A power balancing management method for series-connected lithium battery packs

Abstract

This invention discloses a power balancing management method for series-connected lithium battery packs, comprising: collecting the original dataset of the series-connected lithium battery pack; calculating the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude to obtain a dynamic feature set of battery operation; obtaining a short-circuit risk assessment result when the dynamic feature set exceeds a corresponding preset threshold; matching a pre-established battery pack topology diagram with the short-circuit risk assessment result to determine the target battery identifier that needs to be isolated, generating an isolation command using a relay control algorithm to complete the isolation of the short-circuited battery; adjusting the updated charging current distribution of the battery pack through a balancing management algorithm to obtain the balanced battery operating parameters; and determining that the system stability has been restored and obtaining the final battery pack operating state if the voltage, temperature, and internal resistance data of the remaining individual cells are all within the normal range. This invention can effectively prevent short-circuit risks in lithium battery packs and achieve intelligent management and improved safety of battery packs.

Description

Technical Field

[0001] This invention belongs to the field of new energy battery management and safety control technology, and in particular relates to a power balancing management method for series lithium battery packs. Background Technology

[0002] In the new energy field, lithium battery pack power management is a core technology for ensuring the efficient operation, safety, and lifespan of battery systems, playing a crucial role, especially in applications such as electric vehicles and renewable energy storage. With the increasing prevalence of series-connected lithium battery packs, maintaining the balance between individual cells and the overall stability of the system has become a technical challenge. While current battery balancing management methods can partially alleviate battery inconsistency issues, significant shortcomings remain.

[0003] Traditional solutions typically rely on a single parameter (such as voltage) for equalization control or employ passive energy dissipation methods (such as resistor discharge) to eliminate charge differences. These methods are inefficient and struggle to cope with sudden failures, especially under extreme conditions like short circuits, where response delays can easily lead to serious consequences such as thermal runaway. In series-connected lithium-ion battery packs, short circuits, due to their insidious nature and destructive potential, have become a critical technical bottleneck that urgently needs to be overcome. The main shortcomings of existing technologies are as follows:

[0004] Limited monitoring dimensions: Traditional methods often rely on voltage parameters for judgment, resulting in incomplete capture of short-circuit precursor information and an inability to achieve accurate early warning.

[0005] Insufficient short-circuit feature modeling: Existing models are unable to accurately characterize the micro-dynamic behavior in the early stages of a short circuit, resulting in a lag in fault identification.

[0006] Inefficient fault isolation mechanism: After a short circuit occurs, the lack of a fast and reliable isolation strategy leads to the spread of the fault, threatening the safety of the entire battery pack.

[0007] In summary, existing technologies have significant shortcomings in multi-dimensional real-time monitoring, accurate identification of short-circuit precursors, and rapid fault isolation, resulting in low short-circuit detection accuracy and slow response speed, making it difficult to meet high safety requirements. Therefore, there is an urgent need to propose a power balancing management method for series-connected lithium battery packs to improve their safety and reliability. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention proposes a power balancing management method for series-connected lithium battery packs, thereby resolving the issues present in the prior art.

[0009] To achieve the above objectives, the present invention provides a power balancing management method for a series-connected lithium battery pack, comprising the following steps:

[0010] Real-time acquisition of multi-dimensional monitoring information of series-connected lithium battery packs, including voltage, temperature, and internal resistance data of each individual cell, to construct the original dataset;

[0011] Based on the original dataset, time series analysis algorithms are used to calculate the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude to obtain the dynamic feature set of battery operation.

[0012] A short-circuit precursor analysis model is obtained by classifying and training the dynamic feature set of battery operation using the support vector machine algorithm.

[0013] Based on the output of the short-circuit precursor analysis model, when the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude exceed the corresponding preset thresholds, it is determined that there is a short-circuit precursor in the current single cell, and the short-circuit risk assessment result is obtained.

[0014] The pre-established battery pack topology diagram is matched with the short-circuit risk assessment results to determine the location of individual cells related to short-circuit precursors and obtain the target cell identifiers that need to be isolated.

[0015] Based on the target battery identifier, an isolation command is generated using a relay control algorithm. By disconnecting the connection path of the corresponding relay, the short-circuited battery is isolated, and the updated battery pack status is obtained.

[0016] The voltage, temperature, and internal resistance data of the remaining individual cells are extracted from the updated battery pack status. The charging current distribution is adjusted through a balancing management algorithm to obtain the balanced battery operating parameters.

[0017] Based on the balanced battery operating parameters, if the voltage, temperature, and internal resistance of the remaining individual cells are all within the normal range, then the system stability is determined to have been restored, and the final battery pack operating status is obtained.

[0018] Optionally, the real-time acquisition of multi-dimensional monitoring information of the series-connected lithium battery pack, including the voltage, temperature, and internal resistance data of each individual cell, constructs an original dataset, including:

[0019] By installing corresponding sensors on each individual cell of a series-connected lithium battery pack, the voltage, temperature, and internal resistance data of each individual cell can be collected in real time.

[0020] The voltage, temperature, and internal resistance data of each individual battery cell are filtered, amplified, and converted from analog to digital to obtain multi-dimensional data.

[0021] Based on the battery equivalent circuit model, time-series alignment and outlier removal are performed on multi-dimensional data to form an original dataset containing the three-dimensional coupling relationship of voltage, temperature and internal resistance.

[0022] Optionally, based on the original dataset, a time series analysis algorithm is used to calculate the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude to obtain a dynamic feature set of battery operation, including:

[0023] Based on the original dataset, the voltage change rate is calculated by first-order difference, the temperature rise rate is obtained by derivative after smoothing the temperature curve by Savitzky-Golay filtering, and the internal resistance fluctuation amplitude is quantified by combining the sliding standard deviation.

[0024] Meanwhile, a dynamic time warping algorithm is introduced to align the characteristic timing differences of different battery cells, ultimately obtaining a dynamic feature set of battery operation.

[0025] Optionally, the output results of the short-circuit precursor analysis model, when the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude exceed the corresponding preset thresholds, determine that the current single cell has short-circuit precursors, and obtain short-circuit risk assessment results, including:

[0026] The classification boundary is obtained from the short-circuit precursor analysis model to determine the separation threshold between abnormal data and normal data;

[0027] To determine the segmentation threshold, a sliding window is used to segment the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude to obtain a segmented feature sequence.

[0028] Based on the segmented feature sequence, calculate the statistical values ​​of the features within each segment to obtain a subset of abnormal features;

[0029] If the statistical value of a feature segment in the abnormal feature subset exceeds a preset threshold, it is determined to be an abnormal state through logical judgment, and an abnormal state set is obtained.

[0030] By using the set of abnormal states, the K-means clustering algorithm is used to classify the abnormal data and obtain a set of abnormal categories.

[0031] Based on the mapping relationship between the set of anomaly categories and the running status, the status label corresponding to each anomaly category is determined to obtain the status classification result;

[0032] Based on the state classification results and combined with the operating data of individual cells, the risk assessment results of short circuit precursors are determined.

[0033] Optionally, the step of matching the pre-established battery pack topology diagram with the short-circuit risk assessment results to determine the location of individual cells related to short-circuit precursors and obtain the target cell identifier that needs to be isolated includes:

[0034] By using a pre-established battery pack topology diagram, the location of individual cells corresponding to short-circuit risks is obtained, and the initial target cell identifier is determined.

[0035] Based on the initial target battery identifier, a preset threshold is used to determine the isolation requirements, resulting in the set of batteries that need to be isolated;

[0036] From the set of batteries that need to be isolated, obtain the adjacency relationship of individual batteries in the topology to determine the range of affected battery packs;

[0037] Based on the affected battery pack range, the propagation path of short-circuit precursors is analyzed using structural diagrams to obtain the risk diffusion trend;

[0038] In response to the risk spread trend, if there are battery segments in the propagation path that exceed a preset threshold, they will be identified as a high-risk battery subset through logical judgment.

[0039] From the high-risk battery subset, the K-means clustering algorithm is used to classify the risk assessment results and obtain the abnormal battery categories;

[0040] Based on the abnormal battery category, the isolation requirements are updated through the topology diagram to determine the final set of target battery identifiers.

[0041] Optionally, the step of generating an isolation command using a relay control algorithm based on the target battery identifier, and isolating the short-circuited battery by disconnecting the connection path of the corresponding relay to obtain the updated battery pack status includes:

[0042] By identifying the target battery, the topology data of the corresponding relay path is obtained, and the priority order of disconnection operations is determined.

[0043] Based on the priority order, a relay control algorithm is used to generate an isolated instruction sequence, resulting in a set of instructions to be executed;

[0044] For the instruction set, the disconnection condition of the relay path is determined by a preset threshold. If the load of the relay path exceeds the threshold, a disconnection signal is generated.

[0045] From the disconnection signal, obtain the range of affected connection paths and determine the isolation status of the short-circuited battery;

[0046] Based on the isolation status, the K-means clustering algorithm is used to classify the battery pack status and obtain the distribution characteristics of abnormal batteries;

[0047] By updating the topology record of the battery pack status based on distribution characteristics, the isolation completion rate of the target battery can be determined.

[0048] The updated battery pack operating parameters are obtained from the isolation completion status, and thus the updated battery pack status is obtained.

[0049] Optionally, the step of extracting the voltage, temperature, and internal resistance data of the remaining individual cells from the updated battery pack state, and adjusting the charging current distribution through a balancing management algorithm to obtain balanced battery operating parameters includes:

[0050] The voltage, temperature, and internal resistance data of individual cells are obtained from the battery pack status. The equalization management algorithm is used to calculate the adjustment value of the charging current and obtain a preliminary current distribution scheme.

[0051] For the initial current distribution scheme, a preset threshold is used to determine whether the voltage data of a single battery cell exceeds the range. If it does, the charging current is adjusted to obtain the corrected current distribution result.

[0052] Based on the corrected current distribution results, the temperature data change trend of individual cells is obtained, and statistical methods are used to analyze the distribution characteristics of the temperature data to determine the temperature equilibrium state.

[0053] By obtaining the fluctuation range of internal resistance data through the temperature equilibrium state, it is determined whether the internal resistance data meets the requirements of equilibrium management, and the operating parameters after internal resistance adjustment are obtained.

[0054] The actual distribution value of the charging current is extracted from the operating parameters after internal resistance adjustment, and the current distribution is optimized by the equalization management algorithm to obtain the balanced battery operating parameters.

[0055] Optionally, the step of determining system stability recovery and obtaining the final battery pack operating state based on the balanced battery operating parameters, if the voltage, temperature, and internal resistance data of the remaining individual cells are all within the normal range, includes:

[0056] The voltage difference data of individual cells is obtained from the balanced battery operating parameters. The voltage difference is judged to meet the range requirements by using a preset threshold, and a preliminary stability assessment result is obtained.

[0057] Based on the preliminary stability assessment results, the distribution of temperature data was obtained, and statistical methods were used to analyze the variation characteristics of the temperature data to determine the equilibrium state of the temperature distribution.

[0058] By measuring the equilibrium state of the temperature distribution, the distribution of internal resistance data is obtained, and it is determined whether the internal resistance data is within the preset range, thus obtaining the basis for adjusting the internal resistance distribution.

[0059] Based on the adjustment criteria for internal resistance distribution, real-time data of individual cells are extracted from the operating parameters to obtain an optimized parameter set;

[0060] Based on the optimized parameter set, the overall trend of battery pack status change is obtained, and the system stability is determined by logical judgment to obtain the stability confirmation result.

[0061] Extract operational conclusions from the stability confirmation results, update the recorded data of the battery pack status, and determine the balanced state of the system operation;

[0062] By recording data under equilibrium conditions, the long-term operating trend of individual cells can be obtained, the stability of the battery pack can be determined, and the final operating status of the battery pack can be obtained.

[0063] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0064] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0065] Compared with the prior art, the present invention has the following advantages and technical effects:

[0066] This invention discloses a power balancing management method for series-connected lithium battery packs. This method collects multi-dimensional monitoring information from the battery pack in real time, extracts dynamic features, and constructs a short-circuit precursor analysis model. Based on the model output, the short-circuit risk is assessed, and the at-risk battery is located in the battery pack topology diagram. Subsequently, this invention rapidly isolates the short-circuited battery using a relay control algorithm and adjusts the charging current distribution of the remaining batteries using a balancing management algorithm. Finally, this invention determines whether system stability has been restored, ensuring the safe operation of the battery pack. This method can effectively prevent short-circuit risks in lithium battery packs, achieve intelligent management and safety improvement of the battery pack, extend battery pack lifespan, and improve energy utilization efficiency. Attached Figure Description

[0067] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0068] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of the process for constructing a short-circuit precursor analysis model according to an embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of the process for updating the battery pack operating status according to an embodiment of the present invention. Detailed Implementation

[0071] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0072] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0073] Example 1

[0074] like Figure 1 As shown, this embodiment provides a power balancing management method for a series-connected lithium battery pack, including the following steps:

[0075] Real-time acquisition of multi-dimensional monitoring information of series-connected lithium battery packs, including voltage, temperature, and internal resistance data of each individual cell, to construct the original dataset;

[0076] Based on the original dataset, time series analysis algorithms are used to calculate the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude to obtain the dynamic feature set of battery operation.

[0077] A short-circuit precursor analysis model is obtained by classifying and training the dynamic feature set of battery operation using the support vector machine algorithm.

[0078] Based on the output of the short-circuit precursor analysis model, when the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude exceed the corresponding preset thresholds, it is determined that there is a short-circuit precursor in the current single cell, and the short-circuit risk assessment result is obtained.

[0079] The pre-established battery pack topology diagram is matched with the short-circuit risk assessment results to determine the location of individual cells related to short-circuit precursors and obtain the target cell identifiers that need to be isolated.

[0080] Based on the target battery identifier, an isolation command is generated using a relay control algorithm. By disconnecting the connection path of the corresponding relay, the short-circuited battery is isolated, and the updated battery pack status is obtained.

[0081] The voltage, temperature, and internal resistance data of the remaining individual cells are extracted from the updated battery pack status. The charging current distribution is adjusted through a balancing management algorithm to obtain the balanced battery operating parameters.

[0082] Based on the balanced battery operating parameters, if the voltage, temperature, and internal resistance of the remaining individual cells are all within the normal range, then the system stability is determined to have been restored, and the final battery pack operating status is obtained.

[0083] As a specific implementation method, the following steps are included:

[0084] S101. Real-time collection of multi-dimensional monitoring information of series-connected lithium battery packs through sensors, including voltage, temperature and internal resistance data of each individual cell, to construct a raw dataset reflecting the battery status.

[0085] By installing corresponding sensors on each individual cell of a series-connected lithium battery pack, multi-dimensional monitoring information such as voltage, temperature, and internal resistance of each cell can be collected in real time. After filtering, amplification, and analog-to-digital conversion by the signal conditioning circuit, the multi-dimensional data with timestamp synchronization is transmitted to the host computer by an embedded microcontroller (such as STM32) via CAN bus or DAQs system. Based on the battery equivalent circuit model, such as the Thevenin model, the original data is time-aligned and outliers are removed to form an original dataset containing the three-dimensional coupling relationship of voltage, temperature, and internal resistance.

[0086] It is feasible to collect multi-dimensional information such as voltage, temperature, and internal resistance in real time by installing sensors on each individual cell of a series-connected lithium battery pack. For example, in a battery pack consisting of 10 individual cells, each cell can be equipped with a voltage sensor and a temperature sensor. The voltage acquisition range is set to 2.5V to 4.2V, and the temperature range is -20℃ to 60℃. The internal resistance is indirectly obtained by periodically applying a small AC signal and measuring the response; a typical value may be between 20mΩ and 50mΩ. This multi-dimensional monitoring can comprehensively reflect the battery's health status and prevent the performance of the entire battery pack from deteriorating due to a single point of failure.

[0087] In one possible implementation, the signal conditioning circuit processes the acquired raw signal. For example, the voltage signal is passed through a low-pass filter to remove high-frequency noise, and then amplified by an operational amplifier to a range of 0V to 3.3V to meet the input requirements of the subsequent analog-to-digital converter. The temperature signal is converted to a voltage value by a thermistor bridge circuit and then amplified. This processing ensures the accuracy and stability of the data, providing a reliable foundation for subsequent analysis.

[0088] Furthermore, after analog-to-digital conversion, embedded microcontrollers such as STM32 transmit data via the CAN bus. Understandably, the CAN bus is widely used due to its high reliability and real-time performance. For example, it can transmit 100 sets of data per second, each set containing voltage, temperature, internal resistance, and a timestamp with a timestamp accuracy of 1ms. This synchronous transmission avoids data misalignment, ensuring that the information received by the host computer is highly consistent in timing, thereby supporting accurate status assessment.

[0089] It's important to note that when processing data based on the Thevenin model, time alignment is performed first. For example, if a battery's voltage is 3.8V and temperature is 35℃ at the 5th second, but its internal resistance data appears at the 5.2th second due to acquisition delay, it will be aligned to the 5th second using interpolation. Outlier removal can be achieved by setting thresholds; for instance, data with voltages exceeding 4.3V or temperatures above 70℃ are considered outliers and removed. This method improves the reliability of the dataset.

[0090] In one embodiment, the resulting voltage-temperature-internal resistance three-dimensional coupled dataset can be used for battery state analysis. For example, at a low temperature of 20°C, the internal resistance may rise to 45 mΩ while the voltage drops to 3.6 V, indicating battery capacity degradation; whereas at a high temperature of 50°C, the internal resistance drops to 25 mΩ while the voltage remains at 4.0 V, but this may indicate an overheating risk. Analysis of this three-dimensional relationship helps predict battery life and optimize charge / discharge strategies.

[0091] Preferably, through the above scheme, the host computer can display the status of each battery cell in real time and generate trend charts. For example, after 100 hours of continuous operation, if the internal resistance of a certain battery cell increases from 30mΩ to 40mΩ and the temperature rises by 5°C, maintenance is required. This visualization and predictive capability significantly improves the safety and efficiency of the battery pack.

[0092] In one embodiment, the integration of the sensor and microcontroller also supports dynamic adjustments. For example, when a battery cell's temperature is detected to exceed 45°C, the system automatically reduces the charging current from 2A to 1A, thereby reducing heat buildup. This adaptive management extends battery life and reduces the failure rate.

[0093] S102. Extract dynamic features from the original dataset, and use time series analysis to calculate the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude to obtain the dynamic feature set of battery operation.

[0094] Based on the original dataset, a sliding time window (e.g., a 5-second window) is used to segment the voltage, temperature, and internal resistance data. The voltage change rate (dV / dt) is calculated using first-order difference to reflect the dynamic characteristics of charging and discharging. The temperature curve is smoothed using Savitzky-Golay filtering, and the derivative is used to obtain the temperature surge rate (dT / dt) to capture early signs of thermal runaway. The internal resistance fluctuation amplitude (ΔR) is quantified using the sliding standard deviation (30-second rolling window) to characterize the degree of battery aging. Simultaneously, a dynamic time warping (DTW) algorithm is introduced to align the temporal differences of different battery cells. Finally, a dynamic feature set containing temporal gradient features and statistical features is constructed, providing a multi-dimensional feature vector reflecting transient characteristics and trend evolution for subsequent battery state assessment. This process, by integrating signal processing and time series analysis, transforms static monitoring data into dynamic behavioral indicators, effectively enhancing the sensitivity to abnormal battery operating conditions.

[0095] A feasible approach is to use a sliding time window, a common method for processing time series data. This involves segmenting the data into fixed time intervals for analysis. For example, within a 5-second window, voltage, temperature, and internal resistance data can be divided into multiple consecutive segments, each containing sampling points within a 5-second interval. This method can capture short-term trends.

[0096] Understandably, for a system with a sampling frequency of 10Hz, there will be 50 data points within a 5-second window. The voltage may slowly rise from 3.7V to 3.8V, the temperature may slightly increase from 30℃ to 31℃, while the internal resistance remains around 35mΩ. By segmenting the data, the system can observe battery behavior in a finer granular manner. The purpose of calculating the voltage change rate using first-order differential calculation is to reflect the dynamic characteristics of the battery during charging and discharging.

[0097] For example, for voltage data within a window, assuming adjacent sampling points are 3.75V and 3.76V respectively, with a time interval of 0.1 seconds, the voltage change rate is approximately 0.1V / s. This rate of change may be small at the beginning of charging, but may increase significantly as the battery approaches full charge, indicating a change in charging state.

[0098] Specifically, if a battery cell's voltage rises from 3.6V to 3.9V within 5 seconds, the rate of change is 0.06V / s, indicating that it is in the fast charging phase. This analysis helps determine the battery's real-time operating status. The Savitzky-Golay filter is used to smooth the temperature curve and calculate the rate of temperature rise, in order to identify the risk of thermal runaway in advance.

[0099] In one possible implementation, assuming the temperature data exhibits minor fluctuations, such as 30.1℃, 30.3℃, and 30.2℃, a smoothed value of 30.2℃ is obtained after filtering. The rate of temperature rise is then calculated by taking the derivative. If the temperature rapidly rises from 32℃ to 35℃ within a certain period, at a rate of 0.6℃ / s, it may indicate insufficient heat dissipation or abnormal heating. This smoothing process avoids noise interference and improves the sensitivity of anomaly detection.

[0100] The sliding standard deviation is used to quantify the fluctuation range of internal resistance, characterizing the degree of battery aging. For example, if the internal resistance fluctuates between 33mΩ and 37mΩ within a 30-second window, the standard deviation is approximately 1.5mΩ, indicating good internal resistance stability. However, if the fluctuation range expands to 30mΩ to 45mΩ, and the standard deviation increases to 5mΩ, it may reflect electrode material degradation or poor contact.

[0101] Preferably, this statistical characteristic can intuitively reflect the trend of battery performance degradation over time.

[0102] Dynamic Time Warping (DTW) is used to align the characteristic timing differences between different battery cells. It's important to note that since the sampling times of each battery cell may have slight deviations—for example, the voltage peak of one cell might occur at 10 seconds, while another's occurs at 10.2 seconds—DTW achieves alignment by stretching or compressing the time axis. For instance, in a 10-cell battery pack, the internal resistance changes of cells 1 and 5 may have similar trends but a 0.3-second time offset. DTW can match their characteristic sequences, unifying them to the same time reference. This alignment ensures the consistency of data across multiple battery cells.

[0103] In one embodiment, the constructed dynamic feature set integrates temporal gradient features and statistical features. For example, the feature vector of a certain battery cell may include a voltage change rate of 0.05V / s, a temperature rise rate of 0.2℃ / s, and an internal resistance fluctuation of 2mΩ. These features not only reflect transient characteristics but also reveal long-term trend evolution, providing a multi-dimensional perspective for subsequent state assessment.

[0104] Understandably, this feature set enhances sensitivity to abnormal operating conditions, such as quickly identifying sudden voltage drops or abnormal temperature increases, thereby improving the accuracy of battery management.

[0105] Specifically, this method combines signal processing and time series analysis to transform static data into dynamic indicators. For example, during charging, if the voltage change rate of a battery cell suddenly drops from 0.04V / s to 0.01V / s, while the internal resistance fluctuates more, the system can promptly determine that there may be capacity decay. This construction of multi-dimensional feature vectors lays the foundation for a comprehensive assessment of battery status.

[0106] S103. The voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude are classified and trained using the support vector machine algorithm to obtain a short-circuit precursor analysis model.

[0107] like Figure 2 As shown, based on a dynamic feature set, a Support Vector Machine (SVM) algorithm is employed. Voltage change rate (dV / dt), temperature surge rate (dT / dt), and internal resistance fluctuation amplitude (ΔR) are used as input feature vectors. A kernel function (such as the RBF kernel) maps the nonlinearly separable high-dimensional features to a regenerative Hilbert space, and the classification hyperplane is optimized using the principle of structural risk minimization. During the training phase, historical data with known labels (normal / short-circuit precursor) are used, combined with grid search and cross-validation to determine the optimal penalty coefficient C and kernel parameter γ. This maximizes the classification margin while reducing the risk of overfitting, ultimately obtaining a short-circuit precursor analysis model capable of identifying voltage drops, abnormal temperature rises, and abrupt changes in internal resistance. This model achieves accurate early warning of battery faults by integrating multi-dimensional dynamic features with the small-sample classification advantages of SVM.

[0108] In practice, when constructing a dynamic feature set, the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude can be extracted from real-time battery operation data as input feature vectors. For example, in a charging scenario, suppose a battery's voltage rises from 3.6V to 3.8V in 4 seconds, its temperature rises from 31℃ to 33℃, and its internal resistance fluctuates slightly between 34mΩ and 36mΩ. Through time series segmentation, the voltage change rate can be initially estimated as 0.05V / s, the temperature rise rate as 0.5℃ / s, and the internal resistance fluctuation amplitude as 1mΩ. This feature extraction method provides multi-dimensional input for subsequent SVM classification.

[0109] In one possible implementation, SVM maps these nonlinear features to a high-dimensional space using the RBF kernel function. The core of the RBF kernel lies in measuring the distance difference between samples. For example, given two sets of feature vectors, one set representing 0.03V / s, 0.2℃ / s, and 1mΩ under normal conditions, and the other set representing short-circuit precursors (0.1V / s, 1℃ / s, and 3mΩ), the kernel function amplifies the difference between the two, facilitating the separation of the classification hyperplane. This mapping enhances the model's sensitivity to anomalous features.

[0110] It should be noted that the principle of minimizing structural risk is optimized in SVM by balancing the classification margin and the misclassification penalty.

[0111] Specifically, the penalty coefficient C controls the model's tolerance for misclassification. For example, if C is set to 1, the model tends to pursue a larger margin; if C increases to 10, it focuses more on reducing misclassification. On a battery dataset, assuming normal samples account for 80% and short-circuit precursor samples account for 20%, a higher C value can better focus on rare abnormal samples and improve early warning capabilities.

[0112] Preferably, grid search and cross-validation are used for parameter tuning. For example, in a training set containing 1000 sets of historical data, C is set to a range of 0.1 to 10, and γ is set to a range of 0.01 to 1. The accuracy of each parameter is evaluated using 5-fold cross-validation.

[0113] In one embodiment, when C is 2 and γ is 0.1, the model achieves a classification accuracy of 95% on the validation set, indicating that the parameter combination effectively captures the patterns between features.

[0114] Specifically, the training phase relies on known labeled data. For example, under normal conditions, the voltage change rate is mostly between 0.02V / s and 0.05V / s, and the temperature rise rate is less than 0.3℃ / s. However, short-circuit precursors may manifest as a sudden increase in the voltage change rate to 0.08V / s and a temperature rise rate exceeding 0.8℃ / s. Using these labels, the SVM learns to distinguish between normal and abnormal conditions. In one case, a battery's voltage change rate suddenly dropped from 0.04V / s to 0.01V / s at the end of charging, while the temperature rise rate reached 1.2℃ / s; the model successfully identified this as a short-circuit risk.

[0115] Understandably, SVM's advantage in small-sample classification is particularly prominent in battery fault early warning. For example, a model trained with only 200 sets of data can identify 90% of abnormal temperature rise samples in testing. This efficiency stems from SVM's reliance on support vectors, rather than the total amount of data.

[0116] In one embodiment, feature vectors are fused with information on voltage drops, abnormal temperature rises, and sudden changes in internal resistance. The model can issue an early warning 5 seconds before an anomaly occurs, giving the system time to intervene. For example, in monitoring a 10-cell battery pack, the temperature rise rate of the third cell rapidly increased from 0.2℃ / s to 1℃ / s, and the internal resistance fluctuation increased from 2mΩ to 5mΩ. The SVM model, through feature vector analysis, quickly classified this as a short-circuit precursor. This combination of multi-dimensional feature fusion and SVM ensures the accuracy and reliability of early fault identification.

[0117] S104. Based on the output of the short-circuit precursor analysis model, if the voltage change rate exceeds the preset threshold, the temperature rise rate exceeds the preset threshold, or the internal resistance fluctuation amplitude is abnormal, then it is determined that there is a short-circuit precursor in the current single cell, and the short-circuit risk assessment result is obtained.

[0118] The classification boundary is obtained from the short-circuit precursor analysis model to determine the separation threshold between abnormal and normal data. For this separation threshold, a sliding window technique is used to segment the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude, resulting in segmented feature sequences. Based on these segmented feature sequences, the statistical values ​​of features within each segment are calculated to obtain an abnormal feature subset. If the statistical value of a feature in a segment of the abnormal feature subset exceeds a preset threshold, logical judgment determines that segment of data is in an abnormal state, resulting in an abnormal state set. Using the abnormal state set, the K-means clustering algorithm is employed to classify the abnormal data, obtaining an abnormal category set. Based on the mapping relationship between the abnormal category set and the operating state, the corresponding state label for each abnormal category is determined, yielding the state classification result. Based on the state classification result, combined with the operating data of individual battery cells, the risk assessment result for short-circuit precursors is determined.

[0119] In practice, when obtaining the classification boundary from the short-circuit precursor analysis model, the separation threshold between abnormal and normal data can be determined by analyzing the distribution of support vectors during training. For example, in a battery monitoring scenario, assuming the voltage change rate under normal conditions is mostly between 0.02V / s and 0.05V / s, while under abnormal conditions it may exceed 0.08V / s, the separation threshold can be initially set at 0.06V / s. This threshold setting is based on the distribution characteristics of historical data to ensure discriminative power.

[0120] In one possible implementation, a sliding window technique is used to segment real-time data streams. For example, with a window length of 2 seconds and a step size of 1 second, the voltage change rate, temperature rise rate, and internal resistance fluctuation amplitude can be segmented. Suppose a data record shows that the voltage rises from 3.7V to 3.75V in 2 seconds, the temperature rises from 32℃ to 33℃, and the internal resistance changes from 35mΩ to 36mΩ; then the characteristics of this segment are 0.025V / s, 0.5℃ / s, and 1mΩ, respectively. This segmentation method facilitates the capture of dynamic changes.

[0121] It should be noted that the statistical calculation of segmented feature sequences can include the mean, maximum value, and variance. Specifically, within a 10-second monitoring window, if the maximum value of a certain temperature rise rate reaches 1℃ / s, exceeding the normal range of 0.3℃ / s, it is marked as an anomalous feature subset. This statistical analysis helps to screen key anomalies.

[0122] Preferably, logical judgments are used to determine abnormal states. For example, if a certain segment of characteristic statistics has a voltage change rate of 0.07V / s, a temperature rise rate exceeding 0.8℃ / s, and an internal resistance fluctuation of 3mΩ, exceeding preset thresholds of 0.06V / s, 0.5℃ / s, and 2mΩ, then that segment is classified into the abnormal state set. This multi-condition judgment improves the reliability of anomaly identification.

[0123] In one embodiment, the K-means clustering algorithm classifies the set of anomalous states. Assume the anomalous data falls into two categories: one dominated by voltage spikes, and the other by temperature anomalies. For example, one anomalous data point might have values ​​of 0.1 V / s, 0.2 °C / s, and 1 mΩ, while another data point has values ​​of 0.03 V / s, 1.2 °C / s, and 2 mΩ. Clustering can clearly distinguish the anomalous patterns. This classification helps to refine the anomalous characteristics.

[0124] Understandably, the mapping relationship between anomaly categories and operating states needs to be predefined. For example, voltage surges might correspond to overvoltage risk during charging, while abnormal temperature rises might indicate a short circuit precursor. In one case, a battery's temperature surge rate reached 1.5℃ / s, which was classified as a short circuit risk and mapped to a high-risk state label. This mapping provides a basis for subsequent decision-making.

[0125] Specifically, when conducting risk assessments by combining individual battery operating data, the frequency and duration of anomalies can be analyzed. For example, if a battery experiences three temperature surges exceeding 1°C / s within 5 minutes, each lasting more than 3 seconds, the risk assessment result may be "high risk." This comprehensive analysis enhances the practicality of the early warning system.

[0126] In one embodiment, the effectiveness of the risk assessment is verified from multiple perspectives. For example, by combining historical data, if a battery previously failed due to internal resistance fluctuations exceeding 4mΩ, and a current monitoring shows an internal resistance of 5mΩ, it is inferred that the risk is consistent. Furthermore, by comparing with other battery packs, if a subset of abnormal characteristics is unique within the pack, its abnormality is further confirmed. This multi-dimensional verification ensures the rigor of the assessment.

[0127] S105. After obtaining the short-circuit risk assessment results, the location of individual cells related to short-circuit precursors is determined through a pre-established battery pack topology diagram, and the target battery identifier that needs to be isolated is obtained.

[0128] By establishing a pre-defined battery pack topology diagram, the locations of individual cells corresponding to short-circuit risks are obtained, and target cell identifiers are identified. For each target cell identifier, a preset threshold is used to determine isolation requirements, resulting in a set of cells requiring isolation. From this set, the adjacency relationships of individual cells within the topology are obtained to determine the affected battery pack range. Based on the affected battery pack range, the propagation path of precursor associations is analyzed using the topology diagram to obtain the risk diffusion trend. If a cell location in the propagation path exceeds a preset threshold, a high-risk cell subset is identified through logical judgment. From this high-risk cell subset, the risk assessment results are classified using a K-means clustering algorithm to obtain abnormal cell categories. Based on these abnormal cell categories, the isolation requirements are updated using the topology diagram to determine the final set of target cell identifiers.

[0129] Feasible and feasible, obtaining the location of individual cells corresponding to short-circuit risks through a pre-established battery pack topology diagram is the foundation of risk analysis. For example, in a battery pack consisting of 20 cells connected in series, assuming cell number 5 is marked as a short-circuit risk point, its position in the series chain can be quickly located using the topology diagram, and the target cell is identified as "cell 5". This location method relies on the topology diagram clearly marking the connection relationship of each cell, facilitating subsequent analysis. Based on the target cell identifier, a preset threshold is used to determine the isolation requirements and identify the set of cells that need to be isolated.

[0130] Specifically, assuming the preset thresholds are a voltage change rate of 0.06V / s and a temperature rise rate of 0.5℃ / s, if the real-time data for "Battery 5" are 0.08V / s and 0.7℃ / s, exceeding the thresholds, it is included in the set requiring isolation. This threshold determination uses quantitative indicators to quickly filter out abnormal batteries, ensuring timely isolation. From the set of batteries requiring isolation, the adjacency relationships of individual cells in the topology are obtained to determine the range of affected battery packs.

[0131] For example, if "Battery 5" is directly connected in series with "Battery 4" and "Battery 6", the topology diagram shows that its anomaly may affect neighboring batteries through current or heat conduction. The affected area is initially defined as "Battery 4-6". This adjacency analysis helps to clarify the local impact boundary of the risk. Based on the affected battery pack range, the propagation path of precursor associations is analyzed through a structure diagram to obtain the risk diffusion trend.

[0132] In one possible implementation, if the temperature of "Battery 5" suddenly rises and propagates to "Battery 6," and the rate of temperature rise in "Battery 6" reaches 0.6℃ / s, exceeding the threshold of 0.5℃ / s, then the propagation path is "Battery 5 → Battery 6." This path analysis reveals the dynamic direction of risk expansion. Regarding the risk diffusion trend, if a segment of the battery location in the propagation path exceeds a preset threshold, a subset of high-risk batteries is determined through logical judgment.

[0133] Preferably, if "Battery 6" exhibits an internal resistance fluctuation of 3mΩ in addition to temperature anomalies, exceeding the threshold of 2mΩ, it is grouped into the high-risk subset along with "Battery 5". This multi-condition judgment improves the accuracy of screening. From the high-risk battery subset, the risk assessment results are classified using the K-means clustering algorithm to obtain abnormal battery categories. For example, "Battery 5" is mainly characterized by voltage mutations, with features of 0.08V / s and 0.7℃ / s, while "Battery 6" is mainly characterized by temperature rises, with features of 0.03V / s and 0.6℃ / s. After clustering, they are divided into "Voltage Anomaly Category" and "Temperature Rise Anomaly Category". This classification refines the abnormal features, facilitating targeted processing. Based on the abnormal battery categories, the isolation requirements are updated through the topology diagram to determine the final target battery identifier set.

[0134] Understandably, if "voltage anomaly" maps to overvoltage risk and "temperature rise anomaly" points to short-circuit precursors, then both "Battery 5" and "Battery 6" need to be isolated, ultimately resulting in a set of "Battery 5, Battery 6". This update combines classification results with topological relationships to ensure the comprehensiveness of the isolation strategy.

[0135] It's important to note that the use of topology diagrams is consistent throughout, progressing from localization to diffusion analysis and then to isolation decisions. This approach, through structured analysis, expands risk from a single point to the whole, enhancing the reliability of battery pack management.

[0136] In one embodiment, if the adjacent battery does not exceed the threshold, it can be used as an extension scheme to isolate only the core abnormal battery, balancing efficiency and safety.

[0137] S106. Based on the target battery identifier, an isolation command is generated using a relay control algorithm. By disconnecting the connection path of the corresponding relay, the short-circuited battery is quickly isolated, and the updated battery pack status is obtained.

[0138] By identifying the target battery, the topology data of the corresponding relay path is obtained to determine the priority order of disconnection operations. Based on the priority order, a relay control algorithm is used to generate an isolation command sequence, resulting in a set of commands to be executed. For this command set, a preset threshold is used to determine the disconnection conditions of the relay path; if the load on a path exceeds the threshold, a disconnection signal is generated. From the disconnection signal, the range of affected connection paths is obtained to determine the isolation status of the short-circuited battery. Based on the isolation status, a K-means clustering algorithm is used to classify the battery pack status, obtaining the distribution characteristics of abnormal batteries. Using these distribution characteristics, the topology record of the battery pack status is updated to determine the isolation completion rate of the target battery. From the isolation completion rate, the updated battery pack operating parameters are obtained to determine the system's stable state.

[0139] It is feasible to obtain the topology data of the corresponding relay path through the target battery identifier to determine the priority order of disconnection operations. Understandably, this process depends on the distribution and connection logic of the relays within the battery pack.

[0140] For example, in a battery pack consisting of 10 individual cells connected in series, assuming "Battery 3" is marked as the target battery, its corresponding relay path includes a main circuit relay R1 and a branch relay R2. Topology data shows that R1 controls the entire series chain, while R2 is only responsible for connecting "Battery 3" to its neighboring batteries. In terms of priority, R2 is superior to R1 because partial disconnection can isolate anomalies more quickly while reducing the impact on the system. Based on the priority order, a relay control algorithm is used to generate an isolation instruction sequence, resulting in the set of instructions to be executed.

[0141] Specifically, the algorithm sorts the paths based on importance and response speed. In one embodiment, if R2 requires 0.1 seconds to disconnect and R1 requires 0.3 seconds, the instruction sequence is "R2 first, then R1," generating a set such as "R2 disconnected, R1 on standby." This ensures efficient isolation. For the instruction set, a preset threshold is used to determine the disconnection condition of the relay path. For example, assuming the load current threshold is 50A, if the real-time current of the R2 path reaches 60A, exceeding the threshold, a "R2 disconnected" signal is generated. This quantitative judgment improves the accuracy of the operation.

[0142] The affected connection path range is obtained from the disconnection signal to determine the short-circuit battery isolation status. Preferably, if the connection between "Battery 3" and "Battery 4" is interrupted after R2 is disconnected, and the current data returns to zero, the isolation status is "isolated". This helps to quickly verify the effect.

[0143] Based on the isolation status, the K-means clustering algorithm is used to classify the battery pack status and obtain the distribution characteristics of abnormal batteries. In one possible implementation, clustering is based on voltage and temperature data. If "Battery 3" has a voltage of 3.2V and a temperature of 50℃, while normal batteries are around 3.8V and 30℃, the classification result shows that "Battery 3" is an isolated outlier. This classification clearly reveals the abnormal distribution.

[0144] The topology record of the battery pack status is updated by analyzing distribution characteristics to determine the isolation completion rate of the target battery. It should be noted that the topology record will indicate that "Battery 3" is disconnected. If the data of its neighboring batteries is stable, the completion rate can be set to 100%. This provides a basis for subsequent management.

[0145] Updated battery pack operating parameters are obtained from the isolation completion status to determine the system's stable state. For example, if the total voltage recovers from 36V to 35V and the average temperature drops to 32°C, the system state can be determined as "stable." This parameter analysis ensures the overall operational safety.

[0146] In one embodiment, if the parameters of neighboring batteries fluctuate slightly but do not exceed the threshold, only the core abnormal battery can be isolated as an extended solution, which is both efficient and flexible. This multi-level analysis, from the local to the global, ensures the reliability of the battery pack.

[0147] S107. Extract the voltage, temperature, and internal resistance data of the remaining individual cells from the updated battery pack status, adjust the charging current distribution through the equalization management algorithm, and obtain the balanced battery operating parameters.

[0148] The system obtains voltage, temperature, and internal resistance data of individual cells from the battery pack status. An equalization management algorithm is used to calculate the adjustment value of the charging current, resulting in a preliminary current allocation scheme. For this preliminary scheme, a preset threshold is used to determine if the voltage data of individual cells exceeds the range. If so, the charging current is adjusted to obtain a corrected current allocation result. Based on the corrected current allocation result, the temperature data variation trend of individual cells is obtained, and statistical methods are used to analyze the distribution characteristics of the temperature data to determine the temperature equilibrium state. Using the temperature equilibrium state, the fluctuation range of the internal resistance data is obtained to determine if the internal resistance data meets the requirements of equalization management, resulting in the adjusted operating parameters. From the adjusted operating parameters, the actual allocation value of the charging current is extracted, and the equalization management algorithm is used to optimize the current allocation, resulting in the balanced battery operating parameters.

[0149] It is feasible to obtain individual cell voltage, temperature, and internal resistance data from the battery pack status. Understandably, this data forms the basis for equalization management.

[0150] For example, in a battery pack consisting of 8 individual cells connected in series, assuming voltage data are 3.7V, 3.8V, 3.6V, etc., temperature data are 28℃, 30℃, 32℃, etc., and internal resistance data are 20mΩ, 22mΩ, 19mΩ, etc., when calculating the charging current adjustment value using an equalization management algorithm.

[0151] Specifically, the amount of current that needs to be increased or decreased can be determined by comparing the voltage differences of each battery. For example, if the target average voltage is 3.75V, then the batteries with a voltage of 3.6V need to have their charging current increased. The initial allocation plan might be "increase the charging current of battery 1 by 0.2A and decrease the charging current of battery 2 by 0.1A".

[0152] For the initial current distribution scheme, a preset threshold is used to determine if the voltage exceeds the range. In one possible implementation, assume the safe voltage range is 3.5V to 4.0V. If a battery voltage reaches 4.1V, its charging current is adjusted to decrease by 0.3A. The corrected distribution might be "Battery 1 at 0.5A, Battery 2 at 0.4A". This adjustment ensures the voltage remains within the safe range.

[0153] Based on the corrected current distribution, the temperature data change trend is obtained. Preferably, the temperature change within 10 minutes after adjustment can be observed. For example, the temperature of battery 1 rises from 28°C to 29°C, while the temperature of battery 2 drops from 30°C to 29.5°C.

[0154] Statistical methods were used to analyze the distribution characteristics. Specifically, the calculated temperature mean of 29.5℃ and standard deviation of 0.5℃ indicate a good temperature equilibrium. This analysis helps to determine the effectiveness of thermal management.

[0155] The fluctuation range of internal resistance data is obtained by measuring the temperature equilibrium state. It should be noted that internal resistance is significantly affected by temperature. In one embodiment, if the internal resistance fluctuation range decreases from 19mΩ-22mΩ to 20mΩ-21mΩ after temperature stabilization, it is considered to meet the equalization management requirements. The adjusted operating parameters might be "average internal resistance 20.5mΩ, fluctuation ±0.5mΩ". This stability improves battery consistency.

[0156] The actual charging current distribution value is extracted from the operating parameters after internal resistance adjustment. For example, the actual current of battery 1 is 0.45A, and that of battery 2 is 0.42A.

[0157] An optimization algorithm based on balanced management is employed. In one possible implementation, the operating parameters are adjusted comprehensively based on internal resistance and voltage to obtain balanced operating parameters, such as "Battery 1 current 0.43A, voltage 3.75V, temperature 29℃". This optimization improves overall efficiency. For example, when considering multiple aspects, if a battery's temperature suddenly rises to 35℃ while its voltage remains normal, its current can be adjusted to 0.3A, and the internal resistance observed. If the internal resistance returns to normal, this is the core solution; if it remains too high, the approach can be extended to reduce the current of adjacent batteries to share the heat load. This multi-level method is both precise and flexible, ensuring the long-term reliability of the battery pack.

[0158] S108. Based on the balanced battery operating parameters, if the voltage difference is less than the preset threshold and the temperature and internal resistance are within the normal range, then the system stability is determined to be restored, and the final battery pack operating state is obtained.

[0159] like Figure 3 As shown, voltage difference data of individual cells is obtained from the balanced operating parameters. A preset threshold is used to determine whether the voltage difference meets the required range, yielding a preliminary stability assessment result. Based on the preliminary stability assessment result, the distribution of temperature data is obtained, and statistical methods are used to analyze the temperature data variation characteristics to determine the balanced state of the temperature distribution. Using the balanced temperature distribution, the distribution of internal resistance data is obtained, and it is determined whether the internal resistance data is within a preset range, providing a basis for adjusting the internal resistance distribution. Based on the adjustment basis for the internal resistance distribution, real-time data of individual cells is extracted from the operating parameters, and a balancing algorithm is used to optimize the data distribution, resulting in an optimized parameter set. Based on the optimized parameter set, the overall trend of the battery pack state is obtained, and logical judgment is used to determine whether system stability has recovered, yielding a stability confirmation result. Operating conclusions are extracted from the stability confirmation result, and the recorded data of the battery pack state is updated to determine the balanced state of system operation. Using the recorded data under the balanced state, the long-term operating trend of individual cells is obtained, and it is determined whether the battery pack state remains stable, resulting in the final operating parameter adjustment scheme.

[0160] It is feasible to obtain the voltage difference data of individual cells from the balanced operating parameters. Understandably, voltage differences reflect the consistency of the individual cells within the battery pack. For example, in a battery pack with eight cells connected in series, the balanced voltage might be 3.74V, 3.76V, 3.73V, etc., with a difference ranging from 0.01V to 0.03V.

[0161] Preferably, a preset threshold, such as 0.05V, is used to determine the stability. If the difference is less than this value, the stability is initially considered to be good. This assessment lays the foundation for subsequent analysis.

[0162] The distribution of temperature data is obtained based on the preliminary stability assessment results. Specifically, the temperatures of individual batteries, such as 28.5℃, 29℃, and 28.8℃, can be observed. In one possible implementation, a statistical method is used to calculate the mean of 28.8℃ and the maximum deviation of 0.2℃, indicating a uniform temperature distribution. This uniformity helps reduce the risk of localized overheating.

[0163] When obtaining the distribution of internal resistance data by assessing the equilibrium state of temperature distribution, it's important to note that internal resistance is closely related to temperature. For example, if the temperature is stable, the internal resistance data might be 20.2 mΩ, 20.5 mΩ, or 20.1 mΩ, with a fluctuation range of only 0.4 mΩ. The system determines whether this is within a preset range, such as ±0.5 mΩ. If it is, the internal resistance distribution is reasonable, providing a reliable basis for adjustment.

[0164] The adjustment of internal resistance distribution is based on the extraction of real-time data. For example, a battery has a voltage of 3.73V, a temperature of 29℃, and an internal resistance of 20.3mΩ.

[0165] When using a balanced algorithm for optimization, in one embodiment, the current can be finely adjusted to 0.4A based on the high internal resistance, resulting in optimized parameters such as a voltage of 3.75V and an internal resistance of 20.2mΩ. This optimization improves data consistency.

[0166] When obtaining the overall trend of battery pack status changes based on the optimized parameter set, preferably, the average voltage increases from 3.74V to 3.75V, and the temperature fluctuation decreases to 0.1℃. Through logical judgment, if the trend stabilizes, system stability is restored. This confirmation ensures operational reliability.

[0167] When updating recorded data by extracting operational conclusions from stability confirmation results, for example, if the records show that the voltage difference has remained within 0.02V for a long period and the average temperature has remained stable at 29°C, this update provides data support for long-term monitoring.

[0168] When obtaining long-term operating trends from recorded data under equilibrium conditions, in one possible implementation, a battery with a voltage of 3.75V and an internal resistance of 20.2mΩ remains stable for 10 hours without significant fluctuations, indicating a continuously stable state.

[0169] For example, if a battery's temperature suddenly rises to 31°C, its current can be reduced to 0.35A to observe the trend; if the internal resistance subsequently decreases to 20.1mΩ, this is the core solution. If it is still too high, the current of a nearby battery can be adjusted to 0.38A to share the load. This multi-level adjustment ensures long-term stability. For instance, when considering multiple aspects, assuming a battery's voltage drops to 3.70V while the temperature remains normal, the current can be increased to 0.45A to observe changes in internal resistance. If the internal resistance remains stable, the solution is effective; if fluctuations increase, it can be adjusted to 0.42A in conjunction with temperature data analysis. This multi-directional verification enhances the flexibility and accuracy of the solution.

[0170] Example 2

[0171] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0172] Example 3

[0173] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0174] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A series type lithium battery pack power equalization management method, characterized by, The method comprises the following steps: Real-time acquisition of multi-dimensional monitoring information of the series lithium battery pack, including voltage, temperature, and internal resistance data of each single battery, to construct an original data set; Based on the original data set, a time series analysis algorithm is used to calculate the voltage change rate, temperature sudden rise rate, and internal resistance fluctuation amplitude to obtain a dynamic characteristic set of the battery operation; A support vector machine algorithm is used for classification training of the dynamic characteristic set of the battery operation to obtain a short-circuit precursor analysis model; Based on the output result of the short-circuit precursor analysis model, when the voltage change rate, temperature sudden rise rate, and internal resistance fluctuation amplitude exceed the corresponding preset threshold values, it is determined that the current single battery has a short-circuit precursor, and a short-circuit risk assessment result is obtained; The pre-established battery pack topology structure diagram is matched with the short-circuit risk assessment result to determine the position of the single battery related to the short-circuit precursor, and a target battery identifier needing to be isolated is obtained; According to the target battery identifier, a relay control algorithm is used to generate an isolation instruction, the connection path of the corresponding relay is disconnected, the isolation of the short-circuit battery is completed, and an updated battery pack state is obtained; The voltage, temperature, and internal resistance data of the remaining single batteries are extracted from the updated battery pack state, a charge current distribution is adjusted through an equalization management algorithm, and equalized battery operation parameters are obtained; According to the equalized battery operation parameters, if the voltage, temperature, and internal resistance data of the remaining single batteries are all within the normal range, it is determined that the system stability is restored, and a final battery pack operation state is obtained; The output result based on the short-circuit precursor analysis model, when the voltage change rate, temperature sudden rise rate, and internal resistance fluctuation amplitude exceed the corresponding preset threshold values, it is determined that the current single battery has a short-circuit precursor, and a short-circuit risk assessment result is obtained, comprising: The classification boundary is obtained from the short-circuit precursor analysis model to determine the separation threshold of abnormal data and normal data; For the separation threshold, a sliding window is used to segment the voltage change rate, temperature sudden rise rate, and internal resistance fluctuation amplitude to obtain a segmented feature sequence; According to the segmented feature sequence, the statistical values of the features in each segment are calculated to obtain an abnormal feature subset; If the statistical values of the feature segments in the abnormal feature subset exceed the preset threshold values, it is determined through logical judgment that the state is abnormal to obtain an abnormal state set; Through the abnormal state set, a K-means clustering algorithm is used to classify the abnormal data to obtain an abnormal category set; According to the mapping relationship between the abnormal category set and the operation state, the state labels corresponding to each abnormal category are determined to obtain a state classification result; For the state classification result, the risk assessment result of the short-circuit precursor is determined in combination with the operation data of the single battery.

2. The method of claim 1, wherein, The real-time acquisition of multi-dimensional monitoring information of the series lithium battery pack, including voltage, temperature, and internal resistance data of each single battery, to construct an original data set, comprises: By installing corresponding sensors on each single battery of the series lithium battery pack, the voltage, temperature, and internal resistance data of each single battery are acquired in real time; The voltage, temperature, and internal resistance data of each single battery are filtered, amplified, and analog-digital converted to obtain multi-dimensional data. The multi-dimensional data is time-series aligned and outlier removed based on a battery equivalent circuit model to form an original data set containing a three-dimensional coupling relationship of voltage-temperature-internal resistance.

3. The method of claim 1, wherein, The original data set is used to calculate the voltage change rate, temperature sudden rise rate and internal resistance fluctuation amplitude by using a time series analysis algorithm to obtain a dynamic feature set of the battery operation, including: Based on the original data set, the voltage change rate is calculated by first-order difference, the temperature sudden rise rate is obtained by deriving the smoothed temperature curve after Savitzky-Golay filtering, and the internal resistance fluctuation amplitude is quantified by sliding standard deviation; At the same time, the dynamic time warping algorithm is introduced to align the feature time series difference of different battery monomers, and finally the dynamic feature set of the battery operation is obtained.

4. The method of claim 1, wherein, The pre-established battery pack topology structure graph is matched with the short circuit risk assessment result to determine the monomer battery position related to the short circuit precursor, and the target battery identification needing isolation is obtained, including: The monomer battery position corresponding to the short circuit risk is obtained through the pre-established battery pack topology structure graph, and the initial target battery identification is determined; For the initial target battery identification, a preset threshold is used to judge the isolation requirement to obtain a battery set that needs to be isolated; From the battery set that needs to be isolated, the adjacency relationship of the monomer battery in the topology structure is obtained, and the affected battery pack range is determined; According to the affected battery pack range, the propagation path of the short circuit precursor is analyzed through the structure graph to obtain the risk diffusion trend; For the risk diffusion trend, if there is a battery segment position exceeding the preset threshold in the propagation path, a high-risk battery subset is determined through logical judgment; From the high-risk battery subset, the K-means clustering algorithm is used to classify the risk judgment result to obtain the abnormal battery category; According to the abnormal battery category, the isolation requirement is updated through the topology structure graph to determine the final target battery identification set.

5. The method of claim 1, wherein, According to the target battery identification, a relay control algorithm is used to generate an isolation instruction to disconnect the connection path of the corresponding relay, complete the isolation of the short circuit battery, and obtain the updated battery pack state, including: Through the target battery identification, the topology data of the corresponding relay path is obtained to determine the priority order of the disconnection operation; According to the priority order, a relay control algorithm is used to generate an isolation instruction sequence to obtain an instruction set to be executed; For the instruction set, the disconnection condition of the relay path is judged by a preset threshold, and if the relay path load exceeds the threshold, a disconnection signal is generated; From the disconnection signal, the affected connection path range is obtained to determine the isolation state of the short circuit battery; According to the isolation state, the K-means clustering algorithm is used to classify the battery pack state to obtain the distribution characteristics of the abnormal battery; Through the distribution characteristics, the topology record of the battery pack state is updated to judge the isolation completion degree of the target battery; From the isolation completion degree, the updated battery pack operation parameters are obtained, and then the updated battery pack state is obtained.

6. The method of claim 1, wherein, The voltage, temperature and internal resistance data of the remaining monomer batteries are extracted from the updated battery pack state, and the charging current distribution is adjusted by an equalization management algorithm to obtain the equalized battery operation parameters, including: The voltage data, temperature data and internal resistance data of the single battery are obtained from the battery pack state, the adjustment value of the charging current is calculated by using the equalization management algorithm, and a preliminary current distribution scheme is obtained; For the preliminary current distribution scheme, the voltage data of the single battery is judged whether it is out of range by a preset threshold, if it is out of range, the charging current is adjusted, and a corrected current distribution result is obtained; According to the corrected current distribution result, the temperature data change trend of the single battery is obtained, the distribution characteristics of the temperature data are analyzed by using statistical method, and the temperature equalization state is determined; Through the temperature equalization state, the fluctuation range of the internal resistance data is obtained, whether the internal resistance data meets the requirements of equalization management is judged, and the running parameter after internal resistance adjustment is obtained; From the running parameter after internal resistance adjustment, the actual distribution value of the charging current is extracted, the current distribution is optimized by using the equalization management algorithm, and the equalized battery running parameter is obtained.

7. The method of claim 1, wherein, According to the equalized battery running parameter, if the voltage, temperature and internal resistance data of the remaining single battery are all within the normal range, the system stability recovery is determined, and the final battery pack running state is obtained, including: The voltage difference data of the single battery is obtained from the equalized battery running parameter, whether the voltage difference meets the range requirement is judged by a preset threshold, and a preliminary stability evaluation result is obtained; According to the preliminary stability evaluation result, the distribution of the temperature data is obtained, the change characteristics of the temperature data are analyzed by using statistical method, and the equalization state of the temperature distribution is determined; Through the equalization state of the temperature distribution, the distribution of the internal resistance data is obtained, whether the internal resistance data is within the preset range is judged, and the adjustment basis of the internal resistance distribution is obtained; For the adjustment basis of the internal resistance distribution, the real-time data of the single battery is extracted from the running parameter, and an optimized parameter set is obtained; According to the optimized parameter set, the overall change trend of the battery pack state is obtained, whether the system stability recovers is determined by logical judgment, and a stability confirmation result is obtained; The running conclusion is extracted from the stability confirmation result, the record data of the battery pack state is updated, and the equalization state of the system running is determined; Through the record data under the equalization state, the long-term running trend of the single battery is obtained, whether the battery pack state is continuously stable is judged, and the final battery pack running state is obtained.

8. A computer apparatus comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, causes the processor to perform the method of any one of claims 1 to 7. The processor executes the computer program to realize the steps of the method of any one of claims 1-7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1-7.

Images