A method and apparatus for managing a battery

Abstract

The application discloses a kind of management method and device of battery, it is related to battery management field, including the feature extraction of battery monitoring data, and input to pre-trained neural network, obtain the state of charge, health state and remaining useful life of battery;According to the state of charge and health state, determine the equalization strategy of battery;According to health state, determine early warning threshold, early warning threshold is used to assess battery risk;According to monitoring data, remaining useful life and early warning threshold, determine the risk level of battery;According to risk level and equalization strategy, battery is managed.Through obtaining electrical, thermal, mechanical three kinds of monitoring data, dimension is more, determination result is more accurate.Early warning threshold can be adjusted with health state, equalization strategy not only considers state of charge, but also considers the influence of health state on equalization effect, finally the management of battery is more reasonable and accurate, improve the life of battery.

Description

Technical Field

[0001] This invention relates to the field of battery management, and in particular to a battery management method and apparatus. Background Technology

[0002] In related technologies, lithium battery management system lifespan prediction and risk warning schemes often pre-set fixed warning thresholds. When these thresholds are reached, a warning is issued regarding the battery's health status. However, as the battery ages through cycles, the normal ranges of parameters such as temperature and internal resistance drift. Because the warning thresholds are completely fixed, they cannot adapt to changes in the failure characteristics of the battery after aging. Furthermore, fixed thresholds may be too sensitive in the early stages of the battery's lifespan and too insensitive in later stages, leading to a high probability of false alarms and missed alarms. At the same time, the balancing strategy is also fixed and cannot adjust the state-of-charge control range and charge / discharge rate limits according to the degree of battery aging. Summary of the Invention

[0003] The purpose of this invention is to provide a battery management method and apparatus that acquires three types of monitoring data—electrical, thermal, and mechanical—providing more dimensions and more accurate judgment results. The warning threshold can be adjusted according to the battery's health status, and the balancing strategy considers not only the state of charge but also the impact of the battery's health status on the balancing effect. Ultimately, this results in more rational and accurate battery management, improving battery life.

[0004] To address the aforementioned technical problems, this invention provides a battery management method, comprising:

[0005] Acquire battery monitoring data, including the battery's electrical parameters, thermal parameters, and mechanical parameters;

[0006] Feature extraction is performed on the monitoring data, and the feature-extracted monitoring data is input into a pre-trained neural network to obtain the battery's state of charge, state of health, and remaining lifespan.

[0007] A battery balancing strategy is determined based on the state of charge and the state of health.

[0008] A warning threshold is determined based on the health status, and the warning threshold is used to assess battery risk.

[0009] The risk level of the battery is determined based on the monitoring data, the remaining service life, and the warning threshold.

[0010] The battery is managed according to the risk level and the balancing strategy.

[0011] On the other hand, acquiring battery monitoring data includes:

[0012] The voltage of a single battery, the total voltage of the series circuit of the battery, and the operating current in the charging and discharging circuit of the battery are obtained as the electrical parameters.

[0013] Temperature values ​​collected at multiple preset locations on the battery are obtained as the thermal parameters;

[0014] The equivalent internal resistance, the electrode deformation between the internal electrode and the external casing, and the module preload applied to the battery by the module containing the battery are obtained as the mechanical parameters, which are determined by applying a pulse excitation to the battery and collecting the response signal of the battery.

[0015] On the other hand, after acquiring the battery monitoring data, it also includes:

[0016] Determine the rate of change of each parameter in the monitoring data;

[0017] The battery is determined to be in a steady-state or transient state based on the rate of change.

[0018] When it is determined that the battery is in a steady-state condition, the sampling frequency is updated to the first frequency;

[0019] When it is determined that the battery is in a transient operating condition, the sampling frequency is updated to a second frequency, which is higher than the first frequency;

[0020] Monitoring data of the battery is acquired based on the updated sampling frequency.

[0021] On the other hand, after acquiring the battery monitoring data, it also includes:

[0022] The monitoring data is then filtered and denoised.

[0023] Amplitude calibration is performed on the filtered and denoised monitoring data;

[0024] Time-series alignment is performed on the amplitude-calibrated monitoring data;

[0025] The time-aligned monitoring data is then normalized.

[0026] Feature extraction is performed on the monitoring data, and the feature-extracted monitoring data is input into a pre-trained neural network to obtain the battery's state of charge, state of health, and remaining lifespan, including:

[0027] The normalized monitoring data is decomposed to filter out signal components associated with environmental interference and hardware errors, and retain signal components associated with battery aging, operating condition fluctuations and deformation stress as aging characteristic data.

[0028] Using the aging characteristic data as input, and under the constraint of satisfying the physical laws of the battery aging process, the state of charge, health status, and remaining service life output by the neural network are obtained.

[0029] The neural network is trained based on historical monitoring data and corresponding state of charge, health status, and remaining service life.

[0030] On the other hand, determining a battery balancing strategy based on the state of charge and the state of health includes:

[0031] When the health status is greater than or equal to the first health threshold, the charge state control interval is set as the first charge interval;

[0032] When the health status is greater than or equal to the second health threshold and less than the first health threshold, the charge state control interval is set as the second charge interval, which is narrower than the first charge interval.

[0033] When the health status is less than the second health threshold, the state of charge control range is set to a third charge range, which is narrower than the second charge range, and the maximum charge / discharge rate of the battery is limited.

[0034] The charging and discharging management strategy corresponding to the state of charge control interval is used as the balancing strategy.

[0035] On the other hand, determining the warning threshold based on the health status includes:

[0036] The aging stage of the battery is determined based on its health status;

[0037] The initial temperature warning threshold is corrected according to the aging stage, and the corrected initial temperature warning threshold is used as the temperature warning threshold.

[0038] The voltage warning threshold and internal resistance warning threshold are determined based on the aging stage.

[0039] On the other hand, determining the risk level of the battery based on the monitoring data, the remaining service life, and the warning threshold includes:

[0040] Based on the currently collected temperatures, determine the highest temperature of a single cell and the maximum temperature difference between cells in the series circuit of the battery.

[0041] When the highest temperature is greater than the temperature threshold or the maximum temperature difference is greater than the temperature difference threshold, it is determined to be the first risk level.

[0042] On the other hand, determining the risk level of the battery based on the monitoring data, the remaining service life, and the warning threshold includes:

[0043] The hidden state of the prediction model is updated, and the relationship for the hidden state update is:

[0044] ;

[0045] in, Let be the hidden state at time t, where the hidden state includes the battery aging characteristics and temperature change trends memorized by the prediction model. For activation function, Let be the weighting parameters for temperature and equivalent internal resistance at time t. Let be the temperature and equivalent internal resistance at time t. The weight parameters of the hidden state at time t-1 are... The hidden state at time t-1 For bias terms;

[0046] Temperature and equivalent internal resistance are input into the prediction model to obtain the prediction result output by the prediction model. The expression of the prediction result is as follows:

[0047] ;

[0048] in, For the output of the prediction model Time to The prediction results at time, including predicted temperature and predicted equivalent internal resistance, Let be the temperature and equivalent internal resistance at time t. The temperature and equivalent internal resistance at time t-1 are given. Let be the temperature and equivalent internal resistance at time tN;

[0049] exist to If, at any given time, either the predicted value of temperature or the equivalent internal resistance exceeds the corresponding safety upper limit threshold, it is determined to be at the second risk level.

[0050] On the other hand, determining the risk level of the battery based on the monitoring data, the remaining service life, and the warning threshold includes:

[0051] Based on the monitoring data, the remaining service life, and the warning threshold, the scoring model is used to obtain a score for the battery's failure probability, which is output by the scoring model. The expression for the failure probability score is as follows:

[0052] ;

[0053] in, For the i-th set of monitoring data The score of the failure probability obtained under k decision trees Let F be the discrimination result of the m-th decision tree, and F be the function space of all decision trees. There are a total of k decision trees.

[0054] The scoring is constrained by the following relationship:

[0055] ;

[0056] in, This is the total error. For the actual faults in the i-th set of monitoring data With predicting failure deviation, For regularization terms, Used to control the number of decision trees Used for weight decay;

[0057] Under the conditions of satisfying the constraints, when the predicted fault score is higher than the high-risk threshold, the change in the equivalent internal resistance of the battery within a preset time period exceeds the mutation threshold, and the correspondence between the open-circuit voltage and the state of charge of the battery deviates from the benchmark correspondence and the deviation exceeds the distortion threshold, it is determined to be the third risk level.

[0058] To address the aforementioned technical problems, the present invention also provides a battery management device, comprising:

[0059] Memory, used to store computer programs;

[0060] A processor is used to implement the battery management method described above when executing the computer program.

[0061] This application provides a battery management method and apparatus, relating to the field of battery management. The method includes extracting features from battery monitoring data and inputting the extracted features into a pre-trained neural network to obtain the battery's state of charge (SOC), state of health (SOH), and remaining lifespan. It then determines a battery balancing strategy based on the SOC and SOH; determines a warning threshold based on the SOH, which is used to assess battery risk; determines the battery's risk level based on the monitoring data, remaining lifespan, and warning threshold; and manages the battery according to the risk level and balancing strategy. By acquiring electrical, thermal, and mechanical monitoring data, the method provides more dimensions and more accurate judgment results. The warning threshold can be adjusted according to the SOH, and the balancing strategy considers not only the SOC but also the SOH's influence on the balancing effect, ultimately leading to more reasonable and accurate battery management and improved battery lifespan. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the prior art and embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 A flowchart of a battery management method provided by the present invention;

[0064] Figure 2 This is a schematic diagram of a battery management device provided by the present invention. Detailed Implementation

[0065] The core of this invention is to provide a battery management method and apparatus that acquires three types of monitoring data—electrical, thermal, and mechanical—providing more dimensions and more accurate judgment results. The warning threshold can be adjusted according to the battery's health status, and the balancing strategy considers not only the state of charge but also the impact of the battery's health status on the balancing effect. Ultimately, this results in more reasonable and accurate battery management, improving battery life.

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Figure 1 A flowchart of a battery management method provided by the present invention, the battery management method comprising:

[0068] S11: Acquire battery monitoring data, including battery electrical parameters, thermal parameters and mechanical parameters;

[0069] Traditional battery management systems typically only collect two types of electrical parameters: voltage and current. However, the health status and safety hazards of a battery are not only reflected in its electrical characteristics but are also related to physical quantities such as temperature distribution, mechanical deformation, and internal stress. Relying solely on electrical parameters for condition assessment cannot detect the subtle mechanical changes and stress anomalies that occur in the early stages of thermal runaway, easily leading to the neglect of early risks.

[0070] Therefore, this application collected electrical, thermal, and mechanical parameters. Electrical parameters, including voltage and current, characterize the battery's energy storage level, charge / discharge rate, and external load characteristics. Thermal parameters, obtained through multi-point temperature measurements, characterize the thermal field distribution inside the battery and at its connections, reflecting the Joule heat and electrochemical reaction heat generated during charging and discharging. Mechanical parameters include equivalent internal resistance, electrode deformation, and module preload. Equivalent internal resistance reflects the battery's internal ionic and electronic conductivity and is a sensitive indicator of battery aging. Electrode deformation characterizes the volume expansion and contraction of the internal electrodes during repeated charging and discharging. Module preload reflects the external constraint force on the battery within the module, which dynamically changes with the battery's expansion and contraction.

[0071] S12: Extract features from the monitoring data and input the extracted monitoring data into a pre-trained neural network to obtain the battery's state of charge, health status, and remaining lifespan.

[0072] Raw monitoring data contains a large amount of noise and redundant information, such as measurement noise in voltage data, ripple components in current data, and ambient temperature fluctuations in temperature data. If these interference signals are directly used in neural network calculations, they will affect the accuracy of the prediction results. Therefore, feature extraction is necessary before inputting the monitoring data into the neural network to extract effective features closely related to battery state evolution from the raw data. This application eliminates noise and redundant components in the raw monitoring data through feature extraction, improving the input quality of the neural network and making the output state of charge, health status, and remaining service life more accurate and reliable. Simultaneously, the neural network possesses nonlinear mapping and generalization capabilities, enabling it to adapt to battery state prediction needs under different aging levels and operating conditions, overcoming the poor generalization of traditional physical models.

[0073] S13: Determine the battery balancing strategy based on the state of charge and state of health;

[0074] In a battery pack composed of multiple individual cells connected in series, differences in manufacturing consistency, uneven temperature during use, and varying aging rates can gradually lead to variations in the state of charge (SOC) of each individual cell. If left unchecked, these differences will accumulate and amplify, causing some cells to overcharge or over-discharge, accelerating battery aging, and potentially leading to safety accidents in severe cases. The purpose of battery equalization management is to adjust the energy distribution of each individual cell and reduce the differences in SOC between them.

[0075] Different types of batteries require different balancing strategies at different aging stages. New batteries or slightly aged batteries have similar self-discharge rates and internal resistance growth rates among their individual cells, and conventional balancing strategies can maintain consistency. However, for batteries with more severe aging, internal side reactions intensify, self-discharge rates increase, and the differences between individual cells increase at a faster rate, requiring more stringent balancing strategies, including narrowing the control range of state of charge and limiting the charge and discharge rates.

[0076] Specifically, batteries with higher health status values ​​correspond to a wider state of charge control range, allowing for normal charging and discharging operations within a wider charge range. Batteries with lower health status values ​​correspond to a narrower state of charge control range, limiting their operation to a safer charge window. Limiting the maximum charge and discharge rate can reduce the current rate during charging and discharging, thereby reducing heat generation and slowing down the aging rate.

[0077] S14: Determine the warning threshold based on the health status. The warning threshold is used to assess battery risk.

[0078] Throughout a battery's lifespan, parameter ranges drift with aging. For example, as the number of cycles increases, the battery's equivalent internal resistance gradually increases. This means that if the warning threshold based on internal resistance remains constant, it may be too lenient in the new battery stage, leading to missed risks, while in the aging stage, it may be too strict, resulting in frequent false alarms. Similarly, the thermal characteristics of an aging battery also change; its heat capacity and thermal conductivity change with internal structural variations, resulting in different temperature rise rates for the same amount of heat generated. Therefore, temperature-related warning thresholds also need to be dynamically adjusted. The state of health (SHS) is the ratio of the battery's current capacity to its factory rated capacity and is a core indicator for measuring the degree of battery aging. The lower the SHS value, the more severe the battery aging.

[0079] S15: Determine the risk level of the battery based on monitoring data, remaining service life, and warning thresholds;

[0080] By integrating comprehensive monitoring data, remaining lifespan predictions, and adaptively updated warning thresholds, the current safety risk level of the battery is determined. Risk assessment based on a single parameter is prone to misjudgment. For example, temperature data might temporarily exceed limits due to sensor noise, even if the battery is not actually facing a real risk. Conversely, some early-stage thermal runaway faults might not cause any single parameter to significantly exceed limits, but multiple parameters could lead to different outcomes.

[0081] The overall changes have shown an abnormal pattern.

[0082] The monitoring data provides the current actual operating parameter values, including voltage, current, temperature, equivalent internal resistance, electrode deformation, and module preload. The remaining service life reflects the battery's current ability to maintain its health. A short remaining service life means that the battery is about to reach the end of its life, and its safety margin is significantly reduced. The warning threshold provides a judgment boundary that is appropriate for the current aging state.

[0083] Specifically, this application classifies risk levels into three levels. The first risk level indicates that the currently measured parameters exceed the safety threshold, corresponding to a visible anomaly in the battery, such as excessive temperature or large temperature difference, which is an abnormal state that has already occurred. The second risk level indicates that the predicted future parameter trend will exceed the safety threshold, corresponding to the battery's developmental risk. This risk has not yet manifested as parameter exceeding limits at the current moment, but it is foreseeable that exceeding limits will occur based on the time-series trend. The third risk level indicates that the comprehensive fault characteristics meet the criteria for high-risk fault identification, corresponding to the battery's deeper safety risks. This level involves the joint identification of multiple characteristics.

[0084] S16: Manage the battery according to the risk level and balancing strategy.

[0085] Battery management includes thermal management, energy balancing, and short-circuit protection for the charge / discharge circuits. Thermal management operations are performed according to the risk level to maintain the battery temperature within a safe range. Thermal management includes two basic operating modes: heating and cooling. Heating is initiated under low-temperature conditions to raise the battery temperature to a suitable operating temperature, while cooling (e.g., forced air cooling or water cooling) is initiated under high-temperature conditions to control the battery temperature below the safe upper limit. Energy balancing operations are performed according to a balancing strategy to adjust the energy distribution among individual cells in the battery pack and reduce the difference in state of charge between cells. This can be achieved using active or passive balancing circuits. Short-circuit protection operations are performed according to the risk level to disconnect the battery from external circuits in emergencies. Short-circuit protection is the last line of defense when the risk level reaches the highest level. It physically disconnects the main charge / discharge circuit by controlling a DC circuit breaker or contactor, completely blocking the current path and preventing further deterioration of the fault.

[0086] This application provides a battery management method, relating to the field of battery management. The method includes extracting features from battery monitoring data and inputting the extracted features into a pre-trained neural network to obtain the battery's state of charge (SOC), state of health (SOH), and remaining lifespan. It then determines a battery balancing strategy based on the SOC and SOH; establishes a warning threshold based on the SOH, which is used to assess battery risk; determines the battery's risk level based on the monitoring data, remaining lifespan, and warning threshold; and manages the battery according to the risk level and balancing strategy. By acquiring electrical, thermal, and mechanical monitoring data, the method provides more dimensions and more accurate judgment results. The warning threshold can be adjusted according to the SOH, and the balancing strategy considers not only the SOC but also the SOH's influence on the balancing effect, ultimately leading to more reasonable and accurate battery management and improved battery lifespan.

[0087] Based on the above embodiments:

[0088] In some embodiments, acquiring battery monitoring data includes:

[0089] The voltage of a single battery, the total voltage of the battery's series circuit, and the operating current in the battery's charging and discharging circuit are obtained as electrical parameters.

[0090] Temperature values ​​collected at multiple preset locations on the battery are used as thermal parameters;

[0091] The equivalent internal resistance, the electrode deformation between the internal electrode and the external casing, and the module preload applied to the battery by the module are obtained as mechanical parameters, which are determined by applying pulse excitation to the battery and collecting the battery's response signal.

[0092] The battery voltage is acquired using an AFE analog front-end chip in conjunction with precision low-temperature drift voltage divider resistors and differential instrumentation amplifiers. A Kelvin four-wire sampling structure is employed to eliminate interference from lead and contact resistance. Based on the chip's multi-channel parallel sampling and high-voltage common-mode suppression characteristics, the individual cell voltage and cluster total voltage are acquired simultaneously, ensuring sampling accuracy and operational stability under high-voltage conditions. The battery current is acquired using a combination of a manganese-copper four-terminal shunt and a closed-loop Hall sensor. In the low-current range, the shunt achieves microampere-level sampling, while in the high-current range, Hall devices enable electrically isolated detection. Bidirectional acquisition of dynamic charge and discharge current timing information provides an accurate current reference for internal resistance calculation and state modeling. Multi-point temperature monitoring utilizes a distributed setup of PT100 platinum resistance thermometers and NTC thermistors, along with a dedicated MAX31865 temperature conditioning circuit. Temperature points are densely distributed in key areas such as the cell surface, tabs, busbars, and module gaps. After signal amplification, filtering, and standardization conversion, the monitoring of local temperature differences and the overall temperature field of the battery is achieved. The battery's equivalent internal resistance is dynamically detected using a controllable pulse excitation source in conjunction with a high-speed voltage and current synchronous sampling branch. The transient response of the battery is acquired through short-time pulse excitation, and the ohmic internal resistance is calculated in real time based on the DC voltage drop method. A high-frequency anti-interference conditioning circuit is then used to eliminate operating noise, accurately characterizing the degree of aging and degradation within the battery. Electrode deformation is detected using a non-contact / contact composite detection method combining a laser triangular displacement sensor and a foil strain gauge. This method collects data on the overall bulging expansion deformation of the cell and the micro-bending and interlayer strain changes of the electrode. The signals are normalized by a strain conditioning circuit, enabling dynamic sensing of micro-deformation throughout the charging and discharging process. Module preload is collected using a flexible thin-film pressure sensor and a flat-plate force sensor. Based on the strain bridge detection principle, the local pressure between cell layers and the overall clamping force distribution of the module are obtained. The signals are converted by a 24-bit high-precision strain ADC, enabling long-term monitoring of module assembly stress and dynamic clamping status.

[0093] Specifically, in battery management, voltage is the most fundamental monitoring parameter, directly reflecting the battery's energy storage level and electrochemical state. In a series-connected battery pack, the voltages of individual cells may differ, and this difference is a crucial basis for assessing battery consistency. Therefore, this embodiment simultaneously acquires voltage parameters at two levels: the voltage of a single cell (i.e., individual cell voltage) and the total voltage of the series circuit (i.e., cluster voltage). Individual cell voltage is used to assess the consistency of the state of charge (SOC) of each individual cell and to detect the presence of abnormal cells, while cluster voltage characterizes the overall voltage level and charge / discharge state of the entire battery pack. Similarly, operating current is the fundamental input for calculating battery charge / discharge energy, evaluating battery power performance, and performing SOC integration calculations.

[0094] Multiple temperature sensors are placed in different areas inside or on the surface of the battery. For a single cell, sensors can be placed in the center of the cell surface, near the positive terminal, near the negative terminal, and on the sidewall of the cell. For battery packs, sensors are also placed in critical locations such as module gaps and busbar connection points.

[0095] The equivalent internal resistance is obtained through the DC voltage drop method. A short-duration pulse excitation current is applied to the battery, and the change in battery terminal voltage during the pulse is simultaneously acquired. According to Ohm's law, the internal resistance equals the voltage change divided by the current change. To ensure measurement accuracy, the pulse excitation source needs to provide a current pulse with sufficient amplitude and a stable waveform, and voltage and current sampling needs to be performed synchronously at high frequency. Various sensing technologies can be used to acquire electrode deformation. Strain gauges are attached to the surface of the electrode or cell, and the minute deformation is measured based on the resistance change of the strain gauge, achieving contact deformation measurement. The module preload is acquired through pressure sensors arranged between cells or on the module frame. In this application, a flexible thin-film pressure sensor is arranged at the cell interlayer interface to acquire local pressure between cells; a flat plate force sensor is mounted on the module end plate or side plate to acquire the overall clamping force distribution of the module.

[0096] In some embodiments, after acquiring the battery monitoring data, the method further includes:

[0097] Determine the rate of change of each parameter in the monitoring data;

[0098] Determine whether the battery is in a steady-state or transient condition based on the rate of change.

[0099] When the battery is determined to be in a steady-state condition, the sampling frequency is updated to the first frequency;

[0100] When the battery is determined to be in a transient condition, the sampling frequency is updated to the second frequency, which is higher than the first frequency.

[0101] Battery monitoring data is acquired based on the updated sampling frequency.

[0102] First, the rate of change of each parameter in the monitoring data is determined. The rate of change of a parameter refers to the magnitude of change of the parameter value per unit time, which can be obtained by calculating the difference between the parameter value at the current moment and the parameter value at the previous moment and dividing by the sampling interval. When the rate of change is lower than a preset threshold, the battery is determined to be in a steady-state condition; when the rate of change is higher than the preset threshold, the battery is determined to be in a transient condition. When the battery is determined to be in a steady-state condition, the sampling frequency is updated to a first frequency (low frequency), which can be selected in the range of 1Hz to 10Hz. In a preferred embodiment of this application, 5Hz is used. When the battery is determined to be in a transient condition, the sampling frequency is updated to a second frequency (high frequency), which is higher than the first frequency and can be selected in the range of 100Hz to 1kHz. In a preferred embodiment of this application, 500Hz is used. After the sampling frequency is updated, subsequent monitoring data acquisition steps are performed based on the updated sampling frequency.

[0103] In some embodiments, after acquiring the battery monitoring data, the method further includes:

[0104] The monitoring data is filtered and denoised.

[0105] Amplitude calibration is performed on the filtered and denoised monitoring data;

[0106] Time-series alignment is performed on the amplitude-calibrated monitoring data;

[0107] The time-aligned monitoring data is then normalized.

[0108] Feature extraction is performed on the monitoring data, and the feature-extracted monitoring data is input into a pre-trained neural network to obtain the battery's state of charge, state of health, and remaining lifespan, including:

[0109] The normalized monitoring data is decomposed to filter out signal components related to environmental interference and hardware errors, and retain signal components related to battery aging, operating condition fluctuations and deformation stress as aging characteristic data.

[0110] Using aging characteristic data as input, and under the constraint of satisfying the physical laws of the battery aging process, the state of charge, health status and remaining service life output by the neural network are obtained.

[0111] The neural network was trained based on historical monitoring data and the corresponding state of charge, health status, and remaining service life.

[0112] The raw monitoring data acquired by sensors inevitably contains various types of noise, including the sensor's own electronic noise, electromagnetic interference coupling noise, and noise caused by power supply ripple. If this noise is used directly for subsequent analysis and modeling without processing, it will reduce the accuracy of feature extraction and the precision of neural network prediction. Monitoring data after filtering and denoising has a higher signal-to-noise ratio, removes noise components that are not beneficial to subsequent analysis, and provides a clean input signal for subsequent amplitude calibration, feature extraction, and neural network prediction.

[0113] A signal decomposition method is employed to break down monitoring data into different signal components. Signal components associated with environmental interference and hardware errors are filtered out, while effective signal components related to battery aging, operating condition fluctuations, and deformation stress are retained. The aging feature data obtained after feature extraction is input into a pre-trained neural network. This neural network uses historical monitoring data and corresponding labels for state of charge, health, and remaining service life as training samples, establishing a mapping relationship from monitoring data to battery state parameters through supervised learning. During training, the neural network continuously adjusts its internal weight parameters to minimize the error between predicted values ​​and true labels. After sufficient training, the neural network can simultaneously infer the current state of charge, health, and remaining service life of the battery based on the input current monitoring data.

[0114] Different sensors and signal conditioning circuits may have gain and bias errors, leading to deviations between measured values ​​and actual physical quantities. For example, the voltage divider resistors in a voltage sampling circuit may have resistance errors, and the output characteristics of a temperature sensor may vary individually. Therefore, amplitude calibration is required for each signal channel. Amplitude calibration eliminates systematic errors in each measurement channel, ensuring the comparability of measurement data and enabling comprehensive analysis and fusion processing of data acquired from different locations and sensors under a unified measurement benchmark.

[0115] Different sensors and acquisition channels may have varying signal delays and acquisition timing deviations. For example, voltage and current acquisition may be performed by different analog-to-digital converters at different trigger times, resulting in slight timing discrepancies between voltage and current data. For applications requiring precise synchronization of voltage and current, such as internal resistance calculations, this timing deviation can introduce measurement errors. Timing alignment eliminates timing discrepancies between multi-source data, ensuring the timing consistency of subsequent signal decomposition and neural network input data—a crucial prerequisite for accurate condition assessment and risk prediction.

[0116] Different physical quantities such as voltage, current, temperature, deformation, and preload have different dimensions and vastly different numerical ranges. If unnormalized data is directly input into a neural network, variables with larger numerical ranges will have higher weights in the network weight updates, affecting the network's training effect and convergence speed. Normalization makes the data of different physical quantities comparable in magnitude, eliminating the impact of dimensional differences on subsequent signal decomposition and neural network modeling, which is beneficial to improving the model's convergence speed and prediction accuracy.

[0117] Even after normalization, the monitoring data still contains signal components from various sources and with different characteristics. Among these, signal components associated with environmental interference and hardware errors are invalid or harmful, interfering with the subsequent neural network's identification of the battery's true state. Conversely, signal components associated with battery aging, operating condition fluctuations, and deformation stress contain crucial information about the battery's state evolution. Therefore, it is necessary to separate the effective signal components from the original signal.

[0118] In some embodiments, determining a battery balancing strategy based on state of charge and state of health includes:

[0119] When the health status is greater than or equal to the first health threshold, the state of charge control interval is set to the first state of charge interval.

[0120] When the health status is greater than or equal to the second health threshold and less than the first health threshold, the charge status control interval is set to the second charge interval, which is narrower than the first charge interval.

[0121] When the health status is less than the second health threshold, the state of charge control range is set to the third charge range, which is narrower than the second charge range, and the maximum charge and discharge rate of the battery is limited.

[0122] The charging and discharging management strategy corresponding to the state of charge control range is used as the balancing strategy.

[0123] A battery health status greater than or equal to the first health threshold indicates that the battery is relatively new or mildly aged. In this state, the loss of active materials is minimal, the internal resistance increase is small, and the electrochemical performance is excellent. Under these conditions, the battery can withstand a wide range of charge and discharge operations without excessive restrictions on the state of charge (SCC) range, thus fully utilizing the battery's capacity. The SCC control range refers to the range of SCC within which the battery is allowed to charge and discharge. For batteries in good health, the control range is set to a relatively wide first SCC range, such as 20% to 100%. This means the battery can be charged and discharged normally within this range.

[0124] When the battery's health status falls between the second and first health thresholds, it has entered a moderate aging stage. At this point, the battery has experienced some loss of active material and increased internal resistance, narrowing its safe operating window. Therefore, the state-of-charge (SOC) control range needs to be compressed to ensure the battery operates within a safer charge range and avoids continued operation under extreme charge conditions. An example value for the second SOC threshold is 80%, but this value can be flexibly set according to actual conditions. The second SOC range is narrower than the first SOC range. For example, when the first SOC range is 20% to 100%, the second SOC range can be set to 30% to 90%.

[0125] When the battery's health status falls below the second health threshold, it has entered a severe aging stage. At this point, a significant amount of active material is lost, internal resistance increases substantially, heat generation increases, and the risk of thermal runaway rises significantly. Therefore, it is necessary to further narrow the state-of-charge (SOC) control range and limit the charge / discharge current rate to control heat generation and prevent localized overheating. In this application, when the SOC is below the second SOC threshold, the maximum charge / discharge rate is limited to below 0.5C to ensure the battery's operational safety during the severe aging stage. The third SOC range is further narrower than the second SOC range. For example, when the second SOC range is 30% to 90%, the third SOC range can be set to 40% to 80%. The maximum charge / discharge rate limit refers to limiting the magnitude of the battery's charge / discharge current. The C-rate is a unit representing the charge / discharge rate; 1C indicates that the current value equals the battery capacity. Limiting the maximum charge / discharge rate means reducing the charge / discharge speed by controlling the current magnitude within the system.

[0126] In some embodiments, determining an early warning threshold based on health status includes:

[0127] The aging stage of the battery is determined based on its health status;

[0128] The initial temperature warning threshold is adjusted according to the aging stage, and the adjusted initial temperature warning threshold is used as the temperature warning threshold.

[0129] The voltage warning threshold and internal resistance warning threshold are determined based on the aging stage.

[0130] There is a correlation between battery health status and battery aging stages. The current aging stage of the battery can be determined by the numerical range of the health status. The aging stage is divided into at least two stages. This stage division essentially discretizes the continuous change in health status into several levels. Batteries at different aging stages have different thermal characteristics. Due to changes in their internal structure, aged batteries have reduced thermal conductivity and changed heat capacity, and may experience a higher temperature rise for the same amount of heat generated, thus requiring stricter temperature warning thresholds. The initial temperature warning threshold is a baseline value set for a new battery state. The initial temperature warning threshold is corrected according to the aging stage. The voltage warning threshold and internal resistance warning threshold also need to be adjusted according to the aging stage. The voltage warning threshold is mainly used to assess the voltage consistency between individual cells and to detect voltage anomalies. The determination of the voltage warning threshold and internal resistance warning threshold is also based on the aging stage.

[0131] In some embodiments, determining the risk level of a battery based on monitoring data, remaining lifespan, and warning thresholds includes:

[0132] Based on the currently collected temperature, determine the highest temperature of a single cell and the maximum temperature difference between cells in the series circuit of the battery.

[0133] When the highest temperature exceeds the temperature threshold or the maximum temperature difference exceeds the temperature difference threshold, it is determined to be at the first risk level.

[0134] When an internal battery malfunctions, temperature is often the first parameter to change. By monitoring the maximum temperature and maximum temperature difference, two different types of temperature anomalies can be identified: global overheating and localized overheating. Global overheating is characterized by an excessive maximum temperature, which may be caused by overcharging, over-discharging, external short circuits, or excessively high ambient temperatures. Localized overheating is characterized by a large temperature difference between individual battery cells, which may be caused by abnormal heat generation in individual cells or uneven heat dissipation within the module. Clearly defined criteria are established; when either of these two indicators exceeds its corresponding threshold, it indicates that the battery has experienced a current temperature anomaly, requiring a Level 1 risk warning.

[0135] The Level 1 warning is triggered by measured thresholds, current visible anomalies, and physical threshold criteria. The core criterion formula is a maximum temperature > 45℃ or a maximum temperature difference > 0.7℃.

[0136] The Level 1 warning uses real-time hard threshold determination of physical quantities, eliminating the need for intelligent model calculations. It directly calculates the highest cell temperature and the maximum temperature difference within the cell cluster based on collected multi-point temperature data. When any indicator exceeds the set safety threshold, the battery is deemed to have a minor, noticeable anomaly. It iterates through all cell temperature measurement points, calculates the extreme temperature difference within the cluster in real time, and triggers a local audible and visual warning immediately if the threshold condition is met. This level of anomaly only provides status indication, data retention, and trend monitoring; it does not interfere with the battery's charging and discharging operations and is used to identify early, minor, and recoverable temperature-related anomalies.

[0137] In some embodiments, determining the risk level of a battery based on monitoring data, remaining lifespan, and warning thresholds includes:

[0138] The hidden state of the prediction model is updated, and the relationship for the hidden state update is:

[0139] ;

[0140] in, Let be the hidden state at time t. The hidden state includes the battery aging characteristics and temperature change trends remembered by the prediction model. For activation function, Let be the weighting parameters for temperature and equivalent internal resistance at time t. Let be the temperature and equivalent internal resistance at time t. The weight parameters of the hidden state at time t-1 are... The hidden state at time t-1 For bias terms;

[0141] The temperature and equivalent internal resistance are input into the prediction model to obtain the prediction result output by the model. The expression for the prediction result is as follows:

[0142] ;

[0143] in, For the output of the prediction model Time to The prediction results at any given time include predicted temperature and predicted equivalent internal resistance. Let be the temperature and equivalent internal resistance at time t. The temperature and equivalent internal resistance at time t-1 are given. Let be the temperature and equivalent internal resistance at time tN;

[0144] exist to If, at any given moment, either the predicted value of temperature or the equivalent internal resistance exceeds the corresponding safety upper limit threshold, it is determined to be at the second risk level.

[0145] The role of LSTM hidden state update is to remember the historical operating conditions and aging trends of LSTM. It is the core memory unit of time series prediction, allowing the model to remember the past changes of the battery, so as not to misjudge due to instantaneous noise and to capture the slow deterioration trend.

[0146] The Level 2 early warning system employs an LSTM (Long Short-Term Memory) temporal neural network to address the limitation of traditional thresholds in predicting future risks. The model takes N historical frames of temperature and internal resistance time-series data as input, and uses a gating mechanism to memorize battery aging and operating condition changes, continuously predicting parameter sequences for the next hour (60 sets). Specifically, within the next hour prediction window, if either the predicted temperature or internal resistance value is about to exceed the safety upper limit threshold, it is immediately classified as a developing moderate risk.

[0147] The time-series dataset is updated in real time and fed into the LSTM model for rolling simulation, outputting the parameter change trend for the next hour. Under the premise that the physical parameters have not exceeded the standard, the deterioration trend close to the threshold can be identified in advance, and the operation and maintenance early warning prompts can be pushed proactively, providing a predictive basis for flexible interventions such as battery operating rate reduction and heat dissipation adjustment.

[0148] The determination of the first risk level relies on parameter values ​​that have already exceeded limits at the current moment, essentially a response to an anomaly that has already occurred. However, some risks show a worsening trend before the parameters actually exceed limits. To avoid reactive responses after a failure occurs, it is necessary to identify such trend risks in advance. The prediction model employs a long short-term memory network structure. The prediction model outputs a sequence of predicted temperatures and equivalent internal resistances for a preset future time period. When determining the second risk level, if any predicted value exceeds the corresponding safety upper limit threshold, it is classified as a second risk level.

[0149] In some embodiments, determining the risk level of a battery based on monitoring data, remaining lifespan, and warning thresholds includes:

[0150] Based on monitoring data, remaining service life, and warning thresholds, the scoring model is used to obtain a battery failure probability score. The expression for the failure probability score is as follows:

[0151] ;

[0152] in, For the i-th set of monitoring data The score of the failure probability obtained under k decision trees Let F be the discrimination result of the m-th decision tree, and F be the function space of all decision trees. There are a total of k decision trees.

[0153] The scoring is constrained by the following relationship:

[0154] ;

[0155] in, This is the total error. For the actual faults in the i-th set of monitoring data With predicting failure deviation, For regularization terms, Used to control the number of decision trees Used for weight decay;

[0156] Under the condition of satisfying the constraints, when the predicted fault score is higher than the high-risk threshold, the change in the equivalent internal resistance of the battery within a preset time period exceeds the mutation threshold, and the correspondence between the open-circuit voltage and the state of charge of the battery deviates from the benchmark correspondence and the deviation exceeds the distortion threshold, it is judged as the third risk level.

[0157] The first and second risk levels are determined based on current measured parameters and future single-parameter trends, respectively. However, both have limitations in handling complex fault types with multiple coupled parameters. Therefore, a third risk level is established. The scoring model is an extreme gradient boosting tree model. This model uses open-circuit voltage curve distortion rate, instantaneous surge gradient of internal resistance, temperature change rate, deformation increment, and preload fluctuation coefficient as input features. It obtains the predicted fault score through cumulative reasoning of multiple decision trees. Three conditions are judged simultaneously: the predicted fault score is higher than the high-risk threshold, the change in equivalent internal resistance within a preset time period exceeds the abrupt change threshold, and the correspondence between open-circuit voltage and state of charge deviates from the baseline correspondence and the deviation exceeds the distortion threshold.

[0158] The three-level early warning system employs the Extreme Gradient Boosting Tree (XGBoost) algorithm. XGBoost outputs a battery fault risk score by stacking multiple decision trees. It reduces battery fault prediction bias through error fitting terms, improving the accuracy of hidden fault identification. Regularization constraints suppress model overfitting and filter out misjudgments caused by operating noise and sensor interference. This enables the XGBoost model to stably identify high-risk hidden defects such as OCV curve distortion, abnormal internal resistance step changes, micro-short circuits, and precursors to thermal runaway, achieving highly reliable high-risk early warning. A multi-dimensional fault classification model is constructed. Input features include high-dimensional coupled features such as OCV open-circuit voltage curve distortion rate, instantaneous internal resistance surge gradient, temperature change rate, deformation increment, and preload fluctuation coefficient. Fault scores are obtained through cumulative inference using multiple decision trees, and regularization terms suppress overfitting, improving the generalization ability of fault identification.

[0159] When the model outputs a fault probability score higher than the high-risk threshold, and simultaneously meets the two major characteristics of a sudden change in internal resistance and nonlinear distortion of the OCV curve, it is judged as a precursor to thermal runaway and a high-risk irreversible fault hazard.

[0160] Real-time extraction of battery electrical aging distortion features is fed into the trained XGBoost model for high-precision fault identification. It can identify deep faults such as hidden micro-short circuits, abnormal SEI thickening, and local polarization runaway that cannot be captured by traditional threshold methods, triggering the highest level of safety warning and forced power-off protection mechanism.

[0161] The system employs a multi-condition parallel real-time polling discrimination mechanism for the three-level early warning conditions, simultaneously running three types of algorithms: physical threshold judgment, LSTM time series prediction, and XGBoost fault classification. It achieves hierarchical progression and advanced locking through a risk priority constraint formula.

[0162] Once a low-level warning is triggered, it can be iterated and upgraded to a high-level risk. Once a level 3 high-risk warning is triggered, the system automatically locks into the highest protection state, blocking logic and strategy conflicts of the first and second level low-level warnings.

[0163] Based on the above, the battery control includes: automatically starting PTC heating when the temperature is below -10℃, and charging can only be performed after the temperature rises to 5℃; and starting forced water cooling when the temperature is above 45℃; when the voltage difference between individual cells in the battery cluster is >50mV, starting the 1.5A active balancing mode; if a sign of thermal runaway is detected, the FPGA directly controls the DC circuit breaker to cut off the circuit power supply, meeting the new national standard requirements of 5-minute early warning and 2-hour safety protection.

[0164] The system adaptively optimizes its management strategy based on the battery's real-time State of Health (SOH). For brand-new cells with SOH ≥ 90%, the SOC range is controlled between 20% and 100%. For moderately aged cells with SOH ≤ 90% and SOH < 80%, the SOC range is controlled between 30% and 90%. For severely aged cells with SOH < 80%, the maximum charge / discharge rate is limited to 0.5C. Simultaneously, the maintenance cycle is dynamically adjusted: brand-new cells undergo SOC calibration every 30 days, while aged cells are calibrated every 15 days, comprehensively extending the battery's cycle life.

[0165] Figure 2 This is a schematic diagram of a battery management device provided by the present invention. The battery management device includes:

[0166] Memory 21 is used to store computer programs;

[0167] The processor 22 is used to implement the steps of the battery management method described above when executing a computer program.

[0168] The description of the battery management device provided in this application is based on the above embodiments and will not be repeated here.

[0169] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0170] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0171] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily 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 invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A battery management method, characterized in that, include: Acquire battery monitoring data, including the battery's electrical parameters, thermal parameters, and mechanical parameters; Feature extraction is performed on the monitoring data, and the feature-extracted monitoring data is input into a pre-trained neural network to obtain the battery's state of charge, state of health, and remaining lifespan. A battery balancing strategy is determined based on the state of charge and the state of health. A warning threshold is determined based on the health status, and the warning threshold is used to assess battery risk. The risk level of the battery is determined based on the monitoring data, the remaining service life, and the warning threshold. The battery is managed according to the risk level and the balancing strategy.

2. The battery management method as described in claim 1, characterized in that, Acquire battery monitoring data, including: The voltage of a single battery, the total voltage of the series circuit of the battery, and the operating current in the charging and discharging circuit of the battery are obtained as the electrical parameters. Temperature values ​​collected at multiple preset locations on the battery are obtained as the thermal parameters; The equivalent internal resistance, the electrode deformation between the internal electrode and the external casing, and the module preload applied to the battery by the module containing the battery are obtained as the mechanical parameters, which are determined by applying a pulse excitation to the battery and collecting the response signal of the battery.

3. The battery management method as described in claim 1, characterized in that, After acquiring the battery monitoring data, the following is also included: Determine the rate of change of each parameter in the monitoring data; The battery is determined to be in a steady-state or transient state based on the rate of change. When the battery is determined to be in a steady-state condition, the sampling frequency is updated to the first frequency; When it is determined that the battery is in a transient operating condition, the sampling frequency is updated to a second frequency, which is higher than the first frequency; Monitoring data of the battery is acquired based on the updated sampling frequency.

4. The battery management method as described in claim 1, characterized in that, After acquiring the battery monitoring data, the following is also included: The monitoring data is then filtered and denoised. Amplitude calibration is performed on the filtered and denoised monitoring data; Time-series alignment is performed on the amplitude-calibrated monitoring data; The time-aligned monitoring data is then normalized. Feature extraction is performed on the monitoring data, and the feature-extracted monitoring data is input into a pre-trained neural network to obtain the battery's state of charge, state of health, and remaining lifespan, including: The normalized monitoring data is decomposed to filter out signal components associated with environmental interference and hardware errors, and retain signal components associated with battery aging, operating condition fluctuations and deformation stress as aging characteristic data. Using the aging characteristic data as input, and under the constraint of satisfying the physical laws of the battery aging process, the state of charge, health status, and remaining service life output by the neural network are obtained. The neural network is trained based on historical monitoring data and corresponding state of charge, health status, and remaining service life.

5. The battery management method as described in claim 1, characterized in that, Determining a battery balancing strategy based on the state of charge and the state of health includes: When the health status is greater than or equal to the first health threshold, the charge state control interval is set as the first charge interval; When the health status is greater than or equal to the second health threshold and less than the first health threshold, the charge state control interval is set as the second charge interval, which is narrower than the first charge interval. When the health status is less than the second health threshold, the state of charge control range is set to a third charge range, which is narrower than the second charge range, and the maximum charge / discharge rate of the battery is limited. The charging and discharging management strategy corresponding to the state of charge control interval is used as the balancing strategy.

6. The battery management method as described in claim 1, characterized in that, Determining the warning threshold based on the health status includes: The aging stage of the battery is determined based on its health status; The initial temperature warning threshold is corrected according to the aging stage, and the corrected initial temperature warning threshold is used as the temperature warning threshold. The voltage warning threshold and internal resistance warning threshold are determined based on the aging stage.

7. The battery management method according to any one of claims 1 to 6, characterized in that, The risk level of the battery is determined based on the monitoring data, the remaining service life, and the warning threshold, including: Based on the currently collected temperature, determine the highest temperature of a single cell and the maximum temperature difference between cells in the series circuit of the battery. When the highest temperature is greater than the temperature threshold or the maximum temperature difference is greater than the temperature difference threshold, it is determined to be at the first risk level.

8. The battery management method according to any one of claims 1 to 6, characterized in that, The risk level of the battery is determined based on the monitoring data, the remaining service life, and the warning threshold, including: The hidden state of the prediction model is updated, and the relationship for the hidden state update is: ； in, Let be the hidden state at time t, where the hidden state includes the battery aging characteristics and temperature change trends memorized by the prediction model. For activation function, Let be the weighting parameters for temperature and equivalent internal resistance at time t. Let be the temperature and equivalent internal resistance at time t. Let be the weight parameters of the hidden state at time t-1. The hidden state at time t-1 For bias terms; Temperature and equivalent internal resistance are input into the prediction model to obtain the prediction result output by the prediction model. The expression of the prediction result is as follows: ； in, For the output of the prediction model Time to The prediction results at time, including predicted temperature and predicted equivalent internal resistance, Let be the temperature and equivalent internal resistance at time t. The temperature and equivalent internal resistance at time t-1 are given. Let be the temperature and equivalent internal resistance at time tN; exist to If, at any given moment, either the predicted value of temperature or the equivalent internal resistance exceeds the corresponding safety upper limit threshold, it is determined to be at the second risk level.

9. The battery management method according to any one of claims 1 to 6, characterized in that, The risk level of the battery is determined based on the monitoring data, the remaining service life, and the warning threshold, including: Based on the monitoring data, the remaining service life, and the warning threshold, the scoring model is used to obtain a score for the battery's failure probability, which is output by the scoring model. The expression for the failure probability score is as follows: ； in, For the i-th set of monitoring data The score of the failure probability obtained under k decision trees Let F be the discrimination result of the m-th decision tree, and F be the function space of all decision trees. There are a total of k decision trees. The scoring is constrained by the following relationship: ； in, This is the total error. For the true fault of the i-th set of monitoring data With predicting failure deviation, For regularization terms, Used to control the number of decision trees Used for weight decay; Under the condition of satisfying the above constraints, when the predicted fault score is higher than the high-risk threshold, the change in the equivalent internal resistance of the battery within a preset time period exceeds the mutation threshold, and the correspondence between the open-circuit voltage and the state of charge of the battery deviates from the benchmark correspondence and the deviation exceeds the distortion threshold, it is determined to be the third risk level.

10. A battery management device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the battery management method as described in any one of claims 1 to 9.

Images