Method and system for predicting health of battery pack

Abstract

The invention relates to the technical field of electrical performance detection, in particular to a battery pack health prediction method and system. The method comprises the following steps: carrying out multi-mode sensing data acquisition on a battery pack to obtain voltage data, current data, temperature data and acoustic vibration data, and taking the data as an original monitoring data set; performing preprocessing and timely frequency domain feature extraction on the original monitoring data set, and constructing a battery pack operation feature matrix; acquiring historical operation data of the battery pack; and constructing a battery pack health situation intelligent diagnosis model by using the battery pack historical operation data and the battery pack operation characteristic matrix. According to the invention, through multi-modal data acquisition, intelligent diagnosis model and time sequence decomposition, the health state of the battery pack is accurately monitored, the aging trend and short-term fluctuation are distinguished, and the intelligent level and reliability of battery pack management are improved by combining health early warning and pre-maintenance scheme design.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical performance detection, and in particular to a method and system for battery pack health prediction. Background Art

[0002] The rapid development of the new energy industry has led to the widespread use of lithium-ion batteries in electric vehicles, energy storage systems and consumer electronics. However, during long-term use, battery packs will be affected by multiple factors such as cycle aging, environmental impact and usage patterns, which will cause their performance to gradually decay and cause safety accidents in serious cases. Early battery health prediction methods mainly rely on physical models, such as equivalent circuit models (ECM) and mechanism-based models (PBM), which describe the health status of batteries by establishing internal kinetic equations of batteries. However, such methods are more complex in describing the internal chemical reactions of batteries and require a large amount of experimental data for parameter calibration, making it difficult to adapt to battery packs of different types and working environments. Subsequently, data-driven methods have gradually become the mainstream of research, such as health prediction methods based on statistical analysis, machine learning and deep learning. Statistical analysis methods analyze the trend of battery health status changes through historical data, but cannot effectively capture complex nonlinear characteristics; although traditional machine learning methods improve prediction accuracy, they rely on manual feature extraction and are difficult to fully mine deep information in multimodal data. Although the existing technology has made certain progress, it still has shortcomings. Traditional methods of modeling battery health status rely on a single data source and cannot fully utilize the information of multimodal sensor data. Although machine learning and deep learning methods have improved prediction accuracy, most methods lack in-depth analysis of the battery health status change process and cannot effectively distinguish short-term fluctuations from long-term aging trends. Existing prediction methods still face great challenges in abnormal state detection and health warnings, making it difficult to provide accurate preventive maintenance plans. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a method and system for battery pack health prediction to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for battery pack health prediction includes the following steps: Step S1: performing multi-modal sensing data collection on the battery pack to obtain voltage data, current data, temperature data and acoustic vibration data as an original monitoring data set; Step S2: preprocessing the original monitoring data set and extracting the frequency domain features, and constructing the battery pack operation feature matrix; Step S3: obtaining historical operation data of the battery pack; constructing an intelligent diagnosis model for the health status of the battery pack using the historical operation data of the battery pack and the battery pack operation feature matrix, wherein constructing the intelligent diagnosis model for the health status of the battery pack includes training a generative adversarial network and training a multi-input deep neural network; Step S4: evaluating the battery pack status based on the battery pack health status intelligent diagnosis model to obtain battery pack health status evaluation data; Step S5: performing time series decomposition on the battery pack health status assessment data, separating the aging trend and short-term fluctuation, and obtaining the battery pack life evolution characteristics; performing abnormal level assessment on the battery pack health status assessment data, and obtaining the battery pack abnormal state level; Step S6: Perform health warning and preventive maintenance plan design based on the battery pack life evolution characteristics and the battery pack abnormal status level to obtain the battery pack health prediction result.

[0005] The present invention improves the comprehensiveness of battery pack health status monitoring through multimodal sensor data collection, and can obtain multi-dimensional data of voltage, current, temperature and acoustic vibration, overcoming the limitation of traditional methods relying on a single data source. By preprocessing the original monitoring data set and extracting the frequency domain features in time, the noise in the data is removed, the validity of the feature data is improved, and the battery pack operation feature matrix is ​​constructed, so that the expression of the data is more in line with the needs of health status analysis. The historical operation data of the battery pack is combined with the battery pack operation feature matrix to construct a health status intelligent diagnosis model, which combines the generative adversarial network with the multi-input deep neural network to achieve high-precision health status identification. Compared with traditional statistical analysis and machine learning methods, it can more effectively capture the nonlinear characteristics of the battery pack and improve the accuracy of health status prediction. The state evaluation based on the intelligent diagnosis model can provide detailed health status evaluation data, so that the battery pack operation status can be accurately quantified, laying a data foundation for subsequent life prediction and abnormal analysis. Through the time series decomposition method, the aging trend and short-term fluctuations in the health status data are effectively distinguished, making the battery pack life evolution characteristics clearer, and combined with the abnormal level assessment, the battery pack abnormal state can be finely identified, overcoming the problem that the existing methods are difficult to distinguish between short-term fluctuations and long-term aging trends. Further combined with the life evolution characteristics and abnormal state levels, health warning and pre-maintenance schemes are designed, so that health prediction is not limited to the current state assessment, but can also predict the failure trend of the battery pack in advance, provide accurate health warnings, and formulate targeted pre-maintenance strategies, solving the shortcomings of existing methods in abnormal detection and pre-maintenance, thereby improving the intelligence level of battery pack management, extending the battery pack service life, and improving the safety and reliability of the system.

[0006] Preferably, the present invention further provides a system for battery health prediction, which is used to execute the above-mentioned method for battery health prediction, and the system for battery health prediction comprises: A multi-modal data acquisition module is used to collect multi-modal sensor data of the battery pack to obtain voltage data, current data, temperature data and acoustic vibration data as the original monitoring data set; Data preprocessing and feature extraction module, used to preprocess the original monitoring data set and extract the frequency domain features, and build the battery pack operation feature matrix; An intelligent diagnosis model building module is used to obtain the historical operation data of the battery pack; the battery pack historical operation data and the battery pack operation feature matrix are used to build an intelligent diagnosis model for the health status of the battery pack, wherein the construction of the intelligent diagnosis model for the health status of the battery pack includes training a generative adversarial network and training a multi-input deep neural network; A health status assessment module is used to assess the battery pack status based on the battery pack health status intelligent diagnosis model to obtain battery pack health status assessment data; The life evolution and anomaly detection module is used to perform time series decomposition on the battery pack health status assessment data, separate the aging trend and short-term fluctuations, and obtain the battery pack life evolution characteristics; perform anomaly level assessment on the battery pack health status assessment data to obtain the battery pack abnormal status level; The health prediction and preventive maintenance module is used to perform health warning and preventive maintenance plan design based on the battery pack life evolution characteristics and battery pack abnormal status level, and obtain the battery pack health prediction results.

[0007] The present invention can comprehensively obtain various operating data of the battery pack and provide rich raw data support for subsequent health assessment. By cleaning, feature extraction and constructing an operating feature matrix, the quality and analyzability of the data are ensured, providing an effective information basis for model training. Using historical data and operating characteristics, a powerful intelligent diagnosis model for health status is constructed, which can accurately identify the health status of the battery pack, and improves the diagnostic accuracy and robustness by training generative adversarial networks and multi-input deep neural networks. The state of the battery pack is accurately evaluated and detailed health status data is generated, providing a key reference for subsequent life prediction and anomaly detection. Through time series decomposition and anomaly level evaluation, the aging trend and short-term fluctuations of the battery pack are revealed, which helps to accurately identify potential risks and improve the ability to predict faults. Based on the health assessment results, health warning and pre-maintenance plans are designed to provide a scientific basis for the maintenance strategy of the battery pack and ensure the reliability and service life of the battery pack. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1 A schematic diagram of a process flow of a method for battery health prediction according to the present invention; Figure 2 for Figure 1 Detailed step flow diagram of step S1; Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0009] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0010] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0011] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0012] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for battery pack health prediction, the method comprising the following steps: In the embodiment of the present invention, reference Figure 1 FIG. 1 is a flow chart of steps of a method for battery health prediction according to the present invention. In this example, the method for battery health prediction includes the following steps: Step S1: performing multi-modal sensing data collection on the battery pack to obtain voltage data, current data, temperature data and acoustic vibration data as an original monitoring data set; In an embodiment of the present invention, multimodal sensor data acquisition of a battery pack is achieved by arranging a multi-type sensor array on the surface and periphery of the battery pack. The sensor array consists of 16 voltage acquisition points, 8 current acquisition points, 24 temperature sensor nodes and 12 acoustic vibration sensors. A high-precision digital multimeter is used for voltage data acquisition, with a sampling accuracy of 0.01V and a sampling frequency of 10Hz. A Hall current sensor is used for current data acquisition, with a range of 0-500A, an accuracy of 0.5% of the full scale, and a sampling frequency of 20Hz. A PT100 platinum resistance temperature sensor array is used for temperature data acquisition, with a measurement range of -20°C to 120°C, a resolution of 0.1°C, and a sampling frequency of 1Hz. The sensor arrangement follows the hotspot coverage principle, and is arranged in a 3×8 grid on the surface of the battery pack. Arrangement; Acoustic vibration data acquisition uses a piezoelectric accelerometer with a frequency response range of 20Hz-20kHz, a sensitivity of 100mV / g, and a sampling frequency of 40kHz. The sensors are fixed at the four corners and the center of the battery pack; All sensor signals are digitized by a 16-bit ADC converter and then transmitted to the data acquisition unit; Multimodal data acquisition sets the sampling duration to the entire standard charge and discharge cycle, and records the cumulative number of cycles of the battery pack, ambient temperature, and load type at the beginning of each sampling; The acquisition unit aligns the four types of sensor data according to a unified timestamp and merges them into a structured data set to form an original monitoring data set containing voltage vectors, current vectors, temperature matrices, and acoustic spectrum matrices. The data sets are stored and numbered separately for each charge and discharge cycle.

[0013] Step S2: preprocessing the original monitoring data set and extracting the frequency domain features, and constructing the battery pack operation feature matrix; The embodiment of the present invention performs data cleaning on the voltage data, current data, temperature data and acoustic vibration data obtained in step S1, including removing abnormal values, removing pulse interference in the voltage data and current data by a median filtering algorithm, eliminating random fluctuations in the temperature data by a sliding average method, and processing high-frequency noise in the acoustic vibration data by a wavelet threshold denoising method; then, time domain feature extraction is performed on the cleaned data, and the mean, variance, peak factor, kurtosis, and skewness of the voltage data are calculated; the mean, standard deviation, maximum value, minimum value, and volatility of the current data are calculated; the highest temperature, lowest temperature, temperature gradient, and temperature change rate of the temperature data are calculated; and the root mean square of the acoustic vibration data is calculated. value, peak value, waveform factor, and pulse factor; then, frequency domain features are extracted from the cleaned data, and fast Fourier transform is applied to the voltage data and current data to extract the spectral energy distribution, main frequency component, and spectral centroid. Wavelet packet transform is applied to the acoustic vibration data to extract the energy proportion of different frequency bands and the frequency band power spectral density; finally, all the extracted time domain features and frequency domain features are organized into feature vectors according to the acquisition time sequence, and the feature vectors obtained by multiple samplings are combined to construct a battery pack operation feature matrix. Each row of the matrix represents all the features at a time point, and each column represents the value of a feature at different time points. The matrix dimension is n×m, where n is the number of sampling times and m is the total number of features.

[0014] Step S3: obtaining historical operation data of the battery pack; constructing an intelligent diagnosis model for the health status of the battery pack using the historical operation data of the battery pack and the battery pack operation feature matrix, wherein constructing the intelligent diagnosis model for the health status of the battery pack includes training a generative adversarial network and training a multi-input deep neural network; The embodiment of the present invention extracts historical operation data of the battery pack under different charge and discharge cycles and environmental conditions from the battery management system database, including historical voltage fluctuation records, historical charge and discharge current curves, historical temperature distribution data, historical acoustic vibration characteristics, and corresponding battery pack capacity attenuation rate, internal resistance growth rate, and cycle life health status label; then, the historical operation data is time-aligned and normalized with the battery pack operation feature matrix constructed in step S2, and the data is divided into a training set accounting for 70% and a verification set accounting for 30%; then, a generative adversarial network is constructed, in which the generator adopts a four-layer convolutional neural network structure, the input is a 100-dimensional random noise vector, the number of hidden layer neurons is 256, 512, 256, and 128 respectively, the activation function is LeakyReLU, and the output layer uses the Tanh activation function; the discriminator adopts a five-layer fully connected network, the input is the battery pack operation feature matrix, the number of hidden layer neurons is 128, 256, 512, and 256 respectively, the activation function is ReLU, and the output layer uses Sigmoid activation function, and alternate training for 500 rounds with the Adam optimizer at a learning rate of 0.0002 until the generator can generate synthetic feature data close to the real data distribution; then, the original feature data and the generated synthetic feature data are used to jointly train the multi-input deep neural network, which contains four parallel input branches to process voltage features, current features, temperature features and acoustic vibration features respectively. Each branch consists of three fully connected layers with 64, 32 and 16 neurons respectively. The outputs of the four branches are merged through the connection layer and connected to the three fully connected layers with 128, 64 and 32 neurons respectively. The number of neurons in the final output layer is 3, corresponding to the capacity attenuation rate, internal resistance growth rate and remaining life prediction value respectively. The mean square error is used as the loss function, and the stochastic gradient descent algorithm is used to train 1000 rounds with a learning rate of 0.001; finally, when the prediction error on the validation set is less than 3%, it is confirmed that the training of the intelligent diagnosis model for the health status of the battery pack is completed, and the model parameters are saved.

[0015] Step S4: evaluating the battery pack status based on the battery pack health status intelligent diagnosis model to obtain battery pack health status evaluation data; The embodiment of the present invention loads the intelligent diagnosis model of the health status of the battery pack, including generating adversarial network model parameters and multi-input deep neural network model parameters; then, the raw data collected in the current monitoring period is preprocessed and feature extracted according to the processing method of step S2 to form the battery pack operation feature vector at the current moment; then, the battery pack operation feature vector is input into the multi-input deep neural network, and the predicted values ​​of the three key health indicators of capacity attenuation rate, internal resistance growth rate and remaining life are obtained through forward propagation calculation; then, the terminal voltage response curve actually measured by the battery pack is compared with the ideal terminal voltage response curve predicted by the multi-input deep neural network, and the root mean square error, correlation coefficient and consistency index are calculated to quantitatively evaluate the difference between the current operating state of the battery pack and the ideal state. The degree of deviation; at the same time, the discriminator of the generative adversarial network is used to calculate the KL divergence between the probability distribution of the current battery pack operation characteristics and the probability distribution of historical normal samples. When the KL divergence exceeds the preset threshold value of 0.5, it is marked as an abnormal state; further, according to the voltage consistency, temperature distribution uniformity and acoustic vibration signal correlation of each single battery in the battery pack, the weight of each indicator is determined in combination with the hierarchical analysis method, and the comprehensive health score is calculated. The health score ranges from 0 to 100. The higher the score, the better the health status of the battery pack; finally, the capacity attenuation rate, internal resistance growth rate, remaining life prediction value, terminal voltage response deviation, KL divergence and comprehensive health score indicators are combined to form the battery pack health status assessment data, which are stored in the database at the time granularity of hours, days, weeks and months.

[0016] Step S5: performing time series decomposition on the battery pack health status assessment data, separating the aging trend and short-term fluctuation, and obtaining the battery pack life evolution characteristics; performing abnormal level assessment on the battery pack health status assessment data, and obtaining the battery pack abnormal state level; The embodiment of the present invention extracts three key indicators, namely, capacity attenuation rate, internal resistance growth rate and comprehensive health score, from the battery pack health status assessment data obtained in step S4 to construct a time series data set; then, a seasonal decomposition method is used to process the time series data, and the data is decomposed into three parts: a trend term, a seasonal term and a residual term, wherein the trend term reflects the long-term aging trend of the battery pack, the seasonal term reflects the periodic changes in the charge and discharge cycle, and the residual term reflects random fluctuations; then, a polynomial fitting method is applied to the trend term to determine the optimal fitting curve, the fitting polynomial order is 3, the fitting accuracy requires a root mean square error of less than 0.02, and the battery pack aging rate and aging acceleration are calculated by the first-order derivative and the second-order derivative of the fitting curve; then, in combination with the typical aging curve in the battery pack historical aging database, the dynamic time warping algorithm is used to calculate the correlation between the current aging trend and the typical aging pattern. Similarity is used to identify the aging type of the battery pack, which is divided into three categories: uniform aging type, accelerated aging type and sudden aging type. At the same time, the abnormal level of the battery pack health status assessment data is evaluated. Based on the three indicators of KL divergence value, voltage deviation value and temperature uniformity, the fuzzy comprehensive evaluation method is used to construct the abnormal level evaluation matrix, and the abnormal state is divided into five levels, namely normal (level 1, KL divergence <0.3), slight abnormality (level 2, 0.3≤KL divergence <0.5), moderate abnormality (level 3, 0.5≤KL divergence <0.7), serious abnormality (level 4, 0.7≤KL divergence <0.9) and dangerous state (level 5, KL divergence ≥0.9). Finally, the battery pack aging trend characteristics, aging rate, aging acceleration, aging type and abnormal level information are integrated into the battery pack life evolution characteristics and battery pack abnormal state level, and stored in the database.

[0017] Step S6: Perform health warning and preventive maintenance plan design based on the battery pack life evolution characteristics and the battery pack abnormal status level to obtain the battery pack health prediction result.

[0018] The embodiment of the present invention extracts the battery pack life evolution characteristics, including the aging trend curve, aging rate, aging acceleration and aging type, and adopts the exponential smoothing prediction method to predict the capacity decay trajectory of the next 30 days, 90 days and 180 days, with the smoothing coefficient set to 0.85 and the prediction confidence level to 95%; then, in combination with the rated capacity and design life of the battery pack, the expected value and upper and lower boundaries of the remaining service life of the battery pack are calculated, and when the predicted capacity decays to 80% of the rated capacity, it is determined as the end of life point; then, according to the abnormal state level of the battery pack obtained in step S5, a four-level health warning mechanism is established, corresponding to different warning colors: Levels 1-2 correspond to green warnings, indicating normal operating status; Level 3 corresponds to yellow warnings, indicating that the monitoring frequency needs to be increased; Level 4 corresponds to orange warnings, indicating that the charge and discharge current needs to be limited; Level 5 corresponds to red warnings, indicating that it is necessary to stop using and overhaul immediately; Finally, a preventive maintenance plan is formulated for different warning levels and aging types: for yellow warning and uniformly aged battery packs, a balanced charging strategy is formulated to limit the charging cut-off voltage to 95% of the rated value; for orange warning and accelerated aging battery packs, current limiting measures are formulated to limit the maximum discharge rate to 0.5C, and a periodic balanced charging plan is formulated; for red warning and sudden aging battery packs, an emergency shutdown procedure is triggered, and a single cell detection task is generated to mark the location of potential faulty cells; finally, the remaining life prediction value of the battery pack, the prediction confidence interval, the health warning level, the preventive maintenance plan and the replacement time recommendation are integrated into the battery pack health prediction result, which is transmitted to the battery management system through the data interface, and a prediction report is generated, including a capacity decay trend chart, an internal resistance growth trend chart, a timeline of key abnormal events and a preventive maintenance recommendation list, providing data support for battery pack management decisions.

[0019] The present invention improves the comprehensiveness of battery pack health status monitoring through multimodal sensor data collection, and can obtain multi-dimensional data of voltage, current, temperature and acoustic vibration, overcoming the limitation of traditional methods relying on a single data source. By preprocessing the original monitoring data set and extracting the frequency domain features in time, the noise in the data is removed, the validity of the feature data is improved, and the battery pack operation feature matrix is ​​constructed, so that the expression of the data is more in line with the needs of health status analysis. The historical operation data of the battery pack is combined with the battery pack operation feature matrix to construct a health status intelligent diagnosis model, which combines the generative adversarial network with the multi-input deep neural network to achieve high-precision health status identification. Compared with traditional statistical analysis and machine learning methods, it can more effectively capture the nonlinear characteristics of the battery pack and improve the accuracy of health status prediction. The state evaluation based on the intelligent diagnosis model can provide detailed health status evaluation data, so that the battery pack operation status can be accurately quantified, laying a data foundation for subsequent life prediction and abnormal analysis. Through the time series decomposition method, the aging trend and short-term fluctuations in the health status data are effectively distinguished, making the battery pack life evolution characteristics clearer, and combined with the abnormal level assessment, the battery pack abnormal state can be finely identified, overcoming the problem that the existing methods are difficult to distinguish between short-term fluctuations and long-term aging trends. Further combined with the life evolution characteristics and abnormal state levels, health warning and pre-maintenance schemes are designed, so that health prediction is not limited to the current state assessment, but can also predict the failure trend of the battery pack in advance, provide accurate health warnings, and formulate targeted pre-maintenance strategies, solving the shortcomings of existing methods in abnormal detection and pre-maintenance, thereby improving the intelligence level of battery pack management, extending the battery pack service life, and improving the safety and reliability of the system.

[0020] Preferably, step S1 comprises the following steps: Step S11: performing high frequency sampling on the terminal voltage of the battery pack to obtain voltage time series data; Step S12: monitor the charge and discharge current of the battery pack in real time to obtain current time series data; Step S13: performing multi-point measurement on the surface temperature of the battery pack based on a distributed temperature sensor array to obtain temperature distribution data; Step S14: collecting acoustic vibration signals through a piezoelectric sensor attached to the surface of the battery pack to obtain vibration spectrum data; Step S15: time-synchronize and integrate the voltage time series data, current time series data, temperature distribution data and vibration spectrum data to obtain a multi-modal original monitoring data set.

[0021] In the embodiment of the present invention, a precision voltage acquisition module LTC6804-2 is connected to the positive and negative terminals of the battery pack, the sampling resolution is set to 16 bits, the sampling frequency is set to 10kHz, the voltage signal is adjusted to the range of 0-5V through the front voltage divider circuit, and the cutoff frequency is set to 4kHz using an RC low-pass filter to eliminate high-frequency interference. The sampling duration is 60 seconds per cycle, and the sampling cycle is triggered once every hour. The collected voltage data is transmitted to the data processing unit through the CAN bus; the specific implementation method for real-time monitoring of the charge and discharge current of the battery pack is: a Hall current sensor ACS758 is installed in series in the main circuit of the battery pack, and the measurement range is -500A to +500A, with an accuracy of ±0.5% of the full scale, the sampling frequency is set to 5kHz, and digital conversion is performed through a 16-bit ADC. The current waveform and mutation events are recorded in real time, and the average is calculated every 100ms to form current time series data; the specific implementation method of multi-point measurement of the battery pack surface temperature based on a distributed temperature sensor array is as follows: a total of 30 NTC thermistor sensors are installed on the battery pack surface in a 5×6 grid layout. The sensor temperature measurement range is -20℃ to +80℃, the measurement accuracy is ±0.5℃, the sensor spacing is 10cm, the sampling frequency is set to 1Hz, and a 24-bit delta-sigma ADC is used for data acquisition, and the collected discrete temperature point data is reconstructed into a complete temperature distribution heat map through a bilinear interpolation algorithm; the specific implementation method of collecting acoustic vibration signals through a piezoelectric sensor attached to the surface of the battery pack is as follows: 5 high-sensitivity piezoelectric acceleration sensors are installed at the four corners and the center of the battery pack shell, with a sensitivity of 100mV / g, a measurement range of ±50g, a frequency response range of 10Hz to 20kHz, and a sampling frequency set to 44.1kHz. Data is collected through a 24-bit ADC, and short-time Fourier transform is used to convert time domain signals into frequency domain signals. The analysis frequency range is 10Hz to 5kHz to form vibration spectrum data; the specific implementation method of time synchronization and integration of the voltage time series data, current time series data, temperature distribution data and vibration spectrum data is as follows: based on a unified microsecond timestamp to mark all collected data, FPGA is used to achieve hardware-level precise trigger synchronization, and the trigger error is less than 5μs. The data collected at different frequencies are unified to a sampling rate of 100Hz through a data compensation algorithm, the voltage and current data are downsampled by the mean, the vibration data is downsampled after low-pass filtering, and the temperature data is upsampled by linear interpolation. Finally, the four types of data are organized and integrated into a multi-modal original monitoring data set in MATLAB mat format according to the same time axis. The data set contains five main fields: timestamp, voltage matrix, current vector, temperature distribution matrix and vibration spectrum matrix.

[0022] The present invention can accurately capture subtle voltage changes during battery operation by high-frequency sampling of the terminal voltage of the battery pack, reflect internal chemical reactions and aging states, and improve the sensitivity of health status monitoring. Real-time monitoring of the charge and discharge current allows the current response characteristics of the battery pack under different working conditions to be quantified, providing data support for evaluating the load capacity and internal impedance changes of the battery pack. Multi-point measurement based on a distributed temperature sensor array can comprehensively obtain the temperature distribution on the surface of the battery pack, accurately identify local overheating, improve the effectiveness of thermal management, and avoid battery performance degradation or safety hazards caused by local overheating. Acoustic vibration signals are collected by piezoelectric sensors attached to the surface of the battery pack to achieve non-destructive testing of the internal structural state of the battery, identify battery anomalies, and improve the comprehensiveness of battery pack health assessment. Time synchronization and integration of various types of data are performed to form a multi-modal original monitoring data set, ensure the temporal consistency of different types of sensor data, improve the quality of data fusion, make subsequent health status assessment more accurate, overcome the problem of insufficient information from a single data source, and provide comprehensive and reliable input data for the health status intelligent diagnosis model.

[0023] Preferably, step S2 comprises the following steps: Step S21: Perform data cleaning and outlier detection on the original monitoring data set to obtain a preliminary processed data set with noise points and outliers removed; Step S22: performing noise suppression on the preliminary processed data set to obtain a denoised data set; Step S23: standardizing and normalizing the denoised data set to obtain a standardized data set; Step S24: performing continuous wavelet transform on the standardized data set to obtain a time-frequency domain decomposition result; Step S25: extracting energy distribution, scale coefficient and frequency characteristics from the time-frequency domain decomposition result to obtain a multi-dimensional feature vector; Step S26: Perform dimensionality reduction and feature selection on the multi-dimensional feature vector to obtain a battery pack operation feature matrix.

[0024] In the embodiment of the present invention, an outlier detection method based on the interquartile range is adopted to calculate Q1 (25th percentile), Q3 (75th percentile) and IQR (interquartile range, equal to Q3-Q1) for each type of data, and the outlier detection method based on the interquartile range is used to calculate Q1 (25th percentile), Q3 (75th percentile) and IQR (interquartile range, equal to Q3-Q1) for each type of data. Q3+1.5×IQR] are marked as outliers; the sliding window median method is used to detect and mark mutation points for voltage data, and the window size is set to 51 data points. When the voltage difference between adjacent data points exceeds 0.1V, it is determined to be a mutation point; the 3σ criterion is applied to the current data, the mean μ and standard deviation σ are calculated, and the points that deviate from the mean by more than 3 standard deviations are marked as outliers; the spatial consistency test is used for temperature data, and when the temperature of a certain point differs from the average temperature of the adjacent 8 points by more than 5℃, it is marked as an outlier; the peak limiting method is used for vibration spectrum data, and isolated peaks with amplitudes exceeding 80% of the global maximum value are marked as interference points; the specific implementation method for noise suppression of the preliminary processed data set is as follows: a Butterworth low-pass filter is applied to the voltage data, the cutoff frequency is set to 500Hz, the filter order is 4th order, and high-frequency noise is eliminated; an adaptive median filter algorithm is applied to the current data, and the window size is between 3 and 1 5 to remove pulse interference; Gaussian filtering is applied to temperature data, with the kernel size set to 5×5 and the standard deviation set to 1.2 to smooth the temperature distribution graph; wavelet threshold denoising is applied to vibration spectrum data, using db4 wavelet basis, 5 decomposition layers, soft threshold function, and threshold coefficient of 3.5; the specific implementation method of standardizing and normalizing the denoised data set is as follows: Min-Max normalization method is used for voltage data to map the voltage value to the [0,1] interval; Z-score normalization method is used for current data to make the data mean 0 and standard deviation 1; maximum temperature normalization is used for temperature data to divide all temperature points by the highest temperature value of the batch; frequency band energy normalization is used for vibration spectrum data to divide the energy of each frequency band by the total energy; the specific implementation method of continuous wavelet transform for standardized data sets is as follows: Mexican hat wavelet (Ricker The temperature data was transformed by two-dimensional continuous wavelet transform, with the scale range set to 1 to 128 and the scale step size set to 1, to obtain the time-frequency energy spectrum; the temperature data was transformed by two-dimensional continuous wavelet transform, using the Haar wavelet basis, and the decomposition scale was set to 5 to obtain the multi-scale characteristics of the temperature field; the vibration spectrum data was transformed by Morlet wavelet transform, with the center frequency set to 0.8125Hz, the scale range is 1 to 64, and a time-frequency diagram with high time-frequency resolution is obtained; the specific implementation method of extracting energy distribution, scale coefficient and frequency characteristics from the time-frequency domain decomposition result is as follows: extracting 20 features including maximum energy proportion, energy entropy, dominant scale value and wavelet variance from the voltage wavelet coefficient; extracting 25 features including frequency band energy distribution, wavelet entropy, scale correlation coefficient and duration distribution from the current wavelet coefficient; extracting 15 features including temperature gradient feature, spatial distribution entropy, local extreme point density and hot spot migration rate from the temperature wavelet coefficient; extracting characteristic frequency energy ratio, ridge length, ridge curvature, instantaneous The time-frequency distribution has 30 features in total; the specific implementation method of dimensionality reduction and feature selection for multidimensional feature vectors is as follows: first, the Pearson correlation coefficient analysis method is used to screen out redundant features with correlations higher than 0.85; then the principal component analysis method is used for dimensionality reduction, and the cumulative contribution rate threshold is set to 95%. The feature vectors after dimensionality reduction are arranged into an m×n dimensional matrix, where m is the number of sampling times and n is the number of features after dimensionality reduction, usually between 20 and 30; finally, the feature vectors of multiple sampling periods are combined into a battery pack operation feature matrix according to the time series, and each row of the matrix represents all the features at a time point, and each column represents the change of a feature at different time points. .

[0025] The present invention can effectively eliminate invalid data and measurement errors, reduce the interference of abnormal points on health assessment, and improve the reliability of data by performing data cleaning and outlier detection on the original monitoring data set. Noise suppression is performed on the preliminary processing data set to further reduce environmental interference and measurement noise, and ensure the stability and accuracy of the data. The denoised data set is standardized and normalized so that data of different dimensions have a consistent scale, improve the stability of model training, and enhance the comparability between different features. The standardized data set is subjected to continuous wavelet transform to achieve time-frequency domain decomposition, so that the battery pack state characteristics can be accurately characterized in two dimensions of time and frequency, and the analysis ability of non-stationary signals is improved. The energy distribution, scale coefficient and frequency characteristics of the time-frequency domain decomposition results are extracted to construct a multidimensional feature vector, so that the description of the battery operation state is more comprehensive and the recognition ability of battery aging and abnormal modes is enhanced. The multidimensional feature vector is subjected to dimensionality reduction and feature selection to remove redundant information, improve calculation efficiency, and construct a battery pack operation feature matrix to provide high-quality input data for the subsequent health status intelligent diagnosis model, thereby improving the accuracy of health assessment and life prediction.

[0026] Preferably, the step S3 of training a preset generative adversarial network using historical data includes: Filter and classify historical data to obtain a normal behavior pattern data set, where the signal-to-noise ratio is greater than or equal to 25dB; Construct a conditional generative adversarial network based on the normal behavior pattern dataset to obtain the initial network structure; The conditional input dimension in the conditional generative adversarial network is 4-8 dimensions, the number of hidden layers is 3-5 layers, and the number of neurons in each layer is 64-256; Initialize the generator parameters of the initial network structure to obtain the generator model, where the learning rate is 0.0001-0.001; Based on the normal behavior pattern data set and the generator model, the discriminator adversarial training is performed to obtain the initial discriminator, wherein the proportion of real samples is 40%-60%, the proportion of generated samples is 40%-60%, and the update frequency of the initial discriminator is every 1-3 generator updates; Iteratively optimize the generator model and the initial discriminator to obtain a generative adversarial network model; Based on the generative adversarial network model, the discriminator part is extracted to obtain the discriminator model.

[0027] The embodiment of the present invention screens the long-term monitoring data in the historical battery pack operation database, extracts the data segments with a signal-to-noise ratio greater than or equal to 25dB, classifies the data through an expert annotation system, distinguishes the normal behavior pattern data set, retains the data samples with high voltage stability, good temperature uniformity, and smooth internal resistance change, and samples the data in layers according to the battery pack operating environment temperature, charge and discharge rate, and number of cycles to ensure the representativeness of the data set; constructs a conditional generative adversarial network based on the normal behavior pattern data set, and the network structure is designed as a four-dimensional conditional input, including temperature, current, internal resistance, and remaining capacity. The number of hidden layers is set to 4 layers, the number of neurons in each layer is 128, the generator adopts a cascade architecture of a deconvolution layer and an upsampling layer, and the discriminator adopts a multi-scale progressive discriminant structure; the initial network structure The generator parameters were initialized, and the Xavier algorithm was selected as the initialization method. The learning rate was set to 0.0005, the momentum coefficient was 0.9, the weight decay coefficient was 1e-5, the weight matrix was randomly initialized using the truncated normal distribution, and the bias term was initialized to 0 to obtain the generator model. The discriminator adversarial training was performed based on the normal behavior pattern dataset and the generator model. The proportion of real samples in the training dataset was set to 50%, the proportion of generated samples was set to 50%, the discriminator update frequency was once every 2 generator updates, the loss function used the Wasserstein distance, and the gradient penalty term was introduced to prevent mode collapse. The generator model and the initial discriminator were iteratively optimized, and the maximum number of iterations was set to 500. When the generated data distribution was Fréchet The training is stopped when the inception distance is less than 0.1, and overfitting is prevented through dynamic learning rate adjustment strategy and early stopping mechanism to obtain a generative adversarial network model. The discriminator part is extracted based on the generative adversarial network model, and all parameters of the discriminator except the last layer are frozen. The discriminator feature extraction network and output layer are retained, and the extracted features are mapped to the 4-dimensional health state space to obtain the final discriminator model, which can extract key features and perform state discrimination in the intelligent diagnosis of battery pack health status.

[0028] The present invention screens and classifies historical data to ensure that the signal-to-noise ratio of the normal behavior pattern data set used for training is greater than or equal to 25dB, thereby improving data quality, reducing noise interference, and enabling the model to learn more accurate health status characteristics. A conditional generative adversarial network is constructed based on the normal behavior pattern data set to form an initial network structure, so that the generator can generate samples that meet the battery health status characteristics under known conditional constraints, thereby improving the model's ability to fit complex health states. The conditional input dimension is set to 4-8 dimensions, the number of hidden layers is 3-5 layers, and the number of neurons in each layer is 64-256, ensuring that the network's expression ability is strong enough to adapt to the health status modeling requirements of different types of battery packs. The generator parameters of the initial network structure are initialized, and the learning rate is set between 0.0001-0.001, so that the generator model can converge stably during the training process and improve the quality of the generated samples. Based on the normal behavior pattern data set and the generator model, the discriminator adversarial training is carried out. By reasonably allocating the ratio of real samples to generated samples (40%-60%) and setting the discriminator update frequency to every 1-3 generator updates, the balance of adversarial training is ensured, mode collapse is avoided, and the model's ability to recognize the true battery health status is improved. The generator model and the initial discriminator are iteratively optimized to gradually converge the generative adversarial network, improve the authenticity and diversity of the generated samples, and make them closer to the true health status distribution. Finally, the discriminator part is extracted based on the trained generative adversarial network model to obtain the discriminator model, so that it can be used independently for health status assessment, improve the accuracy of battery anomaly detection and health assessment, and provide highly reliable discrimination capabilities for subsequent intelligent diagnosis of health status.

[0029] Preferably, in step S3, taking the battery pack operation characteristic matrix as input and training the preset multi-input deep neural network model comprises: The battery pack operation characteristic matrix is ​​grouped according to the data type to obtain a voltage characteristic sub-matrix, a current characteristic sub-matrix, a temperature characteristic sub-matrix and an acoustic characteristic sub-matrix; The voltage feature dimension is 0-20 dimensions, the current feature dimension is 8-16 dimensions, the temperature feature dimension is 12-24 dimensions, and the acoustic feature dimension is 20-40 dimensions; The voltage and current processing is constructed according to the voltage characteristic submatrix and the current characteristic submatrix to obtain a time series electrical characteristic neural network functional unit, wherein the number of hidden layers is 2-3 layers, the number of hidden units is 64-128, and the sequence length is 20-50 time steps; According to the temperature characteristic submatrix, the temperature data processing is constructed to obtain the temperature distribution spatial characteristic neural network functional unit, in which the number of convolution layers is 2-4, the number of convolution kernels is 16-64, and the size of the convolution kernel is 3×3; According to the acoustic feature submatrix, the acoustic vibration data processing is constructed to obtain the vibration spectrum feature neural network functional unit, where the number of frequency segments is 8-16, the number of attention heads is 4-8, and the output feature dimension is 32-64 dimensions; A multi-head attention mechanism is designed based on the time series electrical characteristics neural network functional unit, the temperature distribution spatial characteristics neural network functional unit and the vibration spectrum characteristics neural network functional unit to obtain a weight calculation model, where the number of attention heads is 4-8; The outputs of the time series electrical characteristic neural network functional unit, the temperature distribution space characteristic neural network functional unit and the vibration spectrum characteristic neural network functional unit are subjected to feature fusion through a weight calculation model to obtain a multi-modal fusion feature vector; A fully connected layer is constructed based on the multimodal fusion feature vector to obtain the battery pack health status classifier and degradation degree regressor; The number of fully connected layers is 2-3, where the number of hidden neurons in the first layer is 128-256, which is halved layer by layer, and the drop rate is 0.2-0.5; The battery pack health status classifier and degradation degree regressor are jointly trained, and parameters are optimized based on the joint training results to obtain a battery pack status recognition model.

[0030] In the embodiment of the present invention, the battery pack operation feature matrix is ​​firstly accurately grouped according to the data type to obtain a 16-dimensional voltage feature submatrix, a 12-dimensional current feature submatrix, a 20-dimensional temperature feature submatrix and a 36-dimensional acoustic feature submatrix; a time series electrical characteristic neural network functional unit is constructed according to the voltage feature submatrix and the current feature submatrix, a gated recurrent unit (GRU) structure is adopted, the hidden layer is set to 3 layers, the number of hidden units in each layer is 96, the sequence length is 40 time steps, and the input gate and forget gate thresholds are dynamically adjusted; a temperature distribution spatial feature neural network functional unit is constructed according to the temperature feature submatrix, a spatial residual convolutional neural network is adopted, the number of convolutional layers is 3 layers, the number of convolution kernels in the first layer is 32, the number of convolution kernels in the second layer is 48, and the number of convolution kernels in the third layer is 64, the size of the convolution kernels is 3×3, and batch normalization and spatial attention mechanism are introduced; a vibration spectrum feature neural network functional unit is constructed according to the acoustic feature submatrix, the spectrum data is divided into 12 frequency bands, a multi-head self-attention mechanism is designed, the number of attention heads is 6, the output feature dimension is 48 dimensions, and the residual connection is used to obtain the vibration spectrum feature neural network functional unit. Enhance the feature extraction capability; design a weight calculation model based on the above three functional units, adopt a multi-head attention mechanism, the number of attention heads is set to 6, and the weight calculation method combines dot product attention and additive attention; the outputs of the three functional units are subjected to feature fusion through the weight calculation model to obtain a multimodal fusion feature vector, and residual connection and layer normalization are introduced in the fusion process; a fully connected layer is constructed based on the multimodal fusion feature vector, with 192 hidden neurons in the first layer and 96 in the second layer, the drop rate is set to 0.3, and the activation function selects LeakyReLU to construct a battery pack health state classifier (4 output categories) and a degradation degree regressor (continuous value output); a multi-task joint training method is adopted for the battery pack health state classifier and the degradation degree regressor, and the loss function combines cross entropy loss and mean square error loss, introduces a weight balancing coefficient, and is trained through the Adam optimizer with an initial learning rate of 0.001 and adaptive decay every 10 training rounds. Parameter regularization and early stopping are performed based on the performance on the validation set, and finally a battery pack state recognition model is obtained.

[0031] The present invention groups the battery pack operation feature matrix according to the data type to form a voltage feature submatrix, a current feature submatrix, a temperature feature submatrix and an acoustic feature submatrix, so that the data structure is clearer and the pertinence of feature extraction is improved. The dimension range of different feature submatrices is set to reasonably allocate various feature dimensions, ensure information integrity and improve the expression ability of the model. A timing electrical characteristic neural network functional unit is constructed based on voltage and current characteristics, which can extract the electrical dynamic characteristics during battery operation and improve the sensitivity to changes in battery health status. A temperature distribution space feature neural network functional unit is constructed based on temperature characteristics, and the battery temperature distribution mode is extracted through multi-layer convolution operations to improve the recognition ability of thermal runaway and abnormal heating conditions. A vibration spectrum feature neural network functional unit is constructed based on acoustic characteristics, and the signal analysis ability is enhanced through frequency segmentation and attention mechanism, so as to improve the detection accuracy of abnormal vibration inside the battery. A multi-head attention mechanism is designed for weight calculation, so that the features output by each functional unit can be dynamically assigned weights, and the rationality of multimodal data fusion is improved. The weight calculation model is used for feature fusion to generate a multimodal fusion feature vector, which fully combines electrical, temperature and acoustic information to improve the accuracy of battery health status assessment. Based on the multimodal fusion feature vector, a fully connected layer is constructed to achieve health status classification and degradation degree regression, improving the prediction ability of the model. By jointly training the battery pack health status classifier and degradation degree regressor and optimizing the parameters, the model can accurately identify the battery health status and accurately predict the battery degradation trend, providing efficient support for subsequent health assessment and life prediction.

[0032] Preferably, step S4 comprises the following steps: Step S41: inputting the battery pack operation feature matrix into the discriminator model, calculating the degree of deviation from the normal battery pack behavior pattern, and obtaining a preliminary abnormality measurement index; Step S42: inputting the battery pack operation characteristic matrix into the battery pack state recognition model, obtaining the error vector between the predicted value and the actual value of each parameter of the battery pack, and obtaining the parameter performance attenuation index; Step S43: integrating the preliminary abnormal measurement index and the parameter performance attenuation index to construct a battery pack health index, wherein the battery pack health index ranges from 0 to 100, where 100 indicates a brand new state and 0 indicates complete failure; Step S44: calculating the abnormal probability distribution map of each component of the battery pack based on the parameter performance attenuation index; Step S45: Integrate the data of the battery pack health indicators and the abnormal probability distribution map to generate a battery pack health status assessment report, wherein the battery pack health status assessment report includes a health score, an abnormal area heat map, and a key parameter deviation.

[0033] In the embodiment of the present invention, firstly, the battery pack operation feature matrix is ​​input into the discriminator model, and the Euclidean distance calculation method is used to determine the deviation between the current battery pack operation state and the standard normal battery pack behavior mode, so as to obtain a preliminary abnormal measurement index, and the specific calculation formula is deviation value = sqrt (sum ((current feature - standard feature) ^ 2)); then the battery pack operation feature matrix is ​​input into the battery pack state recognition model, and the performance attenuation index of each parameter is calculated by comparing the predicted value with the actual value using the root mean square error algorithm to obtain the parameter prediction error vector; then, according to the weight fusion method, the preliminary abnormal measurement index and the parameter performance attenuation index are linearly combined to construct a 0-100 range. The battery pack health indicators around the detection range are calculated, where the weight distribution is set according to the empirical coefficient of historical data; based on the parameter performance attenuation index, the probability density estimation method is used to draw the abnormal probability distribution map of each component of the battery pack, and the probability of abnormal occurrence of each component is calculated by the kernel density estimation algorithm; finally, the battery pack health indicators and the abnormal probability distribution map are integrated to generate a battery pack health status assessment report including a health score, an abnormal area heat map and a key parameter deviation. The health score directly corresponds to a health indicator value of 0-100, the abnormal area heat map is based on the color mapping of the probability distribution map, and the key parameter deviation is obtained by calculating the relative deviation percentage of each parameter from the standard value.

[0034] The present invention inputs the battery pack operation feature matrix into the discriminator model and calculates the degree of deviation from the normal battery pack behavior pattern, so that the system can quickly identify the abnormal situation of the battery pack operation state and improve the accuracy of abnormal detection. By inputting the battery pack operation feature matrix into the battery pack state recognition model, the error vector between the predicted value and the actual value of each parameter is obtained, and the attenuation of each battery operation parameter can be quantified, providing a key basis for battery performance evaluation. The initial abnormal measurement index and the parameter performance attenuation index are integrated to construct a health index, and a 0-100 scoring mechanism is adopted to make the evaluation of the battery pack health state more intuitive and convenient for comparative analysis under different operating conditions. Based on the parameter performance attenuation index, the abnormal probability distribution map of each component of the battery pack is calculated, which can locate the specific abnormal occurrence site and improve the accuracy of fault diagnosis. Finally, the battery pack health index and the abnormal probability distribution map are integrated to generate a battery pack health state assessment report, so that the health assessment results are more systematic. The report contains health scores, abnormal area heat maps and key parameter deviations, providing comprehensive health analysis references for operation and maintenance personnel and improving the intelligence level of the battery management system.

[0035] It is particularly important that the battery pack operation feature matrix is ​​input into the discriminator model to calculate the degree of deviation from the normal battery pack behavior mode, including: Performing data normalization processing on the battery pack operation characteristic matrix to obtain a normalized characteristic matrix; Perform feature matching on the discriminator model based on the normalized feature matrix to obtain the current operating feature deviation of the battery pack; Perform statistical analysis on the deviation of the current operating characteristics of the battery pack to obtain the deviation amplitude of each characteristic variable; Based on the deviation amplitude of each characteristic variable, the overall operation mode of the battery pack is clustered and analyzed to obtain the classification result of the battery pack operation mode; Compare and analyze the classification results of the battery pack operation mode to obtain the abnormal deviation value of the battery pack operation mode; Calculate the abnormal probability distribution of the battery pack based on the abnormal deviation value of the battery pack operation mode to obtain a preliminary abnormal probability vector; Perform feature weighted calculation on the preliminary abnormal probability vector to obtain the preliminary abnormality metric index of the battery pack.

[0036] In the embodiment of the present invention, the minimum-maximum normalization process is first performed on the battery pack operation characteristic matrix to map each characteristic variable to the interval [0,1]. The normalization formula is X_norm = (X - X_min) / (X_max - X_min), to ensure that features of different dimensions can be directly compared; then, a feature matching algorithm is used to calculate the Euclidean distance between the normalized feature matrix and the standard normal battery pack feature matrix to obtain the deviation of the current operating features of the battery pack. The specific distance calculation method is sqrt(sum((standard features-current features)^2)); then, a statistical analysis is performed on the deviation of the current operating features of the battery pack, and a descriptive statistical method is used to calculate the deviation amplitude of each feature variable, including the absolute deviation value and the relative deviation percentage; based on the deviation amplitude of each feature variable, a K-means clustering analysis is performed to divide the battery pack operating mode into four categories: healthy, sub-healthy, warning, and dangerous; the classification results of the battery pack operating mode are compared and analyzed, and the abnormal deviation value of the battery pack operating mode is obtained by calculating the distance between each category and the intra-class variance; according to the abnormal deviation value of the operating mode, the probability density estimation method is used to calculate the abnormal probability distribution of the battery pack to generate a preliminary abnormal probability vector; finally, a feature weighted calculation is performed on the preliminary abnormal probability vector, and the weight coefficient is determined based on historical data and expert experience to obtain the preliminary abnormal measurement index of the battery pack.

[0037] The present invention performs data normalization processing on the battery pack operation feature matrix, which can eliminate the dimensional differences of different feature data, improve the comparability of data, and ensure the stability of model calculation. Based on the normalized feature matrix, the discriminator model is feature matched, so that the system can accurately identify the difference between the current operation characteristics of the battery pack and the normal mode, and improve the sensitivity of abnormal detection. By statistically analyzing the deviation of the current operation characteristics of the battery pack, the deviation amplitude of each feature variable can be quantified, and fine-grained abnormal features can be provided for further pattern recognition. Cluster analysis is performed on the overall operation mode of the battery pack based on the deviation amplitude of each feature variable, so that the division of the health state of the battery pack is clearer and the resolution of abnormal detection is improved. Comparative analysis is performed on the classification results of the battery pack operation mode, which can further extract the mode abnormal deviation value and accurately quantify the abnormal degree of the battery pack operation state. Calculating the abnormal probability distribution based on the abnormal deviation value of the battery pack operation mode helps to construct a more explanatory abnormal probability vector and improve the reliability of health assessment. Finally, the preliminary abnormal probability vector is feature-weighted, and the importance of each feature is comprehensively considered to obtain the preliminary abnormal measurement index of the battery pack, which provides a key basis for subsequent health status assessment and early warning.

[0038] Preferably, the step S5 of performing time series decomposition on the battery pack health status assessment data to separate the aging trend and short-term fluctuations includes: Reconstruct the time series of the battery pack health status assessment data to obtain the battery pack health status time series data set; The trend component, cycle component and random component of the battery pack health status time series data set are decomposed to obtain the multi-dimensional decomposition characteristics of the battery pack health status, where the number of IMF components is limited to 3-8; Perform significance screening on the multi-dimensional decomposition features of the battery pack health status to obtain the key features of the battery pack health status; Perform evolution pattern recognition on key features of battery pack health status to obtain the evolution law of battery pack health status; Based on the evolution law of the battery pack health state, the degradation curve of the key characteristics of the battery pack health state is fitted to obtain the battery pack life degradation model; Perform change point detection on the battery pack life degradation model to obtain the key stages of battery pack life; Based on the key stages of battery life, the degradation characteristics of each stage are extracted to obtain the staged life characteristics of the battery pack; Perform feature fusion and normalization on the stage life characteristics of the battery pack to obtain a comprehensive feature set of the battery pack life; Based on historical data, the comprehensive feature set of battery pack life is verified and optimized to obtain the battery pack life evolution characteristics.

[0039] In the embodiment of the present invention, firstly, the time series reconstruction is performed on the battery pack health status assessment data, and the sliding window method is used to convert the discrete data points into a continuous time series, and the reconstruction window length is set to 24 data points and the step size is 6 data points; then, the empirical mode decomposition method is used to decompose the reconstructed time series into trend components, periodic components and random components, and the number of intrinsic mode function (IMF) components is strictly controlled within the range of 3-8, and the instantaneous frequency and amplitude of each IMF component are extracted by the Hilbert-Huang transform algorithm; then, the multi-dimensional decomposition characteristics of the battery pack health status are screened for significance, and the variance contribution rate criterion is used to retain the feature components with an explained variance contribution rate of more than 85%; the evolution pattern recognition is performed on the screened key features, and the cluster analysis method is used to divide the features into three typical evolution stages: initial, stable and decay; based on the evolution law of the battery pack health status, the nonlinear least squares method is used to fit the degradation curve of the key features, and the specific curve model is a two-parameter Weibull distribution function; the change point detection is performed on the battery pack life degradation model, and the PELT (Pruned Exact Linear Time) algorithm is used to identify the key turning points of life; the degradation characteristics of each stage are extracted according to the key life stages, and the characteristic statistics of each stage are calculated, including mean, variance, skewness and kurtosis; Z-score normalization is performed on the staged life characteristics of the battery pack to eliminate the dimension effect; finally, the comprehensive feature set of the battery pack life is cross-validated based on the annotated data in the historical database, and the prediction accuracy of the feature set is evaluated by repeated sampling method to obtain the evolution characteristics of the battery pack life.

[0040] The present invention can effectively retain the dynamic change information of the health status of the battery pack through time series reconstruction, providing a complete data basis for subsequent analysis. The decomposition of trends, cycles and random components helps to accurately extract the long-term degradation trend and short-term fluctuation characteristics of the health status of the battery pack, and improve the resolution ability of the state assessment. Significance screening can reduce redundant information, extract key features, improve analysis efficiency and enhance the generalization ability of the model. Evolutionary pattern recognition can reveal the changing laws of the health status of the battery pack, making the life prediction more accurate and reliable. Degradation curve fitting ensures the rationality of the life model, so that it can accurately describe the battery's decay process. Change point detection can identify the key stages of battery life and provide a basis for health assessment and maintenance strategies. Staged life feature extraction makes the characteristics of different life stages clearer and provides support for differentiated maintenance. Feature fusion and normalization processing improve the consistency and comparability of data, making the prediction ability of the model more stable and reliable. The verification and optimization of historical data enhance the accuracy of life evolution characteristics and improve the adaptability and practicality of the prediction model.

[0041] Preferably, the abnormality level assessment of the battery pack health status assessment data in step S5 includes: Perform historical statistical analysis on battery pack health status assessment data to obtain battery pack health status benchmark thresholds; Based on the battery pack health status benchmark threshold, the battery pack health status assessment data is subjected to deviation calculation to obtain the battery pack health status deviation index; Perform pattern recognition and classification on the health status deviation indicators of the battery pack to obtain the abnormal pattern characteristics of the health status of the battery pack; Based on the abnormal mode characteristics of the battery pack health status, the abnormal severity is quantitatively evaluated to obtain the battery pack abnormality index; The quantitative level of the quantitative assessment of the severity of the abnormality is 5 levels, where the threshold for a slight abnormality is 5%-10% deviation, the threshold for a warning abnormality is 10%-15% deviation, the threshold for a moderate abnormality is 15%-20% deviation, the threshold for a severe abnormality is 20%-25% deviation, and the threshold for a critical abnormality is greater than 25% deviation; The fault impact assessment is performed on the battery pack abnormality index to obtain the battery pack system risk level, where the safety factor weights are 0.4-0.5, the performance is 0.2-0.3, and the life span is 0.2-0.3. The risk score range is 0-100. Perform multi-dimensional abnormal index analysis based on the risk level of the battery pack system to obtain the abnormal association rules of the battery pack; The battery pack abnormality association rules are hierarchically classified into abnormal levels to obtain the battery pack graded abnormality evaluation standard; Based on the battery pack classification abnormality assessment standard, the current battery pack state abnormality level is mapped to obtain the real-time abnormality level result of the battery pack; The real-time abnormal level results of the battery pack are verified and confirmed by domain knowledge rules to obtain the abnormal status level of the battery pack.

[0042] In the embodiment of the present invention, firstly, by performing descriptive statistical analysis on the historical operation data of the battery pack, the median and quartile range of the key parameters of voltage, current and temperature are calculated, and a battery pack health status benchmark threshold is established, wherein the benchmark threshold is based on the 95% confidence interval of the normal operation interval; then, the Euclidean distance calculation method is used to calculate the deviation between the battery pack health status assessment data and the benchmark threshold, and the battery pack health status deviation index is obtained, and the specific calculation formula is sqrt(sum((current data-benchmark threshold)^2)); then, a cluster analysis method is used to perform pattern recognition and classification on the battery pack health status deviation index, and the abnormal pattern is divided into three typical characteristics: clustered abnormality, sudden abnormality, and progressive abnormality; based on the abnormal pattern characteristics, an evaluation is performed according to a predefined 5-level abnormality severity quantitative standard, wherein a slight abnormality corresponds to 5. %-10% deviation, warning abnormality corresponds to 10%-15% deviation, moderate abnormality corresponds to 15%-20% deviation, severe abnormality corresponds to 20%-25% deviation, and critical abnormality corresponds to more than 25% deviation; implement fault impact assessment on the abnormality degree index, build a risk scoring model, with weight ratios of safety factor 0.45, performance factor 0.25, and life factor 0.30, and calculate the system risk level of 0-100 points; use the association rule mining algorithm to perform multi-dimensional abnormal indicator analysis on the battery pack system risk level and extract abnormal association rules; use the hierarchical analysis method to classify the abnormal association rules and build a multi-level abnormal assessment standard; based on the constructed graded abnormal assessment standard, map the current battery pack state to the corresponding abnormal level; finally, verify and confirm the abnormal level through the domain expert knowledge rule base to obtain the final battery pack abnormal state level.

[0043] The present invention determines the health status benchmark threshold through historical statistical analysis, which can provide an objective reference standard and improve the accuracy of abnormality detection. Deviation calculation can quantify the degree of deviation of the battery pack state, making abnormality identification more intuitive. Pattern recognition and classification help to accurately distinguish different types of abnormal patterns and provide a basis for subsequent abnormality processing. The quantitative assessment of the severity of abnormalities makes the abnormality level division more hierarchical through the grading standard, which is convenient for taking targeted maintenance measures. Fault impact assessment can comprehensively consider safety, performance and life factors to improve the comprehensiveness of system risk assessment. Multidimensional abnormal indicator analysis can explore the correlation between abnormalities and improve the intelligent level of abnormality detection. Hierarchical abnormality level division ensures the systematicness and operability of abnormality evaluation, making the evaluation results more scientific and reasonable. Abnormality level mapping can realize real-time judgment of the current state and improve the timeliness of abnormal warning of battery packs. Domain knowledge rule verification enhances the reliability of evaluation results, reduces misjudgment and missed judgment, and improves the accuracy of abnormal state level.

[0044] Preferably, step S6 comprises the following steps: Step S61: performing time series prediction on the battery pack life evolution characteristics to obtain an estimated value of the remaining service life; Step S62: classifying and identifying the abnormal mode of the battery pack based on the abnormal state level of the battery pack to obtain a fault type diagnosis result; Step S63: performing multi-threshold division on the estimated value of the remaining useful life to obtain a health warning level division standard; Step S64: optimizing the configuration of maintenance resources based on the fault type diagnosis result to obtain a differentiated maintenance strategy library; Step S65: obtaining real-time status data of the battery pack; comparing and analyzing the real-time status data of the battery pack with the health warning level classification standard to obtain health warning information; Step S66: optimizing and determining the maintenance timing of the battery pack based on the differentiated maintenance strategy library to obtain a pre-maintenance execution plan; Step S67: Integrate the health warning information and the pre-maintenance execution plan to obtain the battery pack health prediction result.

[0045] In the embodiment of the present invention, the autoregressive integrated moving average (ARIMA) time series prediction method is firstly adopted to construct a prediction model based on the historical operation data and life evolution characteristics of the battery pack, and the model parameters are estimated by the least square method to predict the remaining service life of the battery pack, and the confidence level of the prediction interval is set to 95%; then, according to the abnormal state level of the battery pack, a cluster analysis of the multidimensional feature space is performed to divide the battery pack failure mode into four typical fault types: insulation failure, electrode material degradation, electrolyte degradation, and internal resistance increase; multi-threshold division is implemented for the estimated value of the remaining service life, and a four-level health warning level standard is constructed, which specifically includes: safe use period (>80% life), warning period (60%-80% life), decline period (40%-60% life), and dangerous period (<40% life); based on the fault type diagnosis result Based on the results, a differentiated maintenance strategy library is built through the expert knowledge base and historical maintenance data, including specific maintenance plans for detection, replacement and repair of different fault types; the battery pack operation data, including key parameters such as voltage, current and temperature, is obtained in real time, and the real-time status data is quantitatively compared with the health warning level classification standard, the deviation and abnormality index are calculated, and detailed health warning information is generated; according to the differentiated maintenance strategy library, combined with the current health status, remaining service life and fault type of the battery pack, the best maintenance time is determined through the decision tree algorithm optimization to form a pre-maintenance execution plan; finally, a multi-criteria comprehensive evaluation is conducted on the health warning information and the pre-maintenance execution plan, and the comprehensive score is calculated by the weighted sum method to generate the final battery pack health prediction result, and a complete diagnostic report including health status, warning level and maintenance recommendations is output.

[0046] The time series prediction of the present invention can capture the evolution trend of battery life and improve the accuracy of remaining life estimation. Abnormal pattern classification and recognition can accurately distinguish different fault types, providing a reliable basis for fault diagnosis and maintenance decisions. Multi-threshold division ensures the rationality of health warning levels and makes warning information clearer and more intuitive. The optimal allocation of maintenance resources improves the scientificity and economy of resource allocation and reduces unnecessary maintenance costs. The acquisition and comparison analysis of real-time status data can dynamically monitor the health status of batteries and improve the timeliness of health warnings. Maintenance timing optimization ensures that maintenance measures are performed at the best time, extending the service life of the battery pack and reducing the risk of failure. Integrated decision-making combines multi-dimensional information to improve the comprehensiveness and reliability of health predictions and provide more accurate support for battery management.

[0047] It is particularly important that the comparison and analysis of the real-time status data of the battery pack with the health warning level classification standard includes: Perform time-series alignment on the real-time status data of the battery pack to obtain a standardized real-time operation feature matrix; Based on the standardized real-time operation characteristic matrix, the health warning level classification standard is matched and calculated to obtain the health warning matching degree vector; Perform optimal classification mapping on the health warning matching degree vector to obtain the current health warning level of the battery pack; Based on the current health warning level of the battery pack, the historical health status data is correlated and analyzed to obtain the health status change trend; Perform abnormal analysis on the trend of health status changes to obtain abnormal health status warning indicators; Generate health warning information based on abnormal health status warning indicators, where the health warning information includes the current health warning level, health status trend and abnormal warning information.

[0048] In the embodiment of the present invention, firstly, the real-time status data of the battery pack is subjected to timing alignment processing, the sampling frequency is unified by the linear interpolation method, the voltage, current and temperature multi-dimensional data are resampled to a unified time interval, and the maximum-minimum normalization method is used to map each feature variable to the [0,1] interval to generate a standardized real-time operation feature matrix; then, based on the standardized real-time operation feature matrix, the Euclidean distance calculation method is used to match the health warning level classification standard, and the specific calculation formula is sqrt(sum((real-time feature-standard feature)^2)), and the health warning matching degree vector is obtained; then, the maximum a posteriori probability criterion is used to perform optimal classification mapping on the health warning matching degree vector. The current health status of the battery pack is divided into four levels: safe, warning, decay, and dangerous. Based on the current health warning level of the battery pack, a correlation analysis is performed on the historical health status data of the battery pack, and the sliding window method is used to calculate the slope and acceleration of the health status change in a continuous time period to obtain the health status change trend. An abnormal analysis is performed on the health status change trend to identify abnormal fluctuation characteristics by calculating the trend line variance, the number of inflection points, and the extreme value frequency, and to generate abnormal health status warning indicators. Finally, based on the abnormal health status warning indicators, comprehensive health warning information is generated, including the current health warning level, health status trend description, and abnormal warning information, and the abnormal characteristics are quantitatively and qualitatively evaluated.

[0049] The present invention ensures the synchronization of real-time data, so that the real-time operation feature matrix can match the standardized warning standard, thereby improving the comparability of the data. The health warning matching degree vector generated by matching calculation makes the warning level determination more accurate. The optimal classification mapping can accurately define the current health warning level of the battery pack, providing a clear basis for subsequent decision-making. The correlation analysis of historical health status data reveals the long-term change trend of the health status of the battery pack, which helps to identify potential problems in advance. Abnormal analysis further extracts abnormal warning indicators and improves the sensitivity and accuracy of abnormal detection. The health warning information finally generated comprehensively reflects the changes in the health status of the battery pack, provides timely warnings, and can support the optimization of decision-making and maintenance plans.

[0050] Preferably, the present invention further provides a system for battery health prediction, which is used to execute the above-mentioned method for battery health prediction, and the system for battery health prediction comprises: A multi-modal data acquisition module is used to collect multi-modal sensor data of the battery pack to obtain voltage data, current data, temperature data and acoustic vibration data as the original monitoring data set; Data preprocessing and feature extraction module, used to preprocess the original monitoring data set and extract the frequency domain features, and build the battery pack operation feature matrix; An intelligent diagnosis model building module is used to obtain the historical operation data of the battery pack; the battery pack historical operation data and the battery pack operation feature matrix are used to build an intelligent diagnosis model for the health status of the battery pack, wherein the construction of the intelligent diagnosis model for the health status of the battery pack includes training a generative adversarial network and training a multi-input deep neural network; A health status assessment module is used to assess the battery pack status based on the battery pack health status intelligent diagnosis model to obtain battery pack health status assessment data; The life evolution and anomaly detection module is used to perform time series decomposition on the battery pack health status assessment data, separate the aging trend and short-term fluctuations, and obtain the battery pack life evolution characteristics; perform anomaly level assessment on the battery pack health status assessment data to obtain the battery pack abnormal status level; The health prediction and preventive maintenance module is used to perform health warning and preventive maintenance plan design based on the battery pack life evolution characteristics and battery pack abnormal status level, and obtain the battery pack health prediction results.

[0051] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is not limited by the above description, so it is intended to include all changes within the meaning and scope of the equivalent elements of the application documents in the present invention.

[0052] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for battery pack health prediction, characterized in that: The following steps are involved: Step S1: performing multi-modal sensing data collection on the battery pack to obtain voltage data, current data, temperature data and acoustic vibration data as an original monitoring data set; Step S2: preprocessing the original monitoring data set and extracting the frequency domain features, and constructing the battery pack operation feature matrix; Step S3: obtaining historical operation data of the battery pack; constructing an intelligent diagnosis model for the health status of the battery pack using the historical operation data of the battery pack and the battery pack operation feature matrix, wherein constructing the intelligent diagnosis model for the health status of the battery pack includes training a generative adversarial network and training a multi-input deep neural network; Step S4: evaluating the battery pack status based on the battery pack health status intelligent diagnosis model to obtain battery pack health status evaluation data; Step S5: performing time series decomposition on the battery pack health status assessment data, separating the aging trend and short-term fluctuation, and obtaining the battery pack life evolution characteristics; Performing abnormality level evaluation on the battery pack health status evaluation data to obtain the abnormal status level of the battery pack; Step S6: Perform health warning and preventive maintenance plan design based on the battery pack life evolution characteristics and the battery pack abnormal status level to obtain the battery pack health prediction result.

2. The method for battery health prediction according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: performing high frequency sampling on the terminal voltage of the battery pack to obtain voltage time series data; Step S12: monitor the charge and discharge current of the battery pack in real time to obtain current time series data; Step S13: performing multi-point measurement on the surface temperature of the battery pack based on a distributed temperature sensor array to obtain temperature distribution data; Step S14: collecting acoustic vibration signals through a piezoelectric sensor attached to the surface of the battery pack to obtain vibration spectrum data; Step S15: time-synchronize and integrate the voltage time series data, current time series data, temperature distribution data and vibration spectrum data to obtain a multi-modal original monitoring data set.

3. The method for battery health prediction according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Perform data cleaning and outlier detection on the original monitoring data set to obtain a preliminary processed data set with noise points and outliers removed; Step S22: performing noise suppression on the preliminary processed data set to obtain a denoised data set; Step S23: standardizing and normalizing the denoised data set to obtain a standardized data set; Step S24: performing continuous wavelet transform on the standardized data set to obtain a time-frequency domain decomposition result; Step S25: extracting energy distribution, scale coefficient and frequency characteristics from the time-frequency domain decomposition result to obtain a multi-dimensional feature vector; Step S26: Perform dimensionality reduction and feature selection on the multi-dimensional feature vector to obtain a battery pack operation feature matrix.

4. The method for battery health prediction according to claim 1, characterized in that: The step S3 of training the preset generative adversarial network using historical data includes: The historical data is screened and classified to obtain a normal behavior pattern data set, where the signal-to-noise ratio is greater than or equal to 25dB; Construct a conditional generative adversarial network based on the normal behavior pattern dataset to obtain the initial network structure; The conditional input dimension in the conditional generative adversarial network is 4-8 dimensions, the number of hidden layers is 3-5 layers, and the number of neurons in each layer is 64-256; Initialize the generator parameters of the initial network structure to obtain the generator model, where the learning rate is 0.0001-0.001; Based on the normal behavior pattern data set and the generator model, the discriminator adversarial training is performed to obtain the initial discriminator, wherein the proportion of real samples is 40%-60%, the proportion of generated samples is 40%-60%, and the update frequency of the initial discriminator is every 1-3 generator updates; Iteratively optimize the generator model and the initial discriminator to obtain a generative adversarial network model; Based on the generative adversarial network model, the discriminator part is extracted to obtain the discriminator model.

5. The method for battery health prediction according to claim 4, characterized in that: In step S3, the battery pack operation feature matrix is ​​used as input to train the preset multi-input deep neural network model, including: The battery pack operation characteristic matrix is ​​grouped according to the data type to obtain a voltage characteristic sub-matrix, a current characteristic sub-matrix, a temperature characteristic sub-matrix and an acoustic characteristic sub-matrix; The voltage feature dimension is 0-20 dimensions, the current feature dimension is 8-16 dimensions, the temperature feature dimension is 12-24 dimensions, and the acoustic feature dimension is 20-40 dimensions; The voltage and current processing is constructed according to the voltage characteristic submatrix and the current characteristic submatrix to obtain a time series electrical characteristic neural network functional unit, wherein the number of hidden layers is 2-3 layers, the number of hidden units is 64-128, and the sequence length is 20-50 time steps; According to the temperature characteristic submatrix, the temperature data processing is constructed to obtain the temperature distribution spatial characteristic neural network functional unit, in which the number of convolution layers is 2-4, the number of convolution kernels is 16-64, and the size of the convolution kernel is 3×3; According to the acoustic feature submatrix, the acoustic vibration data processing is constructed to obtain the vibration spectrum feature neural network functional unit, where the number of frequency segments is 8-16, the number of attention heads is 4-8, and the output feature dimension is 32-64 dimensions; A multi-head attention mechanism is designed based on the time series electrical characteristics neural network functional unit, the temperature distribution spatial characteristics neural network functional unit and the vibration spectrum characteristics neural network functional unit to obtain a weight calculation model, where the number of attention heads is 4-8; The outputs of the time series electrical characteristic neural network functional unit, the temperature distribution space characteristic neural network functional unit and the vibration spectrum characteristic neural network functional unit are subjected to feature fusion through a weight calculation model to obtain a multi-modal fusion feature vector; A fully connected layer is constructed based on the multimodal fusion feature vector to obtain the battery pack health status classifier and degradation degree regressor; The number of fully connected layers is 2-3, where the number of hidden neurons in the first layer is 128-256, which is halved layer by layer, and the drop rate is 0.2-0.5; The battery pack health status classifier and degradation degree regressor are jointly trained, and parameters are optimized based on the joint training results to obtain a battery pack status recognition model.

6. The method for battery health prediction according to claim 5, characterized in that: Step S4 includes the following steps: Step S41: inputting the battery pack operation feature matrix into the discriminator model, calculating the degree of deviation from the normal battery pack behavior pattern, and obtaining a preliminary abnormality measurement index; Step S42: inputting the battery pack operation characteristic matrix into the battery pack state recognition model, obtaining the error vector between the predicted value and the actual value of each parameter of the battery pack, and obtaining the parameter performance attenuation index; Step S43: integrating the preliminary abnormal measurement index and the parameter performance attenuation index to construct a battery pack health index, wherein the battery pack health index ranges from 0 to 100, where 100 indicates a brand new state and 0 indicates complete failure; Step S44: calculating the abnormal probability distribution map of each component of the battery pack based on the parameter performance attenuation index; Step S45: Integrate the data of the battery pack health indicators and the abnormal probability distribution map to generate a battery pack health status assessment report, wherein the battery pack health status assessment report includes a health score, an abnormal area heat map, and a key parameter deviation.

7. The method for battery health prediction according to claim 1, characterized in that: The step S5 described in which the battery pack health status assessment data is decomposed in time series to separate the aging trend and short-term fluctuation includes: Reconstruct the time series of the battery pack health status assessment data to obtain the battery pack health status time series data set; The trend component, cycle component and random component of the battery pack health status time series data set are decomposed to obtain the multi-dimensional decomposition characteristics of the battery pack health status, where the number of IMF components is limited to 3-8; Perform significance screening on the multi-dimensional decomposition features of the battery pack health status to obtain the key features of the battery pack health status; Perform evolution pattern recognition on key features of battery pack health status to obtain the evolution law of battery pack health status; Based on the evolution law of the battery pack health state, the degradation curve of the key characteristics of the battery pack health state is fitted to obtain the battery pack life degradation model; Perform change point detection on the battery pack life degradation model to obtain the key stages of battery pack life; Based on the key stages of battery life, the degradation characteristics of each stage are extracted to obtain the staged life characteristics of the battery pack; Perform feature fusion and normalization on the stage life characteristics of the battery pack to obtain a comprehensive feature set of the battery pack life; Based on historical data, the comprehensive feature set of battery pack life is verified and optimized to obtain the battery pack life evolution characteristics.

8. The method for battery health prediction according to claim 7, characterized in that: The abnormality level assessment of the battery pack health status assessment data in step S5 includes: Perform historical statistical analysis on battery pack health status assessment data to obtain battery pack health status benchmark thresholds; Based on the battery pack health status benchmark threshold, the battery pack health status assessment data is subjected to deviation calculation to obtain the battery pack health status deviation index; Perform pattern recognition and classification on the health status deviation indicators of the battery pack to obtain the abnormal pattern characteristics of the health status of the battery pack; Based on the abnormal mode characteristics of the battery pack health status, the abnormal severity is quantitatively evaluated to obtain the battery pack abnormality index; The quantitative level of the quantitative assessment of the severity of the abnormality is 5 levels, where the threshold for a slight abnormality is 5%-10% deviation, the threshold for a warning abnormality is 10%-15% deviation, the threshold for a moderate abnormality is 15%-20% deviation, the threshold for a severe abnormality is 20%-25% deviation, and the threshold for a critical abnormality is greater than 25% deviation; The fault impact assessment is performed on the battery pack abnormality index to obtain the battery pack system risk level, where the safety factor weights are 0.4-0.5, the performance is 0.2-0.3, and the life span is 0.2-0.

3. The risk score range is 0-100. Perform multi-dimensional abnormal index analysis based on the risk level of the battery pack system to obtain the abnormal association rules of the battery pack; The battery pack abnormality association rules are hierarchically classified into abnormal levels to obtain the battery pack graded abnormality evaluation standard; Based on the battery pack classification abnormality assessment standard, the current battery pack state abnormality level is mapped to obtain the real-time abnormality level result of the battery pack; The real-time abnormal level results of the battery pack are verified and confirmed by domain knowledge rules to obtain the abnormal status level of the battery pack.

9. The method for battery health prediction according to claim 1, characterized in that: Step S6 includes the following steps: Step S61: performing time series prediction on the battery pack life evolution characteristics to obtain an estimated value of the remaining service life; Step S62: classifying and identifying the abnormal mode of the battery pack based on the abnormal state level of the battery pack to obtain a fault type diagnosis result; Step S63: performing multi-threshold division on the estimated value of the remaining useful life to obtain a health warning level division standard; Step S64: optimizing the configuration of maintenance resources based on the fault type diagnosis result to obtain a differentiated maintenance strategy library; Step S65: obtaining real-time status data of the battery pack; comparing and analyzing the real-time status data of the battery pack with the health warning level classification standard to obtain health warning information; Step S66: optimizing and determining the maintenance timing of the battery pack based on the differentiated maintenance strategy library to obtain a pre-maintenance execution plan; Step S67: Integrate the health warning information and the pre-maintenance execution plan to obtain the battery pack health prediction result.

10. A system for battery pack health prediction, characterized in that: For executing the method for battery health prediction according to claim 1, the system for battery health prediction comprises: A multi-modal data acquisition module is used to collect multi-modal sensor data of the battery pack to obtain voltage data, current data, temperature data and acoustic vibration data as the original monitoring data set; Data preprocessing and feature extraction module, used to preprocess the original monitoring data set and extract the frequency domain features, and build the battery pack operation feature matrix; An intelligent diagnosis model building module is used to obtain the historical operation data of the battery pack; the battery pack historical operation data and the battery pack operation feature matrix are used to build an intelligent diagnosis model for the health status of the battery pack, wherein the construction of the intelligent diagnosis model for the health status of the battery pack includes training a generative adversarial network and training a multi-input deep neural network; A health status assessment module is used to assess the battery pack status based on the battery pack health status intelligent diagnosis model to obtain battery pack health status assessment data; The life evolution and anomaly detection module is used to perform time series decomposition on the battery pack health status assessment data, separate the aging trend and short-term fluctuations, and obtain the battery pack life evolution characteristics; perform anomaly level assessment on the battery pack health status assessment data to obtain the battery pack abnormal status level; The health prediction and preventive maintenance module is used to perform health warning and preventive maintenance plan design based on the battery pack life evolution characteristics and battery pack abnormal status level, and obtain the battery pack health prediction results.

Images