the voltage, current and other characteristics in the weighted spatiotemporal feature graph, the capacity reduction in each 5-second time step is calculated as the degradation increment, and then these degradation increments are accumulated to obtain the degradation accumulation. At the same time, the reverse input sequence is backtracked in the hidden state through the reverse recurrent neural network unit, and the life residual component is calculated in each time step. For example, when analyzing the charging process in reverse, the life residual component corresponding to each time step is calculated according to the characteristic changes of the battery during the charging process, and then these life residual components are weighted and aggregated according to certain weights to obtain the life residual coefficient. In addition, the final hidden state of the forward recurrent neural network unit is cross-layer connected with the initial hidden state of the reverse recurrent neural network unit to generate a joint hidden state vector, which is then nonlinearly transformed through a fully connected layer to output correction factors for the degradation accumulation and the life residual coefficient to further improve the accuracy of these two parameters.

[0058] Step S134, jointly mapping the degradation accumulation and the life residual coefficient through the multi-task output layer in the battery state monitoring model, and synchronously outputting the degradation state parameter, health state parameter and remaining life prediction parameter.

[0059] For example, in a possible implementation, step S134 includes: Step S1341, linearly weighted concatenating the degradation accumulation and the life residual coefficient according to the time step to generate a degradation-residual joint feature vector, and performing cross-channel feature scaling on the degradation-residual joint feature vector to generate a standardized degradation feature and a standardized residual feature.

[0060] In this embodiment, during the operation of the electric vehicle, the time step is an important time interval unit for analyzing the battery state. For example, with every 5 minutes as a time step, at different time steps, the degradation accumulation reflects the degree of performance decline of the battery during long-term use, and the life residual coefficient reflects the residual features related to the predicted life. By linearly weighting the degradation accumulation and the life residual coefficient at each time step according to a specific weight, a degradation-residual joint feature vector is formed, which contains comprehensive information related to battery degradation and life. However, since the dimensions and numerical ranges of different features may be different, in order to make these features comparable in subsequent analysis, cross-channel feature scaling is required. Taking the voltage-related features and temperature-related features of the battery as an example, they may differ in the numerical range and the weight of the impact on the battery state. Cross-channel feature scaling can adjust these different features to a suitable numerical range, thereby generating standardized degradation features and standardized residual features. These two standardized features can more accurately reflect the state information of the battery in terms of degradation and life, laying the foundation for subsequent analysis.

[0061] Step S1342, input the standardized degradation feature into the first fully connected branch, generate initial degradation state parameters through the degradation state activation function, and input the standardized residual feature into the second fully connected branch, generate initial life prediction parameters through the life residual activation function.

[0062] In the actual use of electric vehicles, the initial degradation state parameters of the battery reflect the performance changes of the battery from the initial state to the current state. For example, when an electric vehicle has traveled a certain mileage and has undergone multiple charge and discharge cycles, the chemical substances inside the battery will be lost and the electrode structure may change. These factors will cause the battery performance to deteriorate. The standardized degradation features contain these degradation-related information and are processed by the degradation state activation function in the first fully connected branch. The activation function performs a nonlinear transformation on the input standardized degradation features according to a pre-set mathematical model to generate initial degradation state parameters. The initial degradation state parameters can intuitively represent the degree of degradation of the battery in the current state relative to the initial state. For example, a numerical value can be used to represent the attenuation ratio of the battery capacity or the degree of increase in the internal resistance of the battery. At the same time, the standardized residual features are input into the second fully connected branch, and the initial life prediction parameters are generated through the life residual activation function. In the use of electric vehicles, the remaining life of the battery is a key indicator. The standardized residual features contain information related to the remaining life of the battery, such as the remaining power of the battery under different working conditions, the change in the charge and discharge efficiency of the battery, etc. The life residual activation function in the second fully connected branch is used to process the normalized residual features according to the pre-learned model to generate initial life prediction parameters. This initial life prediction parameter can preliminarily estimate the remaining time or remaining number of charge and discharge cycles that the battery can be used normally in the current state.

[0063] Step S1343, cross-attention interaction is performed on the standardized degradation features and the standardized residual features to generate a health state interaction weight matrix, and dynamic weighted aggregation is performed on the standardized degradation features according to the health state interaction weight matrix to generate a health state intermediate vector.

[0064] In detail, the health state of the battery is a comprehensive concept that is affected by multiple factors. There is a complex relationship between the standardized degradation features and the standardized residual features. Through the cross-attention interaction mechanism, the mutual influence relationship between these features can be captured to generate a health state interaction weight matrix. For example, in some cases, the degree of battery degradation may have a greater impact on the remaining life, which corresponds to a higher weight in the health state interaction weight matrix. According to this health state interaction weight matrix, the standardized degradation features are dynamically weighted and aggregated. Taking the voltage fluctuation and current change of the battery under different charge and discharge states as an example, these degradation-related features are weighted and summed according to the weights in the weight matrix to generate a health state intermediate vector, which comprehensively considers the impact of the battery degradation and remaining life-related features on the health state, and can more comprehensively reflect the battery health state information.

[0065] Step S1344, input the health state intermediate vector into the third fully connected branch, generate initial health state parameters through the health state activation function, perform time alignment compensation on the initial degradation state parameters and the initial health state parameters, generate compensated degradation state parameters and compensated health state parameters, perform residual coupling on the compensated degradation state parameters and the initial life prediction parameters, and generate coupled life prediction parameters.

[0066] In the actual use scenario of electric vehicles, this initial health state parameter can comprehensively reflect the health status of multiple aspects of the battery. For example, the comprehensive status of the battery's internal resistance, available effective capacity, and charging and discharging efficiency. The health state activation function processes the health state intermediate vector according to a predetermined mathematical model and converts it into an initial health state parameter. The initial health state parameter can be used as a value or a group of values ​​to represent the health of the battery. For example, a health state scoring system can be set to divide the battery's health state into different levels according to the initial health state parameter. Then, the initial degradation state parameter and the initial health state parameter are aligned and compensated in time to generate the compensated degradation state parameter and the compensated health state parameter. During the operation of electric vehicles, since the degradation and health state of the battery have certain correlations and change rules in time, it is necessary to ensure the consistency of these two parameters in the time dimension. For example, under different mileages and charge and discharge cycles, the changes in the degradation state and health state of the battery may have a certain time delay or advance. Through the timing alignment compensation mechanism, the initial degradation state parameters and initial health state parameters are adjusted according to the actual operation data of the battery and the pre-learned time relationship model to generate compensated degradation state parameters and compensated health state parameters. These compensated parameters can more accurately reflect the degradation and health state of the battery in the same time dimension.

[0067] Step S1345, nonlinearly interpolating and fusing the compensated healthy state parameters and the coupled life prediction parameters to generate interpolated health-life association parameters, and inputting the compensated degradation state parameters into a degradation state correction module, and adjusting the output corrected degradation state parameters through the degradation correction coefficient.

[0068] There is an intrinsic connection between the degradation state and the remaining life of the battery. The compensated degradation state parameter reflects the actual degradation of the battery, and the initial life prediction parameter is a preliminary estimate of the remaining life of the battery. Through the residual coupling mechanism, these two parameters are organically combined. For example, when the degree of battery degradation is high, it may have a greater impact on the remaining life. This impact is taken into account through a specific mathematical model to generate a coupled life prediction parameter. This parameter can more accurately reflect the remaining life of the battery after considering the degradation state. Then, the compensated health state parameter and the coupled life prediction parameter are nonlinearly interpolated and fused to generate an interpolated health-life association parameter. In the actual operation of electric vehicles, the relationship between the health state and the remaining life of the battery is not a simple linear relationship, but a complex nonlinear relationship. Through the nonlinear interpolation fusion mechanism, the compensated health state parameter and the coupled life prediction parameter are fused according to the pre-learned nonlinear model to generate an interpolated health-life association parameter. This parameter can more comprehensively reflect the association between the health state and the remaining life of the battery.

[0069] Step S1346, input the interpolated health-life association parameters into the health status correction module, adjust the output corrected health status parameters through the health correction coefficient, and input the coupled life prediction parameters into the life prediction correction module, adjust the output corrected remaining life prediction parameters through the life correction coefficient.

[0070] Due to the influence of various factors, such as measurement errors, environmental changes, etc., the degradation state parameters obtained by the initial calculation may have certain deviations. The degradation state correction module adjusts the compensated degradation state parameters according to the predetermined degradation correction coefficient. For example, when it is found that the degradation rate of the battery under certain specific working conditions deviates from the model prediction, the degradation state parameters after compensation can be corrected by adjusting the degradation correction coefficient to obtain more accurate corrected degradation state parameters. Similarly, the interpolated health-life association parameters are input into the health state correction module, and the output corrected health state parameters are adjusted by the health correction coefficient, and the coupled life prediction parameters are input into the life prediction correction module, and the output corrected remaining life prediction parameters are adjusted by the life correction coefficient. These corrected parameters can further improve the accurate description of the battery state.

[0071] Step S1347, output consistency constraints are applied to the corrected degradation state parameters, corrected health state parameters and corrected remaining life prediction parameters to generate degradation state parameters, health state parameters and remaining life prediction parameters.

[0072] The output consistency constraints include: The corrected health status parameter and the corrected remaining life prediction parameter are reversely correlated. If the health status parameter is lower than the health threshold and the remaining life prediction parameter is simultaneously lower than the life threshold, it is determined to be consistent.

[0073] If the threshold states of the health state parameter and the remaining life prediction parameter are inconsistent, the corrected health state parameter and the corrected remaining life prediction parameter are dynamically scaled until a reverse correlation condition is met.

[0074] The corrected degradation state parameter and the corrected health state parameter are trend matched and verified. If the increasing rate of the degradation state parameter exceeds the decreasing rate of the health state parameter, a rate limiting factor is applied to the corrected degradation state parameter to force matching with the decreasing trend of the health state parameter.

[0075] The parameter combination that has been verified by reverse correlation and trend matching is used as the final output degradation state parameters, health state parameters and remaining life prediction parameters.

[0076] In detail, in the battery management of electric vehicles, if the health state parameter is lower than the health threshold and the remaining life prediction parameter is simultaneously lower than the life threshold, for example, the health state parameter indicates that the effective capacity of the battery is lower than the set health capacity lower limit, and the remaining life prediction parameter indicates that the remaining number of charge and discharge cycles of the battery is also lower than the set life lower limit, then it is determined to be consistent. However, if the threshold state of the health state parameter and the remaining life prediction parameter are inconsistent, for example, the health state parameter shows that the battery is still in a good state, but the remaining life prediction parameter shows that the remaining life is extremely low, in this case, the corrected health state parameter and the corrected remaining life prediction parameter are dynamically scaled. According to the pre-set scaling rules and the actual operating data of the battery, the two parameters are adjusted until the reverse correlation condition is met. At the same time, the corrected degradation state parameter and the corrected health state parameter are trend matched and verified. During the use of electric vehicles, as the battery is used, the degradation state parameter should show a certain trend relationship with the health state parameter. If the increasing rate of the degradation state parameter exceeds the decreasing rate of the health state parameter, for example, the battery degrades too fast and the health state parameter does not keep up, in this case, a rate limiting factor is applied to the corrected degradation state parameter to force the matching of the decreasing trend of the health state parameter. Through these output consistency constraint operations, the parameter combination that has been verified by reverse correlation and trend matching is used as the final output degradation state parameter, health state parameter and remaining life prediction parameter. These final parameters can accurately reflect the actual state of the electric vehicle power battery and provide a reliable basis for battery management, maintenance and decision-making.

[0077] In a possible implementation, step S140 includes: Step S141 : comparing the degradation state parameter with a preset degradation threshold interval, and generating a first abnormality mark if the degradation state parameter exceeds the degradation threshold interval.

[0078] For example, in the battery management system of an electric vehicle, the preset degradation threshold interval is determined based on the performance degradation data of a large number of batteries of the same type under normal use conditions. Assume that for a certain type of electric vehicle power battery, the preset degradation threshold interval is [0, 0.3]. This interval represents the acceptable degradation range of the battery during normal use, such as after a certain mileage of driving and charge and discharge cycles. When the degradation state parameter of the battery is 0.35, it exceeds this set threshold interval, indicating that the performance degradation of the battery has exceeded normal expectations. This excess may be caused by excessive loss of chemical substances inside the battery, severe damage to the electrode structure, or failure of the battery management system. At this time, a first abnormal mark is generated, which serves as a signal of abnormal degradation of the battery and provides a basis for subsequent maintenance decisions.

[0079] Step S142, calculating the trend deviation between the health status parameter and the historical health status baseline, and if the trend deviation exceeds a dynamic deviation threshold, generating a second abnormal mark.

[0080] The health status of the battery is a key indicator throughout the life cycle of electric vehicles. The historical health status baseline is obtained by statistically analyzing a large amount of health status data of the same model of batteries under normal use. For example, in the early stage of normal battery use, the health status parameters may remain at a relatively stable level. With the increase in mileage and the number of charge and discharge cycles, the health status will gradually decline, but this decline process has certain rules. Assume that the historical health status baseline is a health status value range or change trend curve determined according to different use stages. When calculating the trend deviation between the current battery health status parameters and the historical health status baseline, multiple factors need to be considered, such as the change trends of the battery's effective capacity, charge and discharge efficiency, internal resistance, etc. For example, if the dynamic deviation threshold is set to 0.2, if the calculated trend deviation is 0.25, it means that the battery's health status has deviated significantly from the historical health status under normal conditions. This deviation may be caused by factors such as abnormal temperature effects, excessive charge and discharge, or imbalance between battery cells during use. Once the trend deviation exceeds the dynamic deviation threshold, a second abnormal mark is generated, which indicates that the battery's health status has an abnormal change trend and requires further attention and processing.

[0081] Step S143, performing relative position analysis on the remaining life prediction parameter and the median of the life distribution of batteries in the same batch, and generating a third abnormal mark if the remaining life prediction parameter is lower than the life warning quantile.

[0082] In detail, batteries from the same batch have similar performance and life characteristics because they follow the same manufacturing process and quality standards during production. In the battery management of electric vehicles, the median life distribution can be obtained by performing life tests and statistical analysis on a large number of samples of batteries from the same batch. For example, for a batch of electric vehicle power batteries, after actual use and life monitoring of many battery samples, it was found that the median life distribution was 5 years, indicating that under normal use conditions, half of the batteries in this batch will have a battery life of more than 5 years, and the other half will be less than 5 years. At the same time, the life warning quantile is set to 0.3, which means that when the remaining life prediction parameter of the battery is lower than 30% of the median life distribution of the same batch of batteries, an early warning is required. Assuming that the remaining life prediction parameter of the current battery is 1.5 years, and the median life distribution of the same batch of batteries is 5 years, 1.5 years is lower than 30% of 5 years (i.e. 1.5 years <5×0.3=1.5 years). In this case, the system will generate a third abnormal mark, indicating that the remaining life of the battery is already at a low level and may be about to reach the service life limit, and timely measures need to be taken.

[0083] Step S144, according to the combination pattern of the first abnormal mark, the second abnormal mark and the third abnormal mark, matching the preset maintenance strategy knowledge base, generating the maintenance decision suggestion including the maintenance priority, the maintenance operation type and the maintenance time window.

[0084] In detail, the preset maintenance strategy knowledge base is formulated according to different abnormal situation combinations and the actual use needs and safety requirements of the battery. For example, if only the first abnormal mark appears, it indicates that the battery is mainly abnormal in degradation state. The possible maintenance decision suggestion is medium maintenance priority. The maintenance operation type may be to perform a deep charge and discharge cycle to try to restore battery performance. The maintenance time window can be set to be performed within the next week. If the first abnormal mark and the third abnormal mark appear at the same time, this indicates that the battery not only has degradation abnormalities, but also has a low remaining life. The maintenance decision suggestion at this time may be high maintenance priority. The maintenance operation type may be to replace the battery module or conduct a comprehensive inspection and repair of the battery. The maintenance time window needs to be more urgent, such as within the next 24 hours. If all three abnormal marks appear, this means that the battery is in a very bad state and may need to stop using the vehicle immediately and perform emergency maintenance. The maintenance priority is the highest, the maintenance operation type may be to directly replace the entire battery pack, and the maintenance time window is to be executed immediately. By matching the maintenance strategy knowledge base according to the abnormal mark combination pattern, reasonable and effective maintenance decision suggestions can be provided for the power battery of electric vehicles to ensure the safety, reliability and service life of the battery.

[0085] In a possible implementation, step S150 includes: Step S151, selecting a corresponding weight update rule from a model parameter adjustment strategy library according to the abnormality mark type in the abnormality detection result.

[0086] In detail, the model parameter adjustment strategy library is pre-established, which contains weight update strategies for different types of abnormal markers. For example, when only the first abnormal marker appears in the abnormal detection result (that is, the degradation state parameter exceeds the preset degradation threshold interval), this indicates that the degradation of the battery is abnormal. The weight update rules selected from the model parameter adjustment strategy library may focus on adjusting the model parameter weights related to the degradation state. In the battery state monitoring model, the convolutional time series coding layer is responsible for capturing local patterns in the battery state data, which plays an important role in detecting the degradation characteristics of the battery. Therefore, the weight update rule for this type of abnormal marker may increase the adjustment amplitude of the convolution kernel weights related to the degradation characteristics in the convolutional time series coding layer. For example, if a convolution kernel is mainly used to detect the relationship between battery voltage fluctuation and degradation degree, when the first abnormal marker appears, the weight adjustment amplitude of the convolution kernel may be larger than normal so that the model can capture the degradation characteristics of the battery more accurately.

[0087] Step S152, determining the learning rate adjustment range and regularization strength of the battery status monitoring model according to the maintenance priority in the maintenance decision suggestion.

[0088] For the case where the first abnormal mark and the third abnormal mark appear at the same time (that is, the degradation state is abnormal and the remaining life prediction parameter is lower than the life warning quantile), the weight update rules in the model parameter adjustment strategy library may be more comprehensive. Not only will the weight adjustment related to degradation in the convolutional temporal coding layer be increased, but also the parameter weights related to the remaining life prediction in the bidirectional loop prediction layer will be adjusted. Because this abnormal combination indicates that the battery not only has the current degradation problem, but also involves a serious shortening of the remaining life, it is necessary to adjust the model parameter weights related to these two aspects at the same time to improve the model's prediction ability for such complex situations.

[0089] In detail, the maintenance priority reflects the urgency and importance of the battery status. If the maintenance priority in the maintenance decision recommendation is high, it means that the battery status is already very urgent and the model needs to be adjusted as soon as possible to improve the prediction accuracy. At this time, in order to speed up the convergence of the model, the learning rate will be increased significantly. For example, the original learning rate is 0.001, and in the case of high maintenance priority, the learning rate may be increased to 0.005. At the same time, in order to prevent the model from overfitting during rapid adjustment, the regularization strength will also be appropriately enhanced. Assuming that the original regularization strength is 0.1, it may be increased to 0.2. A higher regularization strength can limit the excessive growth of model parameters and ensure that the model can maintain good generalization ability while quickly learning new data features.

[0090] If the maintenance priority is medium, the learning rate will be adjusted relatively small, such as from 0.001 to 0.002, and the regularization strength will only be adjusted moderately, such as from 0.1 to 0.15. This adjustment range can not only optimize the model to a certain extent according to the new data information, but also avoid excessive adjustment that may lead to unstable model performance.

[0091] Step S153: construct a virtual feedback data set according to the maintenance operation type and the maintenance time window, and perform mixed sampling on the virtual feedback data set and historical training data.

[0092] Different maintenance operation types and maintenance time windows will have different effects on the battery status. For example, when the maintenance operation type is deep charge and discharge repair and the maintenance time window is 24 hours, a virtual feedback data set is constructed based on the understanding of the characteristics and expected changes of the battery during the deep charge and discharge repair process. This data set contains the expected change values ​​of various state parameters of the battery during the deep charge and discharge repair, such as the change trend of battery voltage, current, temperature and other parameters during the repair process. At the same time, considering that the maintenance time window is 24 hours, the time series data in the data set will be constructed according to this time range to reflect the state changes of the battery within this specific maintenance operation and time range.

[0093] Assuming that the mixing ratio is 1:3, the purpose of this is to introduce the new information contained in the virtual feedback data set on the basis of using historical training data, so that the model can learn the changing rules of battery status under specific maintenance operations and time windows. For example, the historical training data contains a large amount of battery status data during normal use, while the virtual feedback data set supplements the status data under special maintenance conditions. Through mixed sampling, the model can learn and optimize based on more comprehensive data.

[0094] Step S154, based on the mixed sampled data set, the updated learning rate and the regularization strength, back-propagation optimization is performed on the parameter weights of the convolutional temporal coding layer, the attention weight allocation layer and the bidirectional recurrent prediction layer of the battery state monitoring model.

[0095] In the electric vehicle battery state monitoring model, backpropagation optimization is the process of gradually adjusting the model parameter weights according to the sample errors in the mixed sampled data set. For the convolutional temporal coding layer, taking one of the convolution kernels as an example, during the backpropagation process, the contribution of the convolution kernel to the error is calculated based on the error between the battery state data in the mixed sampled data set and the model prediction result, and then the weight of the convolution kernel is adjusted according to the updated learning rate. For example, if there is a large error between the actual degradation state of a battery sample in the mixed sampled data set and the degradation state predicted by the model, and this error is related to the features extracted by a convolution kernel in the convolutional temporal coding layer, then the weight of this convolution kernel will be adjusted according to the direction and size of the error to reduce the prediction error.

[0096] Step S155, freezing the parameter weights of the multi-task output layer in the battery status monitoring model, and only iteratively updating the parameter weights of the convolutional temporal coding layer, the attention weight allocation layer and the bidirectional recurrent prediction layer until the prediction error converges.

[0097] For the attention weight allocation layer, during the back-propagation optimization process, the attention weight is adjusted according to the importance of the battery status data at different time points in the mixed sampling data set and the difference between the model's attention to these data and the actual situation. For example, if the battery status data at a certain time point in the battery status change after a maintenance operation (such as the voltage change at the beginning of charging) is more important to the judgment of the overall battery status than the model previously believed, then during the back-propagation optimization, the attention weight corresponding to this time point will be increased, so that the model can more accurately pay attention to the status information of these key time points in subsequent predictions.

[0098] For the bidirectional cycle prediction layer, during back propagation optimization, the weights in the bidirectional cycle neural network unit are adjusted according to the error between the forward and reverse state change data of the battery in the mixed sampling data set and the model prediction results. For example, during the battery's charge and discharge cycle, if the model's prediction error for the remaining battery life is large, and this error is related to the weights of the bidirectional cycle prediction layer when calculating the degradation accumulation and life residual coefficient, then these weights will be adjusted according to the error situation to improve the model's prediction accuracy for the remaining battery life.

[0099] During the entire back-propagation optimization process, the parameter weights of the multi-task output layer in the battery state monitoring model are frozen, and only the parameter weights of the convolutional temporal coding layer, the attention weight allocation layer, and the bidirectional loop prediction layer are iteratively updated until the prediction error converges. The parameter weights of the multi-task output layer remain frozen because this layer is mainly responsible for jointly mapping the results of the previous layers and outputting the final degradation state parameters, health state parameters, and remaining life prediction parameters. Its parameter weights are relatively stable in the overall structure of the model and do not need to be adjusted frequently. The parameter weights of the convolutional temporal coding layer, the attention weight allocation layer, and the bidirectional loop prediction layer are related to the core operations such as feature extraction of battery state data, key time point reinforcement, and cumulative degradation analysis. By continuously iteratively updating these weights, the model can better adapt to changes in battery state and improve prediction accuracy. During the iterative update process, the prediction error of the model is continuously monitored. When the prediction error gradually decreases and reaches the convergence standard (such as the change in the prediction error is less than a certain set threshold), the iterative update is stopped. At this time, the prediction accuracy of the model has been updated and improved, and the state of the electric vehicle power battery can be monitored and predicted more accurately.

[0100] In a possible implementation, before step S110, the method further includes: Step S111, configuring a distributed sensor node network connected to the target battery, wherein the distributed sensor node network includes a voltage acquisition node, a current acquisition node, a temperature acquisition node, and a time synchronization node.

[0101] In order to accurately monitor the status of electric vehicle power batteries during the charge and discharge cycle, it is necessary to first configure a distributed sensor node network connected to the target battery. This distributed sensor node network is a system specially built to obtain comprehensive battery status information, which includes voltage acquisition nodes, current acquisition nodes, temperature acquisition nodes, and time synchronization nodes.

[0102] Step S112: capturing the voltage sequence at a first sampling frequency through the voltage acquisition node, and capturing the current sequence at a second sampling frequency through the current acquisition node.

[0103] The voltage acquisition node undertakes the important task of obtaining the battery voltage sequence in the entire network. It captures the battery voltage at the first sampling frequency. During the operation of the electric vehicle, the battery voltage will fluctuate continuously with the vehicle's driving state, charging state, and the performance of the battery itself. For example, when the electric vehicle is in the acceleration stage, the battery needs to provide a larger current. At this time, the chemical reaction inside the battery intensifies, and the battery voltage may drop briefly; while when the vehicle is driving at a constant speed or charging, the battery voltage will show different trends. The voltage acquisition node collects voltage values ​​at a higher sampling frequency, such as every 5 milliseconds, to ensure that these voltage changes can be accurately captured. Through continuous acquisition, a voltage sequence that reflects the change of battery voltage over time is formed. This voltage sequence contains rich battery status information, such as the battery's state of charge, the degree of polarization inside the battery, etc.

[0104] At the same time, the current acquisition node captures the current sequence at a second sampling frequency. In the operation of electric vehicles, the change in current is closely related to the power demand of the vehicle. When the vehicle starts, climbs a slope, or travels at high speed, it requires greater power, and the battery will output a larger current; when the vehicle decelerates or coasts, the battery may be in a charging state (through the regenerative braking system), at which time the current direction will change and flow into the battery. The current acquisition node records these current changes at a specific sampling frequency, such as collecting current values ​​every 3 milliseconds. Such frequent sampling can accurately reflect the charging and discharging conditions of the battery under different working conditions, and the resulting current sequence can be used to analyze important parameters such as the battery's charging and discharging efficiency and battery internal resistance.

[0105] Step S113, capturing the temperature sequence at a third sampling frequency through the temperature acquisition node, and aligning the sampling moments of the voltage acquisition node, the current acquisition node, and the temperature acquisition node at the microsecond level through the time synchronization node.

[0106] The temperature acquisition node captures the temperature sequence at the third sampling frequency. Battery temperature is a key status indicator because the battery generates heat during the charging and discharging process, and excessively high temperatures may affect the performance, life, and even safety of the battery. In the actual operation of electric vehicles, the battery temperature will increase significantly, such as after fast charging or long-term high-speed driving. The temperature acquisition node collects the temperature of different positions of the battery at an appropriate sampling frequency, such as once every 10 milliseconds. Since the temperature distribution inside the battery may not be uniform, multiple temperature acquisition points can obtain the overall temperature information of the battery, and the changes in these collected temperature values ​​over time constitute a temperature sequence. This temperature sequence helps monitor the thermal management of the battery and detect possible local overheating problems.

[0107] In this distributed sensor node network, the time synchronization node plays a vital role. It aligns the sampling moments of the voltage acquisition node, current acquisition node, and temperature acquisition node at the microsecond level. In the complex operating environment of electric vehicles, accurate time synchronization is the key to ensuring the validity and relevance of data. For example, when a vehicle accelerates sharply at a certain moment, the voltage, current, and temperature of the battery will change at the same time. If the sampling moments of these data are not accurately aligned, the relationship between these parameters cannot be accurately determined when the battery status is analyzed later. The time synchronization node ensures that the voltage, current, and temperature values ​​collected at the same time can accurately reflect the true status of the battery at that moment through microsecond alignment.

[0108] Step S114: Send the voltage sequence, current sequence, temperature sequence and timestamp sequence to the edge computing gateway through an encrypted transmission protocol for temporary caching and integrity verification.

[0109] The use of encrypted transmission protocols is to ensure the security and confidentiality of data and prevent data from being tampered with or stolen during transmission. In the environment of electric vehicles, there may be various electromagnetic interferences and potential network security threats. Encrypted transmission can effectively protect the integrity of battery status data. As a data transfer station, the edge computing gateway receives data from various acquisition nodes and temporarily caches them. This is because in the process of collecting data, there may be fluctuations in data transmission or rate mismatches between the acquisition device and the subsequent analysis system. Temporary caching can ensure that data is not lost. At the same time, the edge computing gateway will perform integrity checks on the received data. Integrity checks are performed by calculating data through specific algorithms, such as checksum algorithms or hash algorithms, and comparing them with the check values ​​pre-calculated by the sender. If there is no error or tampering in the data during transmission, the check values ​​should match.

[0110] Step S115, in response to the integrity check passing, batch reading the real-time operation data from the edge computing gateway.

[0111] Only when the integrity check is successful can the acquired data be ensured to be accurate and reliable. Batch reading of data from the edge computing gateway can improve the efficiency of data acquisition, obtain data at multiple time points at one time, and facilitate subsequent comprehensive analysis of the battery status. These batch read real-time operating data, including voltage series, current series, temperature series, and timestamp series, will become the basis for further analysis of the battery status, such as multi-dimensional feature extraction, battery status monitoring model input, etc., so as to achieve accurate monitoring and management of the electric vehicle power battery status.

[0112] In a possible implementation, step S121 includes: Step S1211, performing extreme point detection on the voltage sequence, identifying all local maximum points and local minimum points, and calculating the absolute difference between adjacent maximum and minimum values, and taking the maximum absolute difference as the voltage fluctuation extreme value.

[0113] During a complete driving and charging cycle of an electric vehicle, the battery voltage changes continuously. For example, when the vehicle starts, the battery needs to output a large current instantly, and the voltage will drop significantly. Then, during the stable driving of the vehicle, the voltage will fluctuate within a certain range. When the vehicle performs brake energy recovery (charging process), the voltage will rise again. In this process, all local maximum points and local minimum points can be identified through the extreme point detection algorithm. For example, during a certain driving process, the local maximum point detected is 4.1 volts, and the following local minimum point is 3.9 volts. The absolute difference between the adjacent maximum and minimum values ​​is calculated to be 0.2 volts. In the entire voltage sequence, such calculations are continuously performed, and finally the maximum absolute difference is taken as the voltage fluctuation extreme value. Assume that during the entire driving and charging cycle, the absolute difference between the largest adjacent maximum and minimum values ​​is 0.3 volts. This 0.3 volt is the voltage fluctuation extreme value of the battery in this cycle. This extreme value can reflect the maximum amplitude of the battery voltage fluctuation under different working conditions and is an important indicator of battery stability. If the voltage fluctuation is too extreme, it may mean that the chemical reaction inside the battery is unstable or there is a problem with the battery management system.

[0114] Step S1212, performing a first-order difference operation on the voltage sequence to obtain a differential voltage sequence, and calculating an arithmetic mean value of the differential voltage sequence in a sliding time window as the voltage change rate mean value.

[0115] The rate of change of battery voltage can reflect the charge and discharge state and health status of the battery. When the vehicle is in the acceleration stage, the battery discharges quickly and the voltage drops quickly; when driving at a constant speed, the voltage drops relatively slowly. The differential voltage sequence is obtained by performing a first-order difference operation on the voltage sequence. For example, in a certain 10-second time interval, the values ​​of the voltage sequence are 4.0 volts, 3.95 volts, 3.9 volts, etc., and the differential voltage sequence obtained after the first-order difference operation may be -0.05 volts, -0.05 volts, etc. Then, the arithmetic mean of this differential voltage sequence in the sliding time window is calculated as the mean voltage change rate. Assuming that the sliding time window is set to 5 seconds, the values ​​of the differential voltage sequence in this window are -0.03 volts, -0.04 volts, -0.03 volts, etc., and their arithmetic mean is -0.033 volts. This -0.033 volt is the mean voltage change rate in the 5-second sliding time window. By calculating the average voltage change rate in different sliding time windows, we can fully understand the voltage change trend of the battery in different time periods, thereby evaluating the performance and status of the battery.

[0116] Step S1213, performing platform interval detection on the voltage sequence, identifying the maximum continuous time period during which the voltage change rate is continuously lower than the platform threshold, and taking the length of the maximum continuous time period as the voltage platform duration.

[0117] During the battery charging process, especially when the constant current charging stage transitions to the constant voltage charging stage, a voltage platform phenomenon will occur. At this time, the voltage change rate continues to be lower than the platform threshold. For example, the platform threshold is set to 0.01 volts / second. During the charging process, the detection algorithm is used to identify the interval in which the voltage change rate continues to be lower than this threshold. Assume that at a certain stage of charging, from the 100th second to the 150th second, the voltage change rate is always lower than 0.01 volts / second. This maximum continuous time period of 50 seconds is the voltage platform duration. The duration of the voltage platform is of great significance for evaluating the charging characteristics and health status of the battery. If the voltage platform duration is abnormal, it may indicate that the state of the chemical substances inside the battery has changed, or there is a problem with the matching between the charging equipment and the battery.

[0118] In a possible implementation, step S129 includes: Step S1291, performing a first-order seasonal difference operation on each feature dimension in the dynamic feature subset to eliminate the trend component and retain the periodic fluctuation component.

[0119] During the operation of electric vehicles, many characteristics related to the battery status have certain trends and periodicities. Taking the dynamic characteristic of battery charge and discharge efficiency as an example, as the battery is used more and ages, its charge and discharge efficiency will generally show a downward trend, but in different seasons or driving conditions, there will be periodic fluctuations. Through the first-order seasonal difference operation, for example, for a time series of charge and discharge efficiency data, the difference between adjacent time points is calculated, and this long-term downward trend can be eliminated, leaving only the periodic fluctuation component. For other dynamic feature dimensions, such as the temperature change rate of the battery under different working conditions, the same first-order seasonal difference operation is also used, so that each feature dimension removes the trend component, leaving only information related to the periodic changes in the battery status.

[0120] Step S1292, performing fast Fourier transform on the differentiated feature dimension, extracting the dominant frequency component, and constructing a reference phase signal based on the dominant frequency component.

[0121] The various state characteristics of the battery have different manifestations in the frequency domain. For example, certain periodic behaviors of the battery may be more obvious at specific frequencies. Taking the dynamic feature of battery temperature change as an example, the temperature change rate data after differentiation can be converted from the time domain to the frequency domain by performing a fast Fourier transform. In the frequency domain, the dominant frequency component with the highest energy or the greatest impact on the feature can be found. Suppose that in the frequency domain analysis of this temperature change rate, it is found that the frequency component of 0.1 Hz has the highest energy, and this 0.1 Hz is the dominant frequency component. A reference phase signal is constructed based on this dominant frequency component, and this reference phase signal can represent a reference phase state of the feature in the frequency domain. For other dynamic feature dimensions, the same fast Fourier transform is performed to find the respective dominant frequency components and construct the corresponding reference phase signal.

[0122] Step S1293, performing cyclic phase difference calculation on the phase component of each characteristic dimension and the reference phase signal to generate a phase offset set.

[0123] There may be phase differences between different dynamic feature dimensions. Taking the two dynamic features of battery charge and discharge current and battery temperature as examples, their phases may be inconsistent due to various factors such as chemical reactions and heat conduction inside the battery. By calculating the cyclic phase difference between the phase component of each feature dimension and the reference phase signal, a phase offset set can be obtained. For example, for the charge and discharge current feature dimension, the calculated phase offset from the reference phase signal is 0.2 radians; for the battery temperature feature dimension, the calculated phase offset is -0.1 radians, etc. This phase offset set can quantify the degree of phase deviation of each feature dimension relative to the reference phase signal.

[0124] Step S1294: performing time domain translation compensation on each characteristic dimension according to the phase offset set, so that the phase difference of the dominant frequency components of all characteristic dimensions is less than a preset phase tolerance.

[0125] In order to more accurately analyze the relationship between different feature dimensions, the phase needs to be aligned. For example, the preset phase tolerance is 0.1 radians. For the charge and discharge current feature dimension, since its phase offset is 0.2 radians, it needs to be shifted and compensated in the time domain so that the phase difference between its dominant frequency component and the reference phase signal is less than 0.1 radians. Similarly, such time domain shift compensation operations are also performed on other dynamic feature dimensions such as battery temperature. In this way, it is ensured that the dominant frequency components of all dynamic feature dimensions are as close as possible in phase to facilitate subsequent analysis and processing.

[0126] Step S1295, performing amplitude normalization and timestamp resampling on the feature dimensions after phase compensation to generate the dynamic alignment feature subset.

[0127] Finally, after completing the phase compensation, the amplitudes of different feature dimensions may vary greatly, which will affect the subsequent comprehensive analysis of the battery status. For example, the battery charge and discharge current amplitude may be in tens of amperes, while the battery temperature amplitude may be in tens of degrees Celsius. Through amplitude normalization, these different amplitude ranges are adjusted to a unified standard range. At the same time, since the time series of the data may have been adjusted in the previous processing, timestamp resampling is required to ensure the consistency of each feature dimension in time. The feature dimensions after amplitude normalization and timestamp resampling constitute a dynamically aligned feature subset. This dynamically aligned feature subset has better consistency in both time and frequency, and can more accurately reflect the characteristic information of the battery in different states, providing a more reliable data basis for subsequent battery status monitoring and analysis.

[0128] In a possible implementation, step S133 includes: Step S1331, dividing the weighted spatiotemporal feature map into a forward input sequence and a reverse input sequence according to the time step.

[0129] In this embodiment, in the actual driving and charging and discharging process of the electric vehicle, the time step is an important time interval unit for analyzing the battery state. For example, with every 5 minutes as a time step, the weighted spatiotemporal feature map contains the characteristic information of the battery at different times and spaces (the space here can be understood as different physical positions of the battery or different state parameter spaces). After segmentation according to this time step, the forward input sequence can be regarded as starting from the initial state of the battery (such as the state of a new car or the initial state after a complete charge), and as time moves forward, it contains the characteristic information of the battery at each time step. This sequence reflects the state change trend of the battery during normal use, such as the change of battery voltage, current, temperature and other characteristics with the increase of mileage and the increase of charging and discharging times. The reverse input sequence starts from a later time point (such as the current time or the time point after the last deep discharge) and goes back to an early state of the battery. This sequence contains state change information in the opposite direction of the forward sequence, such as the charging recovery process and voltage recovery of the battery at different time steps when the battery goes back from the current state to the previous state.

[0130] Step S1332, performing hidden state recursion on the forward input sequence through a forward recurrent neural network unit, calculating the degradation increment of each time step, and accumulating to obtain the degradation accumulation.

[0131] In every driving and charging and discharging cycle of an electric vehicle, the battery will degrade to a certain extent. When the forward recurrent neural network unit processes the forward input sequence, taking the battery capacity degradation as an example, in each time step, according to the battery voltage, current, temperature and other relevant feature information in the weighted spatiotemporal feature map, the capacity degradation increment of the battery at that time step is calculated through the neuron calculation mechanism inside the forward recurrent neural network unit. For example, in a 5-minute time step, according to the average discharge current, voltage drop and battery temperature of the battery in this time period, the degradation increment of the battery capacity is 0.01 (here it is assumed to be measured in some capacity unit, such as ampere-hour) through the calculation method of the weights and activation functions in the pre-trained forward recurrent neural network unit. As time goes by, such calculations are performed for each time step, and then these degradation increments are accumulated in sequence. Assume that after 10 time steps, the degradation increments calculated for each time step are 0.01, 0.015, 0.02, etc., and the sum of these values ​​is 0.18, which is the cumulative degradation amount within these 10 time steps. This cumulative degradation reflects the overall degradation degree of the battery during this period, and it comprehensively considers the performance degradation of the battery caused by various factors at different time steps.

[0132] Step S1333, performing hidden state backtracking on the reverse input sequence through a reverse recurrent neural network unit, calculating the life residual component of each time step, and weighted aggregation to obtain the life residual coefficient.

[0133] For example, the remaining life of a battery is affected by many factors, and the influence of these factors on the remaining life can be better captured through reverse analysis. When the reverse recurrent neural network unit processes the reverse input sequence, it is also based on various feature information in the weighted spatiotemporal feature map. For example, in the process of the battery backtracking from the current state to the previous state, in each time step, the life residual component related to the remaining life is calculated based on factors such as the battery's charging efficiency, voltage recovery speed, and temperature change. Assume that in a certain time step, the life residual component calculated based on these factors is 0.05 (here it is assumed to be a dimensionless value after normalization, representing a relative amount related to the remaining life). After performing such calculations for each time step, different weights need to be assigned according to the importance of the impact of different time steps on the remaining life, and then these weighted life residual components are weighted and aggregated to obtain the life residual coefficient. For example, for the early time step, since its impact on the remaining life is relatively small, the weight assigned is 0.1; while for the more recent time step, the weight assigned is 0.3. The life residual coefficient obtained through such weighted aggregation method can comprehensively reflect the characteristic information related to the remaining life of the battery during the reverse analysis process.

[0134] Step S1334, cross-layer connect the final hidden state of the forward recurrent neural network unit and the initial hidden state of the reverse recurrent neural network unit to generate a joint hidden state vector.

[0135] The final hidden state obtained by the forward recurrent neural network unit after processing the forward input sequence contains the final state information of the battery under the forward time series, such as the state characteristics of the battery after a series of use and degradation; while the initial hidden state of the reverse recurrent neural network unit when processing the reverse input sequence contains the initial state information of the battery when it starts to trace back from the current state. Although these two state information come from different analysis directions, they are both important for a comprehensive understanding of the battery state. By connecting these two hidden states across layers, they are combined into a joint hidden state vector. For example, the final hidden state of the forward recurrent neural network unit is a vector containing multiple elements (such as numerical values ​​representing state-related information such as battery capacity, internal resistance, and voltage), assuming it is [0.1, 0.2, 0.3]; the initial hidden state of the reverse recurrent neural network unit is [0.4, 0.5, 0.6], and the joint hidden state vector obtained by connecting them is [0.1, 0.2, 0.3, 0.4, 0.5, 0.6]. This joint hidden state vector integrates the key state information of the battery in forward and reverse analysis, providing a rich source of information for subsequent more accurate correction of degradation accumulation and life residual coefficient.

[0136] Step S1335, performing a nonlinear transformation on the joint hidden state vector through a fully connected layer, and outputting a correction factor of the degradation accumulation amount and the life residual coefficient.

[0137] The fully connected layer has a powerful nonlinear mapping capability. After the joint hidden state vector enters the fully connected layer, each neuron in the fully connected layer is connected to each element in the joint hidden state vector, and nonlinear transformation is performed through calculation mechanisms such as weights and activation functions inside the neurons. For example, the weights of neurons in the fully connected layer are obtained through pre-training learning. According to these weights, the elements in the joint hidden state vector are weighted and summed, and then nonlinear transformation is performed through activation functions (such as ReLU functions or other suitable nonlinear activation functions). After such a calculation process, the output result is the correction factor of the degradation accumulation and the life residual coefficient. This correction factor can adjust the previously calculated degradation accumulation and life residual coefficient according to the comprehensive information in the joint hidden state vector. For example, if the correction factor is 1.2, the previously calculated degradation accumulation and life residual coefficient will be multiplied by this correction factor respectively, so as to obtain more accurate degradation accumulation and life residual coefficient, so as to better reflect the actual degradation and remaining life of the battery. The operation mode of this bidirectional cyclic prediction layer comprehensively considers the state information of the battery in the forward and reverse time series, as well as the hidden state information in different analysis directions. By outputting the correction factor through nonlinear transformation, it can ultimately generate the degradation accumulation and life residual coefficient more accurately, laying a solid foundation for the subsequent accurate evaluation of the battery's state parameters (such as degradation state parameters, health state parameters and remaining life prediction parameters).

[0138] Figure 2 The hardware structure of the battery state monitoring and analysis system 100 based on machine learning for implementing the above-mentioned battery state monitoring and analysis method based on machine learning provided in an embodiment of the present invention is shown as follows: Figure 2 As shown, the battery status monitoring and analysis system 100 based on machine learning may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0139] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the machine-learning-based battery state monitoring and analysis system 100 uses to execute or use to complete the exemplary method described in the present invention.

[0140] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the battery status monitoring and analysis method based on machine learning in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0141] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned battery status monitoring and analysis system 100 based on machine learning. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0142] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the battery status monitoring and analysis method based on machine learning as described above is implemented.

[0143] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A battery status monitoring and analysis method based on machine learning, characterized in that: The method comprises: Acquire real-time operation data of the target battery in the charge and discharge cycle, wherein the real-time operation data includes a voltage sequence, a current sequence, a temperature sequence, and a timestamp sequence; Performing multi-dimensional feature extraction on the real-time operation data to generate an original feature set, and performing dynamic time series processing on the original feature set to obtain a standardized feature matrix; Inputting the standardized feature matrix into a pre-trained battery state monitoring model to generate degradation state parameters, health state parameters and remaining life prediction parameters of the target battery; Generate an abnormality detection result and maintenance decision suggestion for the target battery according to the degradation state parameter, health state parameter and remaining life prediction parameter; Based on the abnormality detection results and maintenance decision suggestions, the parameter weights of the battery status monitoring model are adaptively adjusted to update the prediction accuracy of the battery status monitoring model.

2. The battery status monitoring and analysis method based on machine learning according to claim 1 is characterized in that: The performing multi-dimensional feature extraction on the real-time operation data to generate an original feature set includes: Extracting voltage fluctuation extreme value, voltage change rate mean and voltage platform duration from the voltage sequence; Extracting current charge and discharge efficiency, current peak interval period and current decay slope from the current sequence; Extracting temperature gradient distribution, temperature rise rate threshold and temperature balance index from the temperature sequence; Extracting the charge and discharge cycle length, the rest phase duration and the cycle interval consistency coefficient from the timestamp sequence; The voltage fluctuation extreme value, voltage change rate mean, voltage platform duration, current charge and discharge efficiency, current peak interval period, current decay slope, temperature gradient distribution, temperature rise rate threshold, temperature balance index, charge and discharge cycle length, static stage length and cycle interval consistency coefficient are normalized and spliced ​​to generate the original feature set.

3. The battery status monitoring and analysis method based on machine learning according to claim 1 is characterized in that: The step of performing dynamic time series processing on the original feature set to obtain a standardized feature matrix includes: Perform missing value filling and noise filtering on each feature dimension in the original feature set to generate a denoised feature set; According to the statistical distribution characteristics of each feature dimension in the denoising feature set, the denoising feature set is divided into a static feature subset and a dynamic feature subset; Performing sliding window mean smoothing processing on the static feature subset to generate a static smooth feature subset; Performing time series difference transformation and phase alignment processing on the dynamic feature subset to generate a dynamic alignment feature subset; The static smooth feature subset and the dynamic alignment feature subset are cross-dimensionally fused according to a timestamp sequence to generate the standardized feature matrix.

4. The battery status monitoring and analysis method based on machine learning according to claim 3 is characterized in that: The step of inputting the standardized feature matrix into a pre-trained battery state monitoring model to generate degradation state parameters, health state parameters and remaining life prediction parameters of the target battery includes: The standardized feature matrix is ​​captured locally through a convolutional temporal coding layer in the battery state monitoring model to generate a primary spatiotemporal feature map; Through the attention weight allocation layer in the battery status monitoring model, the primary spatiotemporal feature graph is strengthened at key time points to generate a weighted spatiotemporal feature graph; Through the bidirectional cyclic prediction layer in the battery state monitoring model, forward cumulative degradation analysis and reverse life residual fitting are performed on the weighted spatiotemporal characteristic graph to generate degradation accumulation and life residual coefficient; The degradation accumulation and life residual coefficient are jointly mapped through the multi-task output layer in the battery state monitoring model, and the degradation state parameters, health state parameters and remaining life prediction parameters are synchronously output.

5. The battery status monitoring and analysis method based on machine learning according to claim 1 is characterized in that: The generating, according to the degradation state parameter, health state parameter and remaining life prediction parameter, an abnormality detection result and a maintenance decision suggestion for the target battery includes: Comparing the degradation state parameter with a preset degradation threshold interval, and generating a first abnormality mark if the degradation state parameter exceeds the degradation threshold interval; Calculate the trend deviation between the health status parameter and the historical health status baseline, and if the trend deviation exceeds a dynamic deviation threshold, generate a second abnormal mark; Performing relative position analysis on the remaining life prediction parameter and the median of life distribution of batteries in the same batch, and generating a third abnormal mark if the remaining life prediction parameter is lower than the life warning quantile; According to the combination pattern of the first abnormal mark, the second abnormal mark and the third abnormal mark, a preset maintenance strategy knowledge base is matched to generate the maintenance decision suggestion including the maintenance priority, the maintenance operation type and the maintenance time window; The method of adaptively adjusting the parameter weights of the battery status monitoring model based on the abnormality detection results and the maintenance decision suggestions to update the prediction accuracy of the battery status monitoring model includes: According to the abnormality mark type in the abnormality detection result, select a corresponding weight update rule from the model parameter adjustment strategy library; Determining a learning rate adjustment range and a regularization strength of the battery status monitoring model according to the maintenance priority in the maintenance decision suggestion; According to the maintenance operation type and the maintenance time window, a virtual feedback data set is constructed, and the virtual feedback data set is mixed and sampled with historical training data; Based on the mixed sampled data set, the updated learning rate and the regularization strength, back-propagation optimization is performed on the parameter weights of the convolutional temporal coding layer, the attention weight allocation layer and the bidirectional recurrent prediction layer of the battery state monitoring model; The parameter weights of the multi-task output layer in the battery state monitoring model are frozen, and only the parameter weights of the convolutional temporal coding layer, the attention weight allocation layer, and the bidirectional recurrent prediction layer are iteratively updated until the prediction error converges.

6. The battery status monitoring and analysis method based on machine learning according to claim 1 is characterized in that: Before acquiring the real-time operation data of the target battery in the charge and discharge cycle, the method further includes: Configuring a distributed sensor node network connected to the target battery, the distributed sensor node network comprising a voltage acquisition node, a current acquisition node, a temperature acquisition node and a time synchronization node; Capturing the voltage sequence at a first sampling frequency through the voltage acquisition node, and capturing the current sequence at a second sampling frequency through the current acquisition node; The temperature acquisition node captures the temperature sequence at a third sampling frequency, and the time synchronization node performs microsecond-level alignment on the sampling moments of the voltage acquisition node, the current acquisition node, and the temperature acquisition node; Sending the voltage sequence, current sequence, temperature sequence and timestamp sequence to the edge computing gateway through an encrypted transmission protocol for temporary caching and integrity verification; In response to the integrity check passing, the real-time operation data is read in batches from the edge computing gateway.

7. The battery status monitoring and analysis method based on machine learning according to claim 2 is characterized in that: The step of extracting the voltage fluctuation extreme value, the voltage change rate mean value and the voltage platform duration from the voltage sequence includes: Perform extreme point detection on the voltage sequence, identify all local maximum points and local minimum points, and calculate the absolute difference between adjacent maximum and minimum values, and use the maximum absolute difference as the voltage fluctuation extreme value; Performing a first-order difference operation on the voltage sequence to obtain a differential voltage sequence, and calculating an arithmetic mean value of the differential voltage sequence within a sliding time window as the voltage change rate mean value; A platform interval detection is performed on the voltage sequence to identify a maximum continuous time period during which the voltage change rate is continuously lower than a platform threshold, and the length of the maximum continuous time period is used as the voltage platform duration.

8. The battery status monitoring and analysis method based on machine learning according to claim 3 is characterized in that: The step of performing time series difference transformation and phase alignment processing on the dynamic feature subset to generate a dynamic alignment feature subset includes: Performing a first-order seasonal difference operation on each feature dimension in the dynamic feature subset to eliminate trend components and retain periodic fluctuation components; Performing a fast Fourier transform on the feature dimension after differentiation, extracting the dominant frequency component, and constructing a reference phase signal based on the dominant frequency component; Perform cyclic phase difference calculation on the phase component of each characteristic dimension and the reference phase signal to generate a phase offset set; Performing time domain translation compensation on each characteristic dimension according to the phase offset set, so that the phase difference of the dominant frequency components of all characteristic dimensions is less than a preset phase tolerance; The feature dimension after phase compensation is amplitude normalized and timestamp resampled to generate the dynamic alignment feature subset.

9. The battery status monitoring and analysis method based on machine learning according to claim 4 is characterized in that: The forward cumulative degradation analysis and reverse life residual fitting are performed on the weighted spatiotemporal feature graph through the bidirectional cyclic prediction layer in the battery state monitoring model to generate the degradation accumulation and life residual coefficient, including: Splitting the weighted spatiotemporal feature map into a forward input sequence and a reverse input sequence according to the time step; Performing hidden state recursion on the forward input sequence through a forward recurrent neural network unit, calculating the degradation increment of each time step, and accumulating to obtain the degradation accumulation; Perform hidden state backtracking on the reverse input sequence through a reverse recurrent neural network unit, calculate the life residual component of each time step, and perform weighted aggregation to obtain the life residual coefficient; Cross-layer connection is performed between the final hidden state of the forward recurrent neural network unit and the initial hidden state of the reverse recurrent neural network unit to generate a joint hidden state vector; The joint hidden state vector is nonlinearly transformed through a fully connected layer to output a correction factor of the degradation accumulation amount and the life residual coefficient.

10. A battery status monitoring and analysis system based on machine learning, characterized in that: The battery status monitoring and analysis system based on machine learning includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the battery status monitoring and analysis method based on machine learning as described in any one of claims 1 to 9 above.

Images