A charging method based on adjustable charging curves for battery health status sensing

Abstract

This invention discloses a charging method based on an adjustable charging curve for battery health status perception, comprising the following steps: synchronously acquiring voltage and current responses through a micro EIS measurement circuit and obtaining the real and imaginary parts of impedance at different frequencies using Fast Fourier Transform, thereby generating an electrochemical impedance spectrum of the battery during charging and discharging; segmenting and fitting the high-frequency, mid-frequency, and low-frequency regions, extracting characteristic values ​​of ohmic internal resistance, charge transfer impedance, and diffusion impedance respectively, and generating impedance spectrum characteristic parameters; inputting the impedance spectrum characteristic parameters to a control chip to fit and estimate the battery health status assessment value, and constructing a battery health status perception model to dynamically generate an adjustable charging curve; performing a charging operation on the battery according to the adjustable charging curve, and dynamically adjusting the parameters of the adjustable charging curve, thereby achieving closed-loop adaptive adjustment of the charging process. This invention can dynamically adjust the charging curve parameters to achieve closed-loop adaptive adjustment of the charging process.

Description

Technical Field

[0001] This invention relates to the field of battery charging technology, and more specifically, to a charging method based on an adjustable charging curve that senses battery health status. Background Technology

[0002] Lithium-ion batteries, with their advantages of high energy density, long cycle life, and low environmental pollution, have been widely used in consumer electronics, electric vehicles, energy storage power stations, and other fields, becoming a core component of modern society's energy storage and supply system. As various electrical devices increasingly demand higher range, charging efficiency, and safety, the optimization and upgrading of battery charging management technology has become a key direction for industry development. The rationality of charging strategies directly determines the battery's cycle life, charging safety, and energy utilization efficiency. Among these, the State of Health (SOH), as a core indicator reflecting the degree of battery aging and remaining performance, requires precise sensing and dynamic adaptation to charging strategies, becoming a key breakthrough in improving charging management.

[0003] Currently, various technical solutions have been proposed in the industry for battery charging management to adapt to charging needs in different scenarios. The existing technologies mainly include the following categories:

[0004] The traditional constant current constant voltage (CC-CV) charging scheme is the most basic and widely used charging method for lithium-ion batteries. This scheme mainly consists of two stages: first, constant current charging is performed with a constant high current; when the battery voltage reaches a preset cutoff voltage, it switches to a constant voltage mode to continue charging until the charging current drops to the cutoff current, at which point charging is terminated. This scheme uses fixed charging parameters (current value, voltage value), which can only meet the standard charging requirements of new batteries. It has a simple structure and low cost, but it cannot adapt to the aging characteristics changes throughout the battery's entire life cycle.

[0005] To overcome the limitations of traditional CC-CV solutions, some technical solutions attempt to adjust the charging switching point based on battery status. The core of this approach is to collect battery status parameters such as voltage, temperature, and internal resistance in real time, and dynamically determine the switching threshold voltage between constant current mode and constant voltage mode based on these parameters. This achieves dynamic adjustment of the charging mode switching timing. However, this type of solution still does not change the current amplitude in the constant current stage and the voltage amplitude in the constant voltage stage. The adjustment range of charging parameters is relatively narrow, and it cannot achieve adaptation for the entire charging cycle.

[0006] Another approach proposes predicting State of Health (SOH) based on external data and adjusting charging strategies accordingly. This type of approach obtains external data such as ambient battery temperature, user charging habits, and vehicle mileage, inputs this data into an SOH assessment model to predict battery health, and then optimizes the charging strategy based on the predicted SOH value. However, this approach relies on indirect external data for SOH assessment, which is a soft measurement method. It cannot reflect the true electrochemical state inside the battery (such as increased internal resistance or loss of active materials), resulting in low accuracy in SOH prediction and difficulty in capturing instantaneous changes in battery health, thus failing to provide a precise basis for adjusting charging strategies.

[0007] In addition, there are device-side battery status detection and charging adjustment solutions. This solution integrates battery status measurement circuits and intelligent decision-making algorithms into the device being charged (such as mobile phones and electric vehicles). By detecting parameters such as battery voltage, current, and temperature, it adjusts charging parameters to reduce battery capacity loss. However, the intelligent diagnosis and decision-making in this solution are all completed on the device side. The charger only acts as a passive power supply device and does not have active sensing and adaptation capabilities. At the same time, the integration of complex measurement circuits and algorithms on the device side also increases the production cost and size of the device.

[0008] While the aforementioned existing technologies have improved battery charging management to some extent, they still have significant technical shortcomings: traditional CC-CV solutions cannot adapt to battery aging characteristics, easily accelerating battery degradation and even causing safety hazards; solutions based on charging switching point adjustment have limited accuracy in SOH assessment and a narrow range of charging parameter adjustments, lacking self-learning capabilities; solutions based on external data to predict SOH cannot reflect the true internal state of the battery, resulting in insufficient prediction accuracy; and device-side detection solutions increase equipment costs, and chargers lack proactive adaptation capabilities. In summary, existing technologies generally lack a solution where the charger proactively, accurately, and quickly senses the battery's health status and generates a refined, adaptive charging strategy covering the entire charging cycle. This makes it difficult to meet the management needs of the entire battery lifecycle and to balance charging efficiency, safety, and battery life. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a charging method based on an adjustable charging curve that senses battery health status, comprising the following steps:

[0010] Step S1: Simultaneously acquire voltage and current responses through a miniature EIS measurement circuit and obtain the real and imaginary parts of impedance at different frequencies through fast Fourier transform; generate the electrochemical impedance spectrum of the battery during the charging and discharging process based on the real and imaginary parts of impedance at different frequencies.

[0011] Step S2: Based on the electrochemical impedance spectrum of the battery during the charging and discharging process, the high-frequency region, mid-frequency region and low-frequency region are segmented and fitted, and the characteristic values ​​of ohmic internal resistance, charge transfer impedance and diffusion impedance are extracted respectively to generate impedance spectrum characteristic parameters; the impedance spectrum characteristic parameters are input into the control chip for pre-storage and fitted with the battery health status mapping table to estimate the corresponding battery health status assessment value.

[0012] Step S3: Obtain the current state of charge of the battery and construct a battery health status perception model with the battery health status assessment value to dynamically generate an adjustable charging curve;

[0013] Step S4: Perform a charging operation on the battery according to the adjustable charging curve, and periodically collect the real-time impedance spectrum and state of charge data of the battery during the charging process. Feed these data back to the battery health status perception model to update the battery health status assessment value. Then, dynamically adjust the parameters of the adjustable charging curve according to the updated battery health status assessment value to achieve closed-loop adaptive adjustment of the charging process.

[0014] The beneficial effects of this application are as follows:

[0015] This method synchronously acquires voltage and current responses using a miniature EIS measurement circuit, and obtains the real and imaginary parts of impedance at different frequencies using Fast Fourier Transform (FFT), thereby generating an electrochemical impedance spectroscopy (EIS) for the charging and discharging process. Existing solutions for estimating battery status based on external data can only judge battery condition indirectly through data such as ambient temperature and usage habits, failing to reflect the true internal state, such as changes in internal resistance and loss of active materials, resulting in a lack of reliable basis for subsequent status assessment. In contrast, this step directly acquires the internal electrochemical response signal of the battery, and the generated impedance spectrum can comprehensively present the ohmic polarization, charge transfer polarization, and concentration polarization characteristics of the battery at different frequencies, fully reflecting the internal state changes of the battery during charging and discharging, and providing direct and comprehensive raw data support for subsequent health status assessment. Secondly, by segmenting and fitting the high-frequency, mid-frequency, and low-frequency regions of the electrochemical impedance spectroscopy, characteristic values ​​of ohmic internal resistance, charge transfer impedance, and diffusion impedance are extracted respectively. These values ​​are then combined with a pre-stored mapping table in the control chip to estimate the battery health status. The segmented fitting extracts characteristic parameters of different types of impedance, which are directly related to the battery's internal aging mechanism and can accurately reflect the battery's aging degree and remaining performance. The fitting estimation using the pre-stored mapping table enables direct assessment of the battery's health status. Furthermore, the assessment results can adapt to the aging changes throughout the battery's entire life cycle, providing a reliable basis for the dynamic adjustment of subsequent charging strategies. Simultaneously, the integrated application of the control chip enables rapid response in the assessment process, meeting the requirements of real-time charging management. Then, the current state of charge of the battery is obtained and combined with the battery health status assessment value to construct a battery health status perception model, and an adjustable charging curve is dynamically generated. By combining the state of charge with the health status assessment value, the constructed perception model can comprehensively consider the real-time state of the battery and dynamically generate an adjustable charging curve that adapts to the current state. This breaks the limitations of fixed charging parameters and realizes personalized customization of the charging curve. It can ensure charging efficiency when the battery health is good and reduce charging damage after the battery ages. At the same time, the adjustability of the curve also provides a basis for subsequent closed-loop adjustment, ensuring that the charging strategy can be adjusted in a timely manner according to changes in the battery status.Finally, charging is performed according to the adjustable charging curve. During the charging process, real-time impedance spectrum and state of charge data are periodically collected and fed back to the sensing model to update the health status assessment value. The charging curve parameters are dynamically adjusted to achieve closed-loop adaptive adjustment of the charging process. By periodically collecting real-time data, the health status assessment value is dynamically updated, which can capture the internal state changes of the battery in a timely manner. Based on the updated assessment value, the charging curve parameters are adjusted to form a closed-loop adjustment mechanism, so that the charging strategy always adapts to the real-time state of the battery. This avoids battery damage caused by charging with fixed parameters and ensures charging efficiency. At the same time, the charger actively completes sensing, assessment and adjustment without the need for complex circuits and algorithms integrated on the device side, which reduces the equipment production cost and improves the initiative and adaptability of charging management, thereby achieving a balance between charging efficiency, safety and battery life. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0017] Figure 1 This is a flowchart illustrating the steps of the charging method based on the adjustable charging curve of battery health status in this embodiment.

[0018] Figure 2 This is a schematic diagram of the electrical principle of the miniature EIS measurement circuit in this embodiment;

[0019] Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3. Detailed Implementation

[0020] The following drawings disclose several embodiments of the present invention. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential. Furthermore, for the sake of simplicity, some conventional structures and components will be shown in the drawings in a simple schematic manner.

[0021] To further understand the invention's content, features, and effects, the following embodiments are provided, and detailed descriptions are given below in conjunction with the accompanying drawings:

[0022] Reference Figure 1-2 , Figure 1 This is a flowchart illustrating the steps of the charging method based on the adjustable charging curve of battery health status in this embodiment. Figure 2This is a schematic diagram of the electrical principle of the miniature EIS measurement circuit in this embodiment. The charging method based on the adjustable charging curve of battery health status sensing in this embodiment includes the following steps:

[0023] Step S1: Simultaneously acquire voltage and current responses through a miniature EIS measurement circuit and obtain the real and imaginary parts of impedance at different frequencies through fast Fourier transform; generate the electrochemical impedance spectrum of the battery during the charging and discharging process based on the real and imaginary parts of impedance at different frequencies.

[0024] In this embodiment of the invention, the process of synchronously acquiring voltage and current responses and generating electrochemical impedance spectroscopy through a miniature EIS measurement circuit is achieved using a miniature EIS measurement circuit integrated inside the charger. This circuit includes a programmable arbitrary waveform generator, a high-precision data acquisition unit, and an analog switch, with overall power consumption controlled below 100mW to ensure that it does not affect the normal charging function of the charger. The programmable arbitrary waveform generator uses a 12-bit resolution DAC device. The control chip sends a control signal and outputs a digital signal corresponding to a pseudo-random binary sequence. The symbol rate is set to 1kHz~1MHz, covering the main frequency range of the battery impedance response. The digital signal is amplified by a signal conditioning circuit composed of operational amplifiers, and the amplitude is controlled to be 20mV~50mV to form an AC excitation signal. An analog switch is connected in series between the battery and the charger power output terminal, and the control chip controls its on / off state to connect the AC excitation signal to both ends of the battery. At the same time, the charger power circuit is isolated by an optocoupler isolation device to avoid interference. The instrumentation amplifier in the high-precision data acquisition unit differentially amplifies the battery voltage response, with a magnification factor set to 100~1000 times. A 16-bit ADC device, with a sampling rate of 2MHz or higher, synchronously acquires the amplified voltage and current response signals to generate time-domain data. The control chip performs a Fast Fourier Transform (FFT) on the time-domain data, using a radix-2 FFT algorithm to convert the 1024 sampling points into frequency-domain data. It then calculates the real and imaginary parts of the impedance at different frequencies. The real part equals the product of the impedance magnitude and the cosine of the phase angle, and the imaginary part equals the product of the impedance magnitude and the sine of the phase angle. After processing, an electrochemical impedance spectrum covering 1kHz~1MHz is generated.

[0025] Step S2: Based on the electrochemical impedance spectrum of the battery during the charging and discharging process, the high-frequency region, mid-frequency region and low-frequency region are segmented and fitted, and the characteristic values ​​of ohmic internal resistance, charge transfer impedance and diffusion impedance are extracted respectively to generate impedance spectrum characteristic parameters; the impedance spectrum characteristic parameters are input into the control chip for pre-storage and fitted with the battery health status mapping table to estimate the corresponding battery health status assessment value.

[0026] In this embodiment of the invention, the process of segmented fitting to extract impedance characteristic parameters and estimate battery health status is completed by the control chip's computing unit. First, the electrochemical impedance spectrum is segmented according to its frequency response: the high-frequency region (100kHz~1MHz) corresponds to ohmic internal resistance, the mid-frequency region (1kHz~100kHz) corresponds to charge transfer impedance and double-layer capacitance, and the low-frequency region (below 1kHz) corresponds to diffusion impedance. In the high-frequency region, the 500kHz~1MHz frequency band is selected, and the real part of the impedance is linearly fitted. The intercept b in the fitting equation y=kx+b is the ohmic internal resistance characteristic value. Simultaneously, the phase angle is extracted for verification, ensuring it is within the range of -5° to 5°. In the mid-frequency region, the 10kHz~50kHz frequency band is selected. The intersection points of the real and imaginary parts corresponding to the semi-circular diameter of the impedance spectrum are analyzed. The charge transfer impedance characteristic value is obtained through fitting an equivalent circuit model, and the peak phase angle is extracted for verification. In the low-frequency region, the 100Hz~1kHz frequency band is selected, and the difference between the real and imaginary parts is linearly fitted to obtain the diffusion impedance characteristic value. The rationality is verified by combining the phase angle change. The control chip's Flash memory stores accelerated aging test data for 20 sets of batteries of the same model, forming a mapping table between characteristic parameters and health status. The mapping table includes threshold values ​​for each impedance characteristic parameter and corresponding health levels. The extracted impedance characteristic parameters are substituted into the mapping table, and the remaining capacity estimate, internal resistance growth rate estimate, and capacity decay trend estimate are calculated through a combination of linear regression and nonlinear fitting. These are used together as the battery health status assessment value.

[0027] Step S3: Obtain the current state of charge of the battery and construct a battery health status perception model with the battery health status assessment value to dynamically generate an adjustable charging curve;

[0028] In this embodiment of the invention, the process of constructing a battery health status perception model and dynamically generating an adjustable charging curve is implemented using an internal model of the control chip. First, the current state of charge (SOC) of the battery is acquired, calculated by integrating the battery voltage and current, and updated every 100ms, ranging from 0% to 100%. A perception model is constructed based on a local lightweight learning algorithm, employing a small-scale feedforward neural network. The input layer has three nodes that receive the health status assessment value, the current SOC, and the battery's internal temperature. The hidden layers each have 16 neurons and include a self-attention gating unit to dynamically adjust the input feature weights. The output layer has three nodes corresponding to the charging curve parameters. The model is trained using a historical charging optimization dataset containing 1000 charge-discharge cycles, divided into a training set and a validation set in an 8:2 ratio. Training is iterated 50 times with a learning rate of 0.001, incorporating knowledge distillation loss to constrain parameter updates, and training continues until the total loss function value is less than 0.001. By combining real-time collected state of charge-internal resistance change gradient data to fine-tune the model, a dynamic weight bias vector is generated and superimposed on the output layer. After the model is deployed, the current health assessment value, state of charge and temperature are input, and the charging current, voltage and cutoff condition parameters are calculated through forward propagation. Based on the parameters, three continuous smooth curves are generated, which together form an adjustable charging curve to adapt to the current battery state.

[0029] Step S4: Perform a charging operation on the battery according to the adjustable charging curve, and periodically collect the real-time impedance spectrum and state of charge data of the battery during the charging process. Feed these data back to the battery health status perception model to update the battery health status assessment value. Then, dynamically adjust the parameters of the adjustable charging curve according to the updated battery health status assessment value to achieve closed-loop adaptive adjustment of the charging process.

[0030] In this embodiment of the invention, based on the adjustable charging curve, the control chip outputs a PWM signal to drive the power conversion module to perform a charging operation on the battery. During the charging process, real-time impedance spectrum and state of charge (SOC) data of the battery are periodically collected every 200ms. The real-time impedance spectrum is collected using a consistent method, and the SOC data is updated in real time through voltage and current integration. The collected real-time data is fed back to the battery health status perception model. The model recalculates the impedance characteristic parameters based on the new data and updates the battery health status assessment value. The update frequency is consistent with the collection frequency. Based on the updated health status assessment value, the model dynamically adjusts the charging curve parameters. The adjustment step size is set according to the trend of health status changes. When the health status declines rapidly, the adjustment step size is 0.05C, and when the change is gradual, it is 0.2C. The adjustment frequency is synchronized with the data collection frequency. After the parameters are adjusted, the adjustable charging curve is regenerated and transmitted to the power conversion module. The charging current, voltage, and cutoff conditions are adjusted in real time, forming a closed-loop adaptive adjustment of "charging-collection-feedback-update-adjustment" to ensure that the charging process always adapts to the real-time health status of the battery, balancing charging efficiency and battery life protection.

[0031] Furthermore, the miniature EIS measurement circuit mentioned in step S1 includes a programmable arbitrary waveform generator, a high-precision data acquisition unit, and an analog switch. The electrochemical impedance spectroscopy of the generated battery during the charging and discharging process includes the following steps:

[0032] Design a programmable arbitrary waveform generator. The digital signal corresponding to the pseudo-random binary sequence output by the DAC is amplified by the signal conditioning circuit composed of operational amplifier Q1 to form an AC excitation signal with an amplitude controlled within 50mV. The symbol rate of the pseudo-random binary sequence is set to a value that covers the main frequency range of the battery impedance response, so as to obtain wideband impedance information in one excitation process.

[0033] An analog switch is connected in series between the battery and the charger. The AC excitation signal is connected to both ends of the battery by controlling the on and off of the analog switch. At the same time, the power circuit of the charger is isolated by U3 and U4.

[0034] By building a high-precision data acquisition unit inside a programmable arbitrary waveform generator, the voltage response of the battery is amplified by an instrumentation amplifier, and the amplified voltage signal and current response signal are synchronously acquired by an ADC to generate time-domain voltage and current response data; wherein the sampling rate is set to be greater than or equal to the highest sampling rate of the frequency component corresponding to the AC excitation signal.

[0035] The time-domain voltage and current response data are transmitted to the control chip MCU. The impedance spectrum analysis unit in the control chip MCU performs a fast Fourier transform on the time-domain voltage and current response data, converting the time-domain signal into a frequency-domain signal, and calculates the real and imaginary parts of the impedance at different frequencies. Based on the real and imaginary parts of the impedance at different frequencies, the electrochemical impedance spectrum of the battery during the charging and discharging process is generated.

[0036] In this embodiment of the invention, the miniature EIS measurement circuit is integrated inside the charger, forming an electrical connection with the charger's power conversion module and control chip MCU. Its programmable arbitrary waveform generator, high-precision data acquisition unit, and analog switches all employ low-power devices compatible with the control chip MCU. The overall circuit power consumption is controlled within 100mW, ensuring that it does not affect the charger's normal charging function. Simultaneously, it enables rapid measurement of the battery impedance spectrum within 1-3 seconds before charging, providing data support for battery health status diagnosis. The design process of the programmable arbitrary waveform generator involves selecting a 12-bit resolution DAC device. The control chip MCU sends a control signal to the DAC device, causing the DAC to output a digital signal corresponding to a pseudo-random binary sequence. The symbol rate of this pseudo-random binary sequence is set to 1kHz~1MHz, a range that can completely cover the main frequency range of the battery impedance response. This allows for the acquisition of wide-frequency impedance information in a single excitation process, eliminating the need for point-by-point frequency sweeping and significantly shortening the measurement time. The digital signal output from the DAC has an amplitude of 0~5V and is connected to a signal conditioning circuit composed of operational amplifier Q1. Operational amplifier Q1 is a high-input-impedance, low-offset-voltage operational amplifier. By adjusting the resistor divider network around operational amplifier Q1, the digital signal is amplified and converted into an AC excitation signal. Simultaneously, the amplitude of the AC excitation signal is strictly controlled between 20mV and 50mV. This amplitude range ensures that the excitation signal strength is sufficient to induce a significant impedance response in the battery, while preventing excessive signal amplitude from damaging the battery, thus ensuring that the battery is not affected during measurement. An analog switch, U2 chip, is connected in series between the battery and the power output terminal of the charger. Its control terminal is connected to the I / O port of the control chip MCU. The control chip MCU outputs high and low level signals to control the on / off state of U2 chip. U2 chip has both signal switching and power loop isolation functions, effectively blocking signal interference between the power loop and the measurement loop. When impedance measurement is required, the MCU control chip outputs a high-level signal, controlling chip U2 to switch to connection with the output of the programmable arbitrary waveform generator, connecting the signal-conditioned AC excitation signal to both ends of the battery. After the measurement is completed, the MCU control chip outputs a low-level signal, controlling chip U2 to switch to connection with the charger's power conversion module, restoring the normal charging circuit. Simultaneously, two isolation devices, U3 and U4, are connected to the programmable arbitrary waveform generator. U3 and U4 are optocouplers, which, together with chip U2, further enhance the isolation effect. Optical isolation completely isolates the measurement circuit containing the AC excitation signal from the charger's power circuit, preventing interference from high current and high voltage in the power circuit. This ensures the stability and accuracy of the measurement data, preventing fluctuations in the power circuit from affecting the amplitude and frequency of the AC excitation signal, as well as the acquisition accuracy of the voltage and current response signals.The high-precision data acquisition unit is directly integrated within the programmable arbitrary waveform generator (PARG), sharing power and control signals with it to reduce circuit redundancy. This unit includes an instrumentation amplifier and an ADC (Analog-to-Digital Converter). The instrumentation amplifier employs a differential input structure, with its inputs connected to the positive and negative terminals of the battery. It differentially amplifies the voltage response signal generated by the battery under AC excitation, with a magnification factor set between 100 and 1000 times. This factor can be flexibly adjusted according to the amplitude of the battery voltage response signal, ensuring that even weak voltage response signals are effectively amplified to meet the acquisition requirements of the ADC. The ADC is a 16-bit resolution model with an adjustable sampling rate. Its sampling rate is set to at least twice the highest value of the frequency component corresponding to the AC excitation signal. Specifically, when the symbol rate of the pseudo-random binary sequence is 1MHz, the ADC sampling rate is set to at least 2MHz to ensure complete acquisition of all frequency components of the voltage and current response signals, avoiding sampling distortion. During data acquisition, the ADC device simultaneously acquires the voltage response signal amplified by the instrumentation amplifier and the current response signal across the sampling resistor connected in series in the battery circuit. After converting the analog signals into digital signals, they are directly transmitted to the control chip MCU to generate continuous time-domain voltage and current response data. The length of the time-domain data is set to 1024 sampling points to ensure the accuracy of the subsequent Fourier transform. The control chip MCU integrates an impedance spectrum analysis unit. After receiving the time-domain voltage and current response data transmitted by the ADC, this unit first preprocesses the data to remove DC components and high-frequency noise. The preprocessing process is achieved by constructing an RC low-pass filter circuit with a filter cutoff frequency set to 10MHz, which can effectively filter out high-frequency noise generated by external interference while retaining the effective frequency components of the voltage and current response signals. After preprocessing, the impedance spectrum analysis unit performs a fast Fourier transform on the time-domain voltage and current response data, converting the time-domain signal into a frequency-domain signal. During the transform, the radix-2 FFT algorithm is used to convert the 1024 sampling points of time-domain data into 1024 frequency-domain data points, obtaining the frequency-domain amplitude and phase of voltage and current at different frequencies. According to the definition of impedance, the impedance magnitude at a given frequency is obtained by calculating the ratio of the voltage amplitude to the current amplitude at the same frequency. The impedance phase angle at that frequency is obtained by calculating the difference between the voltage phase and the current phase. Then, based on the impedance magnitude and phase angle, the real and imaginary parts of the impedance are obtained through trigonometric functions. The real part of the impedance equals the product of the impedance magnitude and the cosine of the phase angle, and the imaginary part equals the product of the impedance magnitude and the sine of the phase angle. The real and imaginary parts of the impedance at all frequencies are then processed to form the electrochemical impedance spectrum of the battery during charging and discharging. This impedance spectrum covers a frequency range of 1 kHz to 1 MHz and can fully reflect the battery's internal impedance characteristics, such as ohmic resistance, charge transfer impedance, and diffusion impedance.

[0037] Furthermore, the generation of impedance spectrum characteristic parameters in step S2 includes the following steps:

[0038] The electrochemical impedance spectrum is divided into frequency bands based on the frequency response characteristics of the battery's internal impedance. The electrochemical impedance spectrum is divided into high-frequency, mid-frequency and low-frequency regions. The high-frequency region corresponds to the ohmic internal resistance characteristics of the battery, the mid-frequency region corresponds to the charge transfer impedance and double-layer capacitance characteristics, and the low-frequency region corresponds to the diffusion impedance and Warburg impedance characteristics.

[0039] In the high-frequency region, the frequency segment corresponding to the extreme point of the impedance modulus is selected, and the real part of the impedance in the frequency segment is linearly fitted to obtain the characteristic value of the ohmic internal resistance. At the same time, the impedance phase angle of the frequency segment is extracted to verify whether the impedance characteristics in the high-frequency region conform to the ohmic internal resistance response law of the battery.

[0040] In the mid-frequency region, the frequency segment corresponding to the diameter of the semicircular arc in the impedance spectrum is selected. The intersection of the real and imaginary parts of the impedance in this frequency segment is analyzed. The characteristic value of the charge transfer impedance is obtained by fitting the equivalent circuit model. At the same time, the peak value of the impedance phase angle in this frequency segment is extracted to help verify the impedance characteristics in the mid-frequency region.

[0041] In the low-frequency region, a frequency band extending at a 45° angle in the impedance spectrum is selected. The difference between the real and imaginary parts of the impedance in this frequency band is linearly fitted to obtain the characteristic value of the diffusion impedance. At the same time, the characteristic value of the diffusion impedance is verified to be reasonable by combining the impedance phase angle variation law in the low-frequency region.

[0042] The extracted ohmic internal resistance, charge transfer impedance, and diffusion impedance characteristic values ​​are used as impedance spectrum characteristic parameters and stored in the corresponding temporary data buffer of the control chip.

[0043] In this embodiment of the invention, by dividing the electrochemical impedance spectrum into frequency bands, key impedance parameters that can reflect the health status of the battery are accurately extracted, providing data support for subsequent battery SOH estimation. The division and extraction processes are completed within the control chip MCU, without the need for additional external devices. First, the generated electrochemical impedance spectroscopy (EIS) was divided into frequency bands based on the frequency response characteristics of the battery's internal impedance. By analyzing the trends of the real and imaginary parts of the impedance spectrum with frequency, the EIS was clearly divided into high-frequency, mid-frequency, and low-frequency regions. The high-frequency region corresponds to a frequency range of 100 kHz to 1 MHz. The impedance characteristics in this band are mainly determined by the battery's ohmic internal resistance, as high-frequency signals can penetrate the battery's double layer and directly reflect the ohmic losses of the electrolyte, electrode materials, etc. The mid-frequency region corresponds to a frequency range of 1 kHz to 100 kHz. The impedance characteristics in this band are mainly determined by charge transfer impedance and double-layer capacitance. The charge transfer process occurs at the interface between the electrode and the electrolyte, and the charge-discharge characteristics of the double-layer capacitance are significant in this band. The low-frequency region corresponds to a frequency range below 1 kHz. The impedance characteristics in this band are mainly determined by diffusion impedance and Warburg impedance, primarily reflecting the diffusion process of lithium ions inside the battery. In the high-frequency region, a frequency range corresponding to the extreme points of the impedance magnitude is selected, ranging from 500kHz to 1MHz. Within this range, the impedance magnitude tends to be stable, and the extreme point fluctuations are small, accurately reflecting the ohmic internal resistance characteristics. Linear fitting is performed on all the real part impedance data within this frequency range. The fitting process is achieved by constructing a linear equation y = kx + b, where y represents the real part of the impedance, x represents the frequency, k is the fitting slope, and b is the fitting intercept. By calculating the sum of squared deviations between all the real part impedance data within this frequency range and the fitted line, the values ​​of k and b are adjusted to minimize the sum of squared deviations. At this point, the intercept b of the fitted line is the characteristic value of the ohmic internal resistance. Simultaneously, the impedance phase angles at all frequency points within this frequency band are extracted. The impedance phase angles within this band should approach 0°, as the phase angle of ohmic internal resistance is close to 0°. By determining whether the extracted phase angles are within the range of -5° to 5°, the impedance characteristics in the high-frequency region are verified to conform to the ohmic internal resistance response law of the battery. If the phase angle exceeds this range, a new high-frequency band is selected, and fitting and verification are performed again to ensure the accuracy of the ohmic internal resistance characteristic value. In the mid-frequency region, the frequency band corresponding to the diameter of the semicircular arc in the impedance spectrum is selected. This frequency band is selected as 10kHz to 50kHz. Within this frequency band, the impedance spectrum exhibits a distinct semicircular arc shape, and the diameter of the semicircular arc corresponds to the magnitude of the charge transfer impedance. The intersection points of the real and imaginary parts of the impedance within this frequency band are analyzed. First, the start and end points of the semicircular arc are determined. The start point corresponds to a higher frequency, and the end point corresponds to a lower frequency. The highest point of the semicircular arc is found, where the imaginary part of the impedance is largest. A straight line parallel to the real axis of the impedance is drawn through this highest point; the difference between this line and the two intersection points of the semicircular arc represents the diameter of the semicircular arc.The impedance data for this frequency band was fitted by constructing an equivalent circuit model. The equivalent circuit model consists of an ohmic internal resistance, a charge transfer impedance, and a double-layer capacitance connected in series. The extracted real and imaginary impedance data for this frequency band were substituted into the equivalent circuit model, and the values ​​of the charge transfer impedance and double-layer capacitance were adjusted through iterative calculations to minimize the deviation between the impedance data calculated by the model and the actual acquired impedance data. The resulting charge transfer impedance value is the characteristic value of the charge transfer impedance. Simultaneously, the peak value of the impedance phase angle within this frequency band was extracted. This peak value should be within the range of -45° to -60°. This peak value is used to help verify whether the impedance characteristics in the mid-frequency region conform to the response laws of charge transfer and double-layer capacitance, ensuring the reliability of the charge transfer impedance characteristic value. In the low-frequency region, a frequency band extending at a 45° angle in the impedance spectrum was selected. This frequency band is selected as 100Hz to 1kHz. In this frequency band, the real and imaginary impedance parts increase linearly with decreasing frequency, and the growth slope is close to 1, exhibiting a clear 45° angle characteristic, corresponding to the lithium-ion diffusion process. A linear fit is performed on the difference between the real and imaginary parts of the impedance within this frequency band. The fitting process also constructs a linear equation: y = kx + b, where y represents the difference between the real and imaginary parts of the impedance, x represents the reciprocal of the frequency, k is the fitting slope, and b is the fitting intercept. The values ​​of k and b are adjusted using the least squares method to minimize the deviation between the fitted line and the actual difference data. The slope k of the fitted line at this point is the characteristic value of the diffusion impedance. Simultaneously, considering the impedance phase angle variation pattern in the low-frequency region, the impedance phase angle within this frequency band should gradually approach -45°. By observing the trend of phase angle variation, the reasonableness of the extracted diffusion impedance characteristic value is determined. If the phase angle variation trend deviates from -45°, the frequency band of the low-frequency region is readjusted, and the fitting calculation is performed again to ensure that the diffusion impedance characteristic value accurately reflects the diffusion characteristics of lithium ions inside the battery. Finally, the extracted ohmic internal resistance, charge transfer impedance, and diffusion impedance characteristic values ​​are stored in the corresponding temporary data buffer of the control chip MCU according to a fixed data format. The capacity of the temporary data buffer is set to 1024 bytes, which can meet the storage requirements of the characteristic parameters and facilitate the subsequent use of these parameters by the control chip MCU to estimate the battery SOH and generate the charging curve.

[0044] Furthermore, the step of selecting a frequency band corresponding to the diameter of the semicircular arc in the impedance spectrum in the mid-frequency region and analyzing the intersection of the real and imaginary parts of the impedance in that frequency band includes the following steps:

[0045] Based on the frequency band corresponding to the semi-circular diameter of the impedance spectrum selected in the mid-frequency region, the real and imaginary parts of the impedance corresponding to each frequency point in the frequency band are extracted to generate a three-dimensional data set of frequency-real-imaginary part.

[0046] Based on the frequency-real-imaginary three-dimensional data set, an equivalent circuit physical model is constructed that includes charge transfer impedance, double-layer capacitance and solid-state diffusion impedance parameters. The solid-state diffusion impedance parameter in the equivalent circuit physical model is used as the associated coupling variable of charge transfer impedance.

[0047] Based on the equivalent circuit physical model, a nonlinear complex plane least squares fitting is performed on the frequency-real-imaginary three-dimensional data set. During the fitting process, the solid-phase diffusion impedance parameter is introduced as a constraint on the charge transfer impedance, and the initial charge transfer impedance estimate is obtained by solving the problem. The peak value of the impedance phase angle at the characteristic frequency in the mid-frequency region of the impedance spectrum is obtained. The initial charge transfer impedance estimate and the peak value of the impedance phase angle are coupled together to calculate and generate the verification correction factor of the charge transfer impedance.

[0048] The verification correction factor is convolved with the initial charge transfer impedance estimate to generate the final charge transfer impedance characteristic value, and the frequency and phase angle information corresponding to the characteristic value are recorded.

[0049] In this embodiment of the invention, the process of selecting the frequency band corresponding to the diameter of the semicircular arc in the impedance spectrum in the mid-frequency region and analyzing the intersection of the real and imaginary parts of the impedance in this frequency band is entirely completed within the control chip MCU. Data extraction, model construction, fitting calculation, and correction verification are achieved by relying on the MCU's computing unit, ensuring the accurate extraction of charge transfer impedance characteristic values. Simultaneously, connecting with the mid-frequency region division results, 10kHz~50kHz is selected as the frequency band corresponding to the diameter of the semicircular arc. Within this frequency band, the semicircular arc characteristic of the impedance spectrum is obvious, accurately reflecting the true characteristics of charge transfer impedance. First, based on this frequency band, the real and imaginary parts of the impedance corresponding to each frequency point are extracted. Frequency points are selected at 100Hz intervals, resulting in 400 frequency points. Each frequency point corresponds to a unique real and imaginary impedance value. These data are organized in ascending order of frequency to generate a three-dimensional data set of frequency-real-imaginary part. Each data set contains a frequency value, the corresponding real impedance value, and the imaginary impedance value, providing basic data for subsequent model construction and fitting calculation. Based on the generated frequency-real-imaginary three-dimensional data set, an equivalent circuit physical model is constructed, including parameters such as charge transfer impedance, double-layer capacitance, and solid-phase diffusion impedance. This model is based on the charge transfer process and solid-phase diffusion process at the interface between the battery electrode and the electrolyte. It consists of ohmic internal resistance, charge transfer impedance, double-layer capacitance, and solid-phase diffusion impedance connected in series. The ohmic internal resistance is based on the extracted ohmic internal resistance characteristic value, the initial value of the double-layer capacitance is set to 10μF~100μF, and the initial value of the solid-phase diffusion impedance parameter is set to 10Ω~100Ω. The solid-phase diffusion impedance parameter in the equivalent circuit physical model is used as the associated coupling variable of the charge transfer impedance. Because the solid-phase diffusion process affects the rate and efficiency of charge transfer, and the two have a close coupling relationship, associating the solid-phase diffusion impedance parameter with the charge transfer impedance can improve the accuracy of model fitting and avoid errors caused by fitting a single parameter. Based on the constructed equivalent circuit physical model, a nonlinear complex plane least squares fitting is performed on the three-dimensional data set of frequency-real-imaginary parts. The fitting process is iteratively implemented by the computing unit of the control chip MCU. First, each data group in the three-dimensional data set is substituted into the equivalent circuit physical model to calculate the theoretical values ​​of the real and imaginary parts of the impedance output by the model. Then, the sum of squares of the deviations between the theoretical values ​​and the actual acquired values ​​is calculated. By iteratively adjusting the values ​​of charge transfer impedance, double-layer capacitance, and solid-state diffusion impedance parameters, the sum of squares of the deviations is continuously reduced until it is less than a preset threshold of 10. -6 Ω 2During the fitting process, a constraint condition for the charge transfer impedance is introduced using the solid-phase diffusion impedance parameter. This constraint is set so that the charge transfer impedance value is no greater than 5 times the solid-phase diffusion impedance parameter value and no less than 0.2 times the solid-phase diffusion impedance parameter value. This constraint is based on the inherent correlation between battery charge transfer and solid-phase diffusion, conforming to the electrochemical reaction rules within the battery. This constraint avoids unreasonable parameter values ​​during the fitting process, thus obtaining the initial estimated charge transfer impedance value. Subsequently, the peak value of the impedance phase angle at the characteristic frequency in the mid-frequency region of the impedance spectrum is obtained. The characteristic frequency is selected as the frequency corresponding to the highest point of the semi-circular arc, where the phase angle change is most significant. The peak value of the impedance phase angle corresponding to this frequency is extracted, and this peak value is controlled between -45° and -60°. The initial estimated charge transfer impedance value and the peak value of the impedance phase angle are coupled together to generate a verification correction factor for the charge transfer impedance. The coupling calculation process is achieved by constructing a coupling equation. The initial value of the verification correction factor is obtained by multiplying the initial estimated charge transfer impedance value by the cosine value of the phase angle peak value, laying the foundation for subsequent correction processing. The verification correction factor is convolved with the initial charge transfer impedance estimate. The convolution operation is implemented by the computing module inside the control chip MCU. During the operation, the frequency is used as the independent variable. The verification correction factor and the initial charge transfer impedance estimate are convolved point by point according to the frequency correspondence to eliminate the error caused by the calculation of a single parameter and generate the final charge transfer impedance characteristic value. At the same time, the frequency and phase angle information corresponding to the characteristic value are recorded. The frequency is the frequency corresponding to the midpoint of the semicircular arc diameter, and the phase angle is the impedance phase angle corresponding to the frequency. The recorded data is stored in the temporary data buffer of the control chip MCU and stored in association with the extracted ohmic internal resistance and diffusion impedance characteristic values. This provides complete impedance characteristic parameter support for subsequent battery SOH estimation and adjustable charging curve generation.

[0050] Furthermore, the verification correction factor for generating charge transfer impedance includes the following steps:

[0051] The initial charge transfer impedance estimate and the peak value of the impedance phase angle are obtained. Based on the charge transfer dynamics theory, the initial charge transfer impedance estimate is mapped to an equivalent charge exchange current density to generate charge exchange current density data.

[0052] The solid-phase diffusion rate at the electrode-electrolyte interface is obtained, and the amplitude phase hysteresis of the charge exchange current density at the solid-phase diffusion rate is calculated by combining the charge exchange current density data.

[0053] Extract the actual impedance imaginary part amplitude corresponding to the characteristic frequency point from the frequency-real-imaginary part three-dimensional data set, and perform cross-correlation analysis between it and the amplitude phase lag to generate the dynamic coupling coefficient between the charge transfer process and the solid-phase diffusion process.

[0054] The dynamic coupling coefficient is substituted into the preset correction factor generation model to perform frequency response convolution inversion and generate impedance verification correction factor; the impedance verification correction factor is subjected to frequency domain weighted filtering based on the ratio between the amplitude phase lag and the actual impedance imaginary part amplitude to filter out error components introduced by high frequency noise or low frequency drift and generate charge transfer impedance verification correction factor.

[0055] In this embodiment of the invention, by obtaining the initial charge transfer impedance estimate and the peak value of the impedance phase angle, the initial charge transfer impedance estimate is mapped to an equivalent charge exchange current density based on charge transfer kinetics theory. This mapping process is achieved by constructing a mapping equation. The charge exchange current density is equal to the product of the Faraday constant, the electrode area, and the reciprocal of the initial charge transfer impedance estimate. The Faraday constant is taken as 96500 C / mol, and the electrode area is set according to the battery's nominal electrode area. The calculated charge exchange current density data is controlled within the range of 1 mA / cm². 2 ~10mA / cm 2 This reflects the charge transfer rate at the electrode-electrolyte interface. The solid-phase diffusion rate at the electrode-electrolyte interface is obtained by analyzing low-frequency impedance data. The difference between the real and imaginary parts of the impedance in the 100Hz–1kHz frequency range is selected, and the diffusion coefficient is obtained through linear fitting. Based on the correlation between the diffusion coefficient and the solid-phase diffusion rate, the solid-phase diffusion rate is calculated, with the rate range controlled within 10... -10 m 2 / s~10 -8 m 2 / s, combined with charge exchange current density data, the amplitude phase lag of the charge exchange current density at the solid-state diffusion rate is calculated. The solution process is achieved by constructing a phase lag equation. The solid-state diffusion rate and charge exchange current density are substituted into the equation to calculate the amplitude phase lag. The lag is controlled within the range of 5°~15°, reflecting the degree of influence of the solid-state diffusion process on the amplitude and phase of the charge transfer current. The actual impedance imaginary amplitude corresponding to the characteristic frequency points is extracted from the three-dimensional data set of frequency-real-imaginary parts. The characteristic frequency points are selected as five key frequencies in the mid-frequency range: 10kHz, 20kHz, 30kHz, 40kHz, and 50kHz. The impedance imaginary part value corresponding to each frequency point is extracted as the actual impedance imaginary amplitude. Cross-correlation analysis is performed on it and the amplitude phase lag. The cross-correlation analysis is achieved by calculating the correlation coefficient between the actual impedance imaginary amplitude and the amplitude phase lag. The correlation coefficient is calculated by dividing the covariance by the product of the standard deviations of the two, and the correlation coefficient ranges from 0.8 to 0.95. Based on the correlation coefficient, the dynamic coupling coefficient between the charge transfer process and the solid-state diffusion process is generated. The coupling coefficient is equal to the product of the correlation coefficient and the amplitude phase lag, and the range is controlled between 0.07 and 0.14. The dynamic coupling coefficients are substituted into a preset correction factor generation model for frequency response convolution inversion. This model, based on frequency response theory, is used for the convolution inversion process, which is iteratively implemented by the MCU's computational unit. After substituting the dynamic coupling coefficients into the model, the frequency response signal is inverted to generate an impedance verification correction factor. The impedance verification correction factor is then subjected to frequency-domain weighted filtering based on the ratio between the amplitude-phase lag and the amplitude of the imaginary part of the actual impedance. This filtering is achieved by constructing a frequency-domain filter, where the weighting coefficients are equal to the ratio of the amplitude-phase lag to the amplitude of the imaginary part of the actual impedance. The impedance verification correction factor is weighted at each frequency point to filter out error components introduced by high-frequency noise or low-frequency drift, keeping these error components within 5%. This results in a final charge transfer impedance verification correction factor, ranging from 0.9 to 1.1, effectively correcting deviations in the initial charge transfer impedance estimate and ensuring that the final charge transfer impedance characteristic value accurately reflects the actual health status of the battery.

[0056] Furthermore, the fitting estimation of the corresponding battery health status assessment value in step S2 includes the following steps:

[0057] Multiple groups of batteries of the same model were selected for accelerated aging tests. Each group of batteries was aged under the same ambient temperature and charge-discharge regime. The electrochemical impedance spectrum of the battery was collected every certain number of cycles and the characteristic parameters of the impedance spectrum were extracted. At the same time, the remaining capacity, capacity decay rate and internal resistance growth rate of each group of batteries were recorded as indicators of battery health status.

[0058] The impedance spectrum characteristic parameters of each group of batteries were correlated with the battery health status characterization index, and the mapping relationship between the characteristic parameters and the battery health status was established by statistical methods.

[0059] The established mapping relationship is stored in the Flash memory of the control chip to form a mapping table between feature parameters and battery health status. This table includes the threshold range of the feature parameters and the corresponding battery health status level, as well as the correspondence between the trend of feature parameter change and the trend of battery health status change.

[0060] Once the impedance spectrum characteristic parameters of the battery are obtained, the corresponding battery health status level is found in the mapping table between characteristic parameters and battery health status. Based on the battery health status level, the battery health status assessment value corresponding to the current impedance spectrum characteristic parameters is calculated, including the remaining capacity estimate, the internal resistance growth rate estimate, and the capacity decay trend estimate.

[0061] In this embodiment of the invention, 20 groups of batteries of the same model were selected for accelerated aging test. The test environment temperature was controlled at 25℃±2℃. The charge and discharge regime was set to constant current and constant voltage mode. The charging current was 1C, the constant voltage was 4.2V, the cut-off current was 0.05C, the discharge current was 0.5C, and the discharge cut-off voltage was 3.0V. Each group of batteries was cycled and aged according to this charge and discharge regime until the remaining capacity of the battery dropped to 80% of the initial capacity and the test was stopped. During the aging process, the electrochemical impedance spectroscopy (EIS) of the battery is collected every 50 charge-discharge cycles. The collection method is consistent and is completed through the micro EIS measurement circuit integrated inside the charger. After each collection, the ohmic internal resistance, charge transfer impedance, and diffusion impedance characteristic values ​​of the battery group are extracted. At the same time, the remaining capacity, capacity decay rate, and internal resistance growth rate of each battery group are recorded through constant current charge-discharge tests. The remaining capacity is calculated by the total discharge capacity when discharged to the cutoff voltage. The capacity decay rate is equal to the ratio of (initial capacity - current remaining capacity) to the initial capacity. The internal resistance growth rate is equal to the ratio of (current ohmic internal resistance - initial ohmic internal resistance) to the initial ohmic internal resistance. The three together serve as indicators of battery health status. Each battery group corresponds to a complete set of impedance spectrum characteristic parameters and health status characterization index data. Correlation analysis was conducted on the impedance spectrum characteristic parameters and battery health status indicators of 20 battery groups. This correlation analysis was performed using external testing and computing equipment controlled by the chip. The ohmic internal resistance, charge transfer impedance, and diffusion impedance of each battery group were used as independent variables, while remaining capacity, capacity decay rate, and internal resistance growth rate were used as dependent variables. The correlation strength was determined by calculating the correlation coefficients between each impedance characteristic parameter and each health status indicator. Specifically, the correlation coefficients between ohmic internal resistance, charge transfer impedance, and internal resistance growth rate were controlled above 0.9; the correlation coefficient between diffusion impedance and capacity decay rate was controlled above 0.85; and the correlation coefficient between ohmic internal resistance and remaining capacity was controlled above -0.88. A mapping relationship between characteristic parameters and battery health status was established using statistical methods, combining linear regression and nonlinear fitting. A linear regression equation was constructed for the linear correlation between ohmic internal resistance and remaining capacity and internal resistance growth rate. A quadratic polynomial fitting equation was constructed for the nonlinear correlation between charge transfer impedance, diffusion impedance, and capacity decay rate. The equation parameters were iteratively adjusted by substituting 20 sets of experimental data to ensure that the sum of squares of the deviations between the calculated values ​​and the experimentally measured values ​​was less than 10. -3This process ultimately yields a complete mapping relationship, which accurately reflects the intrinsic correlation between changes in impedance spectrum characteristic parameters and changes in battery health status. The established mapping relationship is stored in the Flash memory of the control chip, with a capacity of 16MB and 4MB reserved for storing the mapping relationship, forming a mapping table between characteristic parameters and battery health status. The mapping table contains the threshold ranges corresponding to the feature parameters and the corresponding battery health status levels. The health status levels are divided into three levels: Level 1 is a healthy state (remaining capacity ≥ 95%, capacity decay rate ≤ 5%, internal resistance growth rate ≤ 10%), corresponding to ohmic internal resistance ≤ 50mΩ, charge transfer impedance ≤ 100mΩ, and diffusion impedance ≤ 150mΩ; Level 2 is a sub-healthy state (80% ≤ remaining capacity < 95%, 5% < capacity decay rate ≤ 20%, 10% < internal resistance growth rate ≤ 30%), corresponding to 50mΩ < ohmic internal resistance ≤ 80mΩ, 100mΩ < charge transfer impedance ≤ 200mΩ, and 150mΩ < diffusion impedance ≤ 300mΩ; Level 3 is an unhealthy state (remaining capacity < 80%, capacity decay rate > 20%, internal resistance growth rate > 30%), corresponding to ohmic internal resistance > 80mΩ, charge transfer impedance > 200mΩ, and diffusion impedance > 300mΩ. The mapping table simultaneously contains the correspondence between the changing trends of characteristic parameters and the changing trends of battery health status. Specifically, when the ohmic internal resistance, charge transfer impedance, and diffusion impedance increase by more than 5% per month, the battery health status level will decrease by one level; when the remaining capacity decreases by more than 3% per month, the capacity decay rate level will increase accordingly. After obtaining the battery's impedance spectrum characteristic parameters, the control chip's computing unit calls the mapping table in the Flash memory. Based on the extracted ohmic internal resistance, charge transfer impedance, and diffusion impedance characteristic values, it looks up the corresponding battery health status level, and then calculates the battery health status assessment value corresponding to the current impedance spectrum characteristic parameters based on this level. The fitting calculation is achieved by substituting into the regression equation and fitting equation in the mapping relationship to calculate the estimated remaining capacity, estimated internal resistance growth rate, and estimated capacity decay trend. The estimated capacity decay trend is obtained by fitting the capacity decay amount for the next month using the difference in capacity decay rate calculated from the impedance characteristic parameters collected in the last three times. All assessment values ​​are stored in the control chip's temporary data buffer to provide data support for the subsequent generation of charging curve parameters.

[0062] Furthermore, refer to Figure 3 , Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3 in this embodiment. Step S3 in this embodiment includes the following steps:

[0063] S31: A battery health status perception model is established based on a local lightweight learning algorithm, including a battery health status input interface, a state of charge input interface, a charging curve parameter output interface, and a model parameter update interface. The battery health status input interface receives the battery health status evaluation value, and the state of charge input interface receives the current state of charge value of the battery.

[0064] S32: Set the mapping rules between battery health status and charging curve parameters within the model, and set the corresponding upper limit of charging current, upper limit of charging voltage and charging cut-off conditions according to different levels of battery health status assessment value and different ranges of battery state of charge.

[0065] S33: Adjust the adjustment step size and adjustment frequency of the charging curve parameters in real time according to the changing trend of the battery health status assessment value and the changing rate of the battery state of charge, so as to generate the corresponding dynamic adjustment rules; compile the mapping rules and dynamic adjustment rules into program code that can run on the control chip, and load it into the control logic of the battery health status perception model to generate the corresponding charging curve parameters according to the input battery health status assessment value and battery state of charge value.

[0066] S34: Generate adjustable charging curves based on charging curve parameters, including curves of charging current changing with time, charging voltage changing with time, and charging cutoff conditions changing with time. The adjustable charging curves together define the charging strategy for the battery under different health and charge states.

[0067] In this embodiment of the invention, a battery health status perception model is established based on a local lightweight learning algorithm. The local lightweight learning algorithm employs a combination of linear regression and gradient descent, with the computational complexity controlled to within 10. 6Within a speed of less than 1000 times per second, the model is adapted to the computing power of the control chip and includes four core interfaces: The battery health status input interface receives the fitted battery health status assessment values, including estimated remaining capacity, estimated internal resistance growth rate, and estimated capacity decay trend. The data transmission rate of this interface is set to 100kbps to ensure real-time data transmission; the state of charge (SOC) input interface receives the current SOC value of the battery, which is calculated by integrating the battery voltage and current. This calculation process is completed by the control chip's computing unit, and the SOC value is updated and transmitted to this interface every 100ms; the charging curve parameter output interface outputs parameters such as the generated upper limit of charging current, upper limit of charging voltage, and charging cutoff conditions, with the output frequency consistent with the SOC update frequency; and the model parameter update interface receives parameter adjustment signals during subsequent closed-loop learning and optimization processes, enabling dynamic optimization of the model. The model internally sets mapping rules between battery health status and charging curve parameters. These rules are based on the battery health status level and state of charge (SOC) range, which is divided into three ranges: low SOC (0%~30%), medium SOC (30%~80%), and high SOC (80%~100%). For batteries in Level 1 health status, the upper limit of charging current for low SOC is set to 1.2C, for medium SOC to 1C, and for high SOC to 0.3C. The upper limit of charging voltage is set to 4.2V, and the charging cutoff condition is set to the charging current dropping to 0.05C for 30 seconds. For batteries in Level 2 sub-health status, the upper limit of charging current for low SOC is set to 0.8C, for medium SOC to 0.6C, and for high SOC to 0.2C. The upper limit of charging voltage is set to... The voltage is set to 4.15V, and the charging cutoff condition is set to the charging current dropping to 0.03C for 30 seconds. For batteries in a level 3 unhealthy state, the upper limit of the charging current for low charge state is set to 0.5C, the upper limit of the charging current for medium charge state is set to 0.3C, and the upper limit of the charging current for high charge state is set to 0.1C. The upper limit of the charging voltage is set to 4.1V, and the charging cutoff condition is set to the charging current dropping to 0.02C for 30 seconds. Here, C is the current value corresponding to the nominal capacity of the battery, ensuring that batteries in different health states can obtain suitable charging parameters.Adjust the step size and adjustment frequency of the charging curve parameters in real time according to the change trend of the battery health status evaluation value and the change rate of the battery state of charge. The setting of the step size and adjustment frequency is based on the speed of battery state change. When the estimated value of the capacity attenuation trend exceeds 3% per month and the internal resistance growth rate exceeds 5% per month, the step size of the charging curve parameter is set to 0.05C, and the adjustment frequency is set to 50 ms / time to accelerate the parameter adjustment speed and avoid further damage to the battery. When the estimated value of the capacity attenuation trend is between 1% and 3% per month and the internal resistance growth rate is between 2% and 5% per month, the step size is set to 0.1C, and the adjustment frequency is set to 100 ms / time. When the estimated value of the capacity attenuation trend is less than 1% per month and the internal resistance growth rate is less than 2% per month, the step size is set to 0.2C, and the adjustment frequency is set to 200 ms / time to reduce the parameter adjustment frequency and ensure charging stability. Compile the mapping rules and dynamic adjustment rules into program code that can run on the control chip. The compilation process is realized through the compilation tool supporting the control chip, converting the mapping rules and dynamic adjustment rules into machine instructions recognizable by the control chip, and loading them into the control logic of the battery health status perception model. After the control logic receives the data transmitted by the battery health status input interface and the state of charge input interface, it calls the compiled instructions to calculate and generate the corresponding charging curve parameters to ensure the accurate matching of the parameters with the current battery state. Generate an adjustable charging curve based on the charging curve parameters. The charging curve consists of three sub-curves, namely the curve of charging current changing with time, the curve of charging voltage changing with time, and the curve of charging cut-off conditions changing with time. The curve of charging current changing with time is generated according to the state of charge interval and the step size. In the low state of charge stage, the current remains stable at the current upper limit of the corresponding interval. In the middle state of charge stage, the current gradually decreases according to the step size. In the high state of charge stage, the current drops to the minimum value of the corresponding interval and remains until the charging cut-off. The curve of charging voltage changing with time adopts a constant voltage mode, gradually rising to the corresponding voltage upper limit in the low state of charge and middle state of charge stages, and remaining stable at the voltage upper limit in the high state of charge stage. The curve of charging cut-off conditions changing with time is set according to the battery health status level. The better the health status, the higher the cut-off current, and the cut-off duration is fixed at 30 seconds. The three adjustable charging curves jointly define the charging strategy of the battery under different health statuses and states of charge. The curve data is transmitted to the power conversion module of the charger in real time to drive the power conversion module to perform the charging operation according to the curve parameters, realizing adaptive charging based on battery health status perception.

[0068] Further, generating the adjustable charging curve based on the charging curve parameters includes the following steps:

[0069] Obtain the set of output charging curve parameters, which includes charging current parameters, charging voltage parameters and charging cut-off condition parameters, and use this set of charging curve parameters as the initial charging curve control parameter vector, wherein the charging cut-off condition parameters include cut-off voltage, cut-off current and cut-off temperature threshold.

[0070] Based on the current state of charge-internal resistance change gradient data of the battery, the control parameter vector of the initial charging curve is dynamically adjusted to generate a charging current instantaneous adjustment coefficient vector and a charging voltage instantaneous adjustment coefficient vector containing the time dimension; based on the adjustment coefficient vector and combined with the entropy change rate data of the internal electrochemical reaction of the battery, a dynamic response constraint of charging current and charging voltage with time as the independent variable is constructed, wherein the entropy change rate is also introduced as a state constraint variable.

[0071] The real-time polarization voltage data of the battery during the charging process is acquired, and the real-time polarization voltage data is fed back to the dynamic response constraint iterative calculation to generate a continuous smooth curve of charging current changing with time and a continuous smooth curve of charging voltage changing with time.

[0072] Based on the changing trends of the state of charge-internal resistance change gradient data and entropy change rate data, the charging cutoff condition parameters are dynamically corrected to generate an adaptive charging cutoff condition curve that changes synchronously with the charging current curve and the charging voltage curve. The adaptive charging cutoff condition curve defines the charging termination threshold at different time points and together with the aforementioned two curves, constitutes a complete adjustable charging curve.

[0073] In this embodiment of the invention, an output charging curve parameter set is obtained. This parameter set includes charging current parameters, charging voltage parameters, and charging cutoff condition parameters. The charging current parameter represents the upper limit of current and adjustment step size for each state of charge interval. The charging voltage parameter represents the upper limit of voltage corresponding to each healthy state. The charging cutoff condition parameters include cutoff voltage, cutoff current, and cutoff temperature threshold. The cutoff voltage corresponds to the charging termination voltage for each healthy state: 4.2V for level one, 4.15V for level two, and 4.1V for level three. The cutoff current corresponds to the termination current for each healthy state: 0.05C for level one, 0.03C for level two, and 0.02C for level three. The cutoff temperature threshold is uniformly set to 55℃. This charging curve parameter set is used as the initial charging curve control parameter vector. The parameter vector is arranged in the order of current, voltage, and cutoff condition, forming a three-dimensional parameter vector, providing an initial reference for subsequent dynamic adjustments. Based on the current state of charge (SOC)-internal resistance change gradient data of the battery, which is calculated by collecting SOC and internal resistance values ​​every 100ms, the gradient value is equal to the ratio of the difference in internal resistance between two adjacent measurements to the difference in SOC, with a range controlled between 0.5mΩ / % and 2mΩ / %. The initial charging curve control parameter vector is dynamically adjusted. The adjustment process is achieved iteratively through the control chip's computing unit. The gradient data is substituted into the adjustment equation to generate instantaneous adjustment coefficient vectors for charging current and charging voltage, which contain the time dimension. The time interval of the adjustment coefficient vectors is consistent with the SOC update frequency, which is 100ms. The current adjustment coefficient ranges from 0.9 to 1.1, and the voltage adjustment coefficient ranges from 0.98 to 1.02. Based on the adjustment coefficient vector and combined with the entropy change rate data of the internal electrochemical reaction of the battery, the entropy change rate data is calculated by the entropy change coefficient calibrated by the aging test and the current battery temperature, ranging from 0.01 J / (mol·K·s) to 0.05 J / (mol·K·s). A dynamic response constraint for charging current and charging voltage with time as the independent variable is constructed. The constraint equation is constructed based on electrochemical reaction kinetics. The adjustment coefficient vector and entropy change rate data are substituted into the equation, and the entropy change rate is introduced as a state constraint variable to limit the rate of change of current and voltage. The rate of change of current does not exceed 0.1 C / s, and the rate of change of voltage does not exceed 0.05 V / s.Real-time polarization voltage data of the battery during charging is acquired. The polarization voltage data is obtained by collecting the difference between the real-time voltage at both ends of the battery and the open-circuit voltage. The open-circuit voltage is obtained by measuring the voltage after resting for 5 minutes before each charge. Real-time polarization voltage data is collected every 50ms and fed back to the dynamic response constraint iterative calculation. The iterative calculation is completed by the control chip's computing unit. In each iteration, the polarization voltage data is substituted into the dynamic response constraint to adjust the instantaneous values ​​of current and voltage, so that the polarization voltage is controlled between 0.05V and 0.2V. Finally, a continuous smooth curve of charging current changing with time and a continuous smooth curve of charging voltage changing with time are generated. The time resolution of the curve is 100ms to ensure that there are no abrupt changes in the curve and to avoid sudden changes in battery stress during charging. Based on the changing trends of the state of charge-internal resistance gradient data and the entropy change rate data, which are obtained by calculating the difference between the last 10 data collections, when the gradient data and entropy change rate data continue to rise, it indicates that the internal reaction of the battery is intensifying. The charging cutoff condition parameters are dynamically corrected. The correction process is achieved by constructing a correction equation. The difference in the changing trends is substituted into the equation to adjust the cutoff current, cutoff voltage, and cutoff temperature threshold, so that the cutoff current is reduced, the cutoff voltage is lowered, and the cutoff temperature threshold remains unchanged. An adaptive charging cutoff condition curve that changes synchronously with the charging current curve and the charging voltage curve is generated. This curve uses time as the independent variable to define the charging termination threshold at different time points. Together with the two curves mentioned above, it forms a complete adjustable charging curve. The curve data is transmitted to the power conversion module in real time to guide the charging operation.

[0074] Furthermore, the generation of the adaptive curve for the charging cutoff condition includes the following steps:

[0075] The correlation coefficient between the state of charge-internal resistance change gradient data and the entropy change rate data within the sliding time window is calculated to generate a characteristic coupling coefficient characterizing the internal dynamic-thermodynamic coupling state of the battery. Based on the characteristic coupling coefficient, the charging cutoff condition parameters are normalized and corrected, and each corrected cutoff condition parameter is multiplied by the exponential decay constraint of the characteristic coupling coefficient to obtain the first round of cutoff condition parameter adjustment values.

[0076] Obtain the first derivative of the charging current versus time curve within a set time interval under the current charging stage, and perform a convolution operation on the first derivative with the feature coupling coefficient to generate a dynamic adjustment factor for the influence of current change on the cutoff condition.

[0077] The dynamic adjustment factor is weighted and fused with the first round of cutoff condition parameter adjustment value, wherein the weight is dynamically determined by the second derivative of the electrochemical reaction entropy change rate data to generate the second round of cutoff condition parameter adjustment value.

[0078] Based on the second round of cutoff condition parameter adjustment values, and combined with the slope of the charging voltage versus time curve at the current moment, a continuous relationship between the cutoff condition parameters and time is constructed. With time as the independent variable, the cutoff voltage threshold curve, cutoff current threshold curve, and cutoff temperature threshold curve are output, which together constitute the adaptive curve of the charging cutoff condition.

[0079] In this embodiment of the invention, the correlation coefficient between the gradient data of the state of charge-internal resistance change and the entropy change rate data within a sliding time window is calculated. The sliding time window is set to 10 seconds and contains 100 sets of data. The correlation coefficient is obtained by calculating the product of the covariance of the two data points and their respective standard deviations, and is controlled within the range of 0.8 to 0.95. This coefficient characterizes the degree of correlation between the state of charge-internal resistance change and the entropy change rate change, generating a characteristic coupling coefficient characterizing the internal dynamic-thermodynamic coupling state of the battery. The characteristic coupling coefficient is equal to the product of the correlation coefficient and the entropy change rate data, and ranges from 0.0008 to 0.00475. The charging cutoff condition parameters are normalized and corrected based on the characteristic coupling coefficient. The normalization correction is achieved by dividing each cutoff condition parameter by its initial maximum value and then multiplying it by the characteristic coupling coefficient to obtain the first round of cutoff condition parameter adjustment values. The cutoff voltage adjustment value ranges from 4.0 to 4.2V, the cutoff current adjustment value ranges from 0.01 to 0.05C, and the cutoff temperature threshold remains unchanged at 55℃. The corrected cutoff condition parameters are then multiplied by the exponential decay constraint of the characteristic coupling coefficient, which is set to 0.99 and decays once every 100ms to obtain the first round of cutoff condition parameter adjustment values. The first derivative of the charging current versus time curve during the current charging phase is obtained within a set time interval of 1 second, containing 10 sets of current data. The first derivative is calculated by dividing the difference between two adjacent current values ​​by the time interval, reflecting the rate of current change. This first derivative is then convolved with a feature coupling coefficient. The convolution operation is achieved by multiplying point by point and then summing, generating a dynamic adjustment factor for the impact of current change on the cutoff condition. The adjustment factor ranges from 0.005 to 0.02; the faster the current change, the larger the adjustment factor, and the more significant the impact on the cutoff condition. The dynamic adjustment factor is then weighted and fused with the first round of cutoff condition parameter adjustments. The weights are dynamically determined by the second derivative of the entropy change rate data of the electrochemical reaction. The second derivative of the entropy change rate is calculated by dividing the difference between two adjacent entropy change rate data by the time interval, ranging from -0.001 to 0.001 J / (mol·K·s). 2When the second derivative is positive, the weight increases; when it is negative, the weight decreases. Weighted fusion is achieved by multiplying the adjustment factor by the weight and adding the first-round adjustment value multiplied by (1-weight) to generate the second-round cutoff condition parameter adjustment value, ensuring that the adjustment value is more in line with the real-time state of the battery. Based on the second-round cutoff condition parameter adjustment value, combined with the slope of the charging voltage versus time curve at the current moment (the slope is calculated by dividing the difference between two adjacent voltage values ​​by the time interval, ranging from -0.01 to 0.01V / ms), a continuous relationship between the cutoff condition parameters and time is constructed. With time as the independent variable, a cutoff voltage threshold curve, a cutoff current threshold curve, and a cutoff temperature threshold curve are generated through linear interpolation. The cutoff temperature threshold curve remains at 55℃, the cutoff voltage threshold curve gradually decreases over time, and the cutoff current threshold curve gradually decreases over time. The three curves together constitute the charging cutoff condition adaptive curve, which changes synchronously with the generated current and voltage curves. This ensures that during the charging process, when the battery state reaches the corresponding cutoff threshold at the time point, charging termination or adjustment is triggered in a timely manner, avoiding battery overcharging damage and ensuring charging safety and battery life.

[0080] Furthermore, the local lightweight learning algorithm can be trained using a decision tree algorithm or a small-scale neural network. The step of using a small-scale neural network to train and construct the battery health status perception model includes:

[0081] A small-scale feedforward neural network model is constructed. Its input layer nodes receive battery health status assessment values, current state of charge, and current battery internal temperature. Its hidden layer adopts a gated unit structure with a self-attention mechanism to dynamically adjust the contribution weights of different input features in decision-making. Historical charging optimization datasets are loaded from the storage unit of the control chip. These datasets contain battery performance data obtained under different health status, state of charge, and temperature conditions using different charging curve parameters. These datasets serve as training samples and supervision labels for the neural network.

[0082] During the charging interval, the feedforward neural network model is trained online using the historical charging optimization dataset to generate a battery health status perception model. Knowledge distillation loss is introduced during the training process, and the prior mapping rules in the battery health status perception model are used as soft labels to constrain the parameter update direction of the neural network, so as to obtain the pre-trained network weight parameters.

[0083] The pre-trained network weight parameters are convolved and fused with the real-time collected battery state of charge-internal resistance change gradient data to generate a dynamic weight bias vector for the current battery state. This vector is then superimposed on the output layer of the battery health state perception model to complete the online fine-tuning of the model in the current charging scenario.

[0084] The fine-tuned model is deployed in the real-time inference unit of the control chip. The current battery health status assessment value, state of charge and internal temperature are input into the model. Through the forward propagation calculation of the model, the charging curve parameter set consisting of the current optimal charging current, charging voltage and charging cut-off condition is directly output. The charging curve parameter set is then used to generate an adjustable charging curve.

[0085] In this embodiment of the invention, a small-scale feedforward neural network model is constructed, with the total number of parameters controlled to within 1000, to adapt to the computing power of the control chip and avoid response delay caused by excessive computation. The model input layer has three nodes, which respectively receive the battery health status assessment value, the current state of charge, and the current battery internal temperature. The battery health status assessment value is a fusion of the fitted remaining capacity estimate, internal resistance growth rate estimate, and capacity decay trend estimate, obtained by weighted summation of the three values, with weights set to 0.6, 0.25, and 0.15 respectively. The current state of charge is calculated by integrating the battery voltage and current, updated every 100ms, with a range controlled between 0% and 100%. The current battery internal temperature is measured by a temperature acquisition element embedded inside the battery, with a measurement accuracy controlled within ±1℃ and a range controlled between 0℃ and 60℃. All three input features are normalized and converted into values ​​between 0 and 1 before being input to the input layer nodes. The model has two hidden layers, each containing 16 neurons, employing a gating unit structure with a self-attention mechanism. The gating unit consists of an update gate and a reset gate. The update gate controls the retention rate of historical information, while the reset gate controls the forgetting rate. The self-attention mechanism dynamically adjusts the contribution weight of different input features in the decision-making process by calculating the attention weight of each input feature. Specifically, the attention weight for battery health status assessment ranges from 0.4 to 0.6, the attention weight for current state of charge ranges from 0.25 to 0.35, and the attention weight for current battery internal temperature ranges from 0.1 to 0.2, ensuring that key features dominate the model output. The model output layer has three nodes, corresponding to charging current, charging voltage, and charging cutoff condition parameters, respectively. The output values, after inverse normalization, are directly used as parameters for the charging curve. The historical charging optimization dataset is loaded from the storage unit of the control chip. The storage unit reserves 8MB of space specifically for storing this dataset. The dataset contains complete data of the same model battery after 1000 charge-discharge cycles, covering relevant data of the battery under different health states, states of charge and temperature conditions. The health state covers three levels: level 1, level 2 and level 3. The state of charge is divided into 10 intervals in 10% intervals. The temperature is divided into 12 intervals in 5℃ intervals. Each combination of conditions contains 5 sets of test data with different charging curve parameters.The historical charging optimization dataset contains battery performance data obtained under different health states, states of charge, and temperature conditions using different charging curve parameters. Specifically, this includes capacity retention, internal resistance change, and cycle life loss after charging. These data serve as training samples and supervision labels for the neural network. The charging curve parameters act as auxiliary features of the input samples, while the battery performance data serves as supervision labels, corresponding to the three core features of the input layer to form complete training sample pairs. A total of 1800 training sample pairs were constructed, divided into training and validation sets in an 8:2 ratio. The training set was used for model training, and the validation set was used to verify the model's training effect. During the charging interval (the resting period after the battery is fully charged until the next charging begins), the feedforward neural network model was trained online using the historical charging optimization dataset. The charging interval was set to 5 minutes to ensure that the training process did not affect normal charging operations. The training process was executed by the computing unit of the control chip, with 50 training iterations and a learning rate of 0.001 for each iteration. The model parameters were adjusted using the backpropagation algorithm to gradually reduce the deviation between the battery performance data corresponding to the charging curve parameters output by the model and the supervision labels. During training, a knowledge distillation loss is introduced, using the prior mapping rules in the battery health status perception model as soft labels. These prior mapping rules are the mapping rules between the battery health status and charging curve parameters. The knowledge distillation loss incorporates these prior mapping rules into the model training, constraining the parameter update direction of the neural network and preventing overfitting. The knowledge distillation loss weight is set to 0.3 and is weighted and fused with the model's mean squared error loss to form the model's total loss function. Training stops when the total loss function value is less than 0.001, yielding the pre-trained network weight parameters. These weight parameters are stored in the control chip's temporary storage unit for subsequent online fine-tuning. The pre-trained network weight parameters are then convolved and fused with real-time acquired battery state of charge (SOC)-internal resistance change gradient data. This gradient data is calculated using SOC and internal resistance values ​​acquired every 100ms. The gradient value is equal to the ratio of the difference in internal resistance to the difference in SOC between two adjacent acquisitions, within a range of 0.5mΩ / % to 2mΩ / %. The convolutional fusion process is implemented through the computing unit of the control chip. The network weight parameters and gradient data are convolved point by point according to their corresponding dimensions. The convolution kernel size is set to 3×3, and the calculation is performed 3 times to generate a dynamic weight bias vector for the current battery state. The dimension of this vector is consistent with the dimension of the network output layer node, which is a 3-dimensional vector. Each dimension corresponds to a bias value of a charging curve parameter. The bias value is controlled between -0.05 and 0.05 and is used to dynamically adjust the model output deviation.The dynamic weight bias vector is superimposed onto the output layer of the battery health status perception model. This superposition process is achieved by directly adding the bias value to the output value of the output layer node, enabling online fine-tuning of the model under the current charging scenario. The fine-tuned model better adapts to the actual state of the battery, reducing the impact of deviations between historical datasets and the current battery state. The fine-tuned model is then deployed in the real-time inference unit of the control chip, with the computation rate of the real-time inference unit set to 10. 6 The system operates at a rate of [number] times per second to ensure rapid response to real-time data. The current battery health status assessment, state of charge, and internal temperature are input into the model. The input data is sequentially passed to the hidden layer according to the order of the input layer nodes. The gating units in the hidden layer, combined with a self-attention mechanism, dynamically adjust the contribution weight of each input feature. The output value of the hidden layer is then calculated through the activation function of the neurons. The sigmoid function is used to control the output value between 0 and 1. The hidden layer output value is passed to the output layer, where it undergoes forward propagation calculation using the linear activation function. After inverse normalization, the output value is directly output as a set of charging curve parameters, comprising the optimal charging current, charging voltage, and charging cutoff conditions for the current moment. The charging current range is adjusted according to the battery health status level: Level 1 health status is 0.3C~1.2C, Level 2 is 0.2C~0.8C, and Level 3 is 0.1C~0.5C. The charging voltage range is 4.1V~4.2V. The charging cutoff conditions include a cutoff current of 0.02C~0.05C, a cutoff voltage of 4.1V~4.2V, and a cutoff temperature of 55℃. This charging curve parameter set is then transmitted to relevant modules to generate an adjustable charging curve, enabling adaptive charging based on battery health status awareness. This ensures efficient charging while maximizing battery protection and extending battery cycle life.

[0086] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A charging method based on battery state of health aware adjustable charging profile, characterized in that, Includes the following steps: Step S1: Voltage and current responses are synchronously acquired using a miniature EIS measurement circuit, and the real and imaginary parts of the impedance at different frequencies are obtained through Fast Fourier Transform. Based on the real and imaginary parts of the impedance at different frequencies, an electrochemical impedance spectrum of the battery during charging and discharging is generated. The miniature EIS measurement circuit includes a programmable arbitrary waveform generator, a high-precision data acquisition unit, and an analog switch. Generating the electrochemical impedance spectrum of the battery during charging and discharging includes the following steps: Design a programmable arbitrary waveform generator. The digital signal corresponding to the pseudo-random binary sequence output by the DAC is amplified by the signal conditioning circuit composed of operational amplifier Q1 to form an AC excitation signal with an amplitude controlled within 50mV. The symbol rate of the pseudo-random binary sequence is set to a value that covers the main frequency range of the battery impedance response, so as to obtain wideband impedance information in one excitation process. An analog switch is connected in series between the battery and the charger. The AC excitation signal is connected to both ends of the battery by controlling the on and off of the analog switch. At the same time, the power circuit of the charger is isolated by optocoupler isolation devices U3 and U4. By building a high-precision data acquisition unit inside a programmable arbitrary waveform generator, the voltage response of the battery is amplified by an instrumentation amplifier, and the amplified voltage signal and current response signal are synchronously acquired by an ADC to generate time-domain voltage and current response data; wherein the sampling rate is set to be greater than or equal to the highest sampling rate of the frequency component corresponding to the AC excitation signal. The time-domain voltage and current response data are transmitted to the control chip MCU. The impedance spectrum analysis unit in the control chip MCU performs a fast Fourier transform on the time-domain voltage and current response data, converting the time-domain signal into a frequency-domain signal, and calculates the real and imaginary parts of the impedance at different frequencies. Based on the real and imaginary parts of the impedance at different frequencies, the electrochemical impedance spectrum of the battery during the charging and discharging process is generated. Step S2: Based on the electrochemical impedance spectrum of the battery during the charging and discharging process, the high-frequency region, mid-frequency region and low-frequency region are segmented and fitted, and the characteristic values ​​of ohmic internal resistance, charge transfer impedance and diffusion impedance are extracted respectively to generate impedance spectrum characteristic parameters; the impedance spectrum characteristic parameters are input into the control chip for pre-storage and fitted with the battery health status mapping table to estimate the corresponding battery health status assessment value. Step S3: Obtain the current state of charge of the battery and construct a battery health status perception model with the battery health status assessment value to dynamically generate an adjustable charging curve; Step S4: Perform a charging operation on the battery according to the adjustable charging curve, and periodically collect the real-time impedance spectrum and state of charge data of the battery during the charging process. Feed these data back to the battery health status perception model to update the battery health status assessment value. Then, dynamically adjust the parameters of the adjustable charging curve according to the updated battery health status assessment value to achieve closed-loop adaptive adjustment of the charging process.

2. The charging method based on an adjustable charging curve for battery health status sensing according to claim 1, characterized in that, The generation of impedance spectrum characteristic parameters in step S2 includes the following steps: The electrochemical impedance spectrum is divided into frequency bands based on the frequency response characteristics of the battery's internal impedance. The electrochemical impedance spectrum is divided into high-frequency, mid-frequency and low-frequency regions. The high-frequency region corresponds to the ohmic internal resistance characteristics of the battery, the mid-frequency region corresponds to the charge transfer impedance and double-layer capacitance characteristics, and the low-frequency region corresponds to the diffusion impedance and Warburg impedance characteristics. In the high-frequency region, the frequency segment corresponding to the extreme point of the impedance modulus is selected, and the real part of the impedance in the frequency segment is linearly fitted to obtain the characteristic value of the ohmic internal resistance. At the same time, the impedance phase angle of the frequency segment is extracted to verify whether the impedance characteristics in the high-frequency region conform to the ohmic internal resistance response law of the battery. In the mid-frequency region, the frequency segment corresponding to the diameter of the semicircular arc in the impedance spectrum is selected. The intersection of the real and imaginary parts of the impedance in this frequency segment is analyzed. The characteristic value of the charge transfer impedance is obtained by fitting the equivalent circuit model. At the same time, the peak value of the impedance phase angle in this frequency segment is extracted to help verify the impedance characteristics in the mid-frequency region. In the low-frequency region, a frequency band extending at a 45° angle in the impedance spectrum is selected. The difference between the real and imaginary parts of the impedance in this frequency band is linearly fitted to obtain the characteristic value of the diffusion impedance. At the same time, the characteristic value of the diffusion impedance is verified to be reasonable by combining the impedance phase angle variation law in the low-frequency region. The extracted ohmic internal resistance, charge transfer impedance, and diffusion impedance characteristic values ​​are used as impedance spectrum characteristic parameters and stored in the corresponding temporary data buffer of the control chip.

3. The charging method based on an adjustable charging curve for battery health status sensing according to claim 2, characterized in that, The step of selecting a frequency band corresponding to the diameter of a semicircular arc in the impedance spectrum in the mid-frequency region and analyzing the intersection of the real and imaginary parts of the impedance in that frequency band includes the following steps: Based on the frequency band corresponding to the semi-circular diameter of the impedance spectrum selected in the mid-frequency region, the real and imaginary parts of the impedance corresponding to each frequency point in the frequency band are extracted to generate a three-dimensional data set of frequency-real-imaginary part. Based on the frequency-real-imaginary three-dimensional data set, an equivalent circuit physical model is constructed that includes charge transfer impedance, double-layer capacitance and solid-state diffusion impedance parameters. The solid-state diffusion impedance parameter in the equivalent circuit physical model is used as the associated coupling variable of charge transfer impedance. Based on the equivalent circuit physical model, a nonlinear complex plane least squares fitting is performed on the frequency-real-imaginary three-dimensional data set. During the fitting process, the solid-phase diffusion impedance parameter is introduced as a constraint on the charge transfer impedance, and the initial charge transfer impedance estimate is obtained by solving the problem. The peak value of the impedance phase angle at the characteristic frequency in the mid-frequency region of the impedance spectrum is obtained. The initial charge transfer impedance estimate and the peak value of the impedance phase angle are coupled together to calculate and generate the verification correction factor of the charge transfer impedance. The verification correction factor is convolved with the initial charge transfer impedance estimate to generate the final charge transfer impedance characteristic value, and the frequency and phase angle information corresponding to the characteristic value are recorded.

4. The charging method based on an adjustable charging curve for battery health status sensing according to claim 3, characterized in that, The verification correction factor for generating the charge transfer impedance includes the following steps: The initial charge transfer impedance estimate and the peak value of the impedance phase angle are obtained. Based on the charge transfer dynamics theory, the initial charge transfer impedance estimate is mapped to an equivalent charge exchange current density to generate charge exchange current density data. The solid-phase diffusion rate at the electrode-electrolyte interface is obtained, and the amplitude phase hysteresis of the charge exchange current density at the solid-phase diffusion rate is calculated by combining the charge exchange current density data. Extract the actual impedance imaginary part amplitude corresponding to the characteristic frequency point from the frequency-real-imaginary part three-dimensional data set, and perform cross-correlation analysis between it and the amplitude phase lag to generate the dynamic coupling coefficient between the charge transfer process and the solid-phase diffusion process. The dynamic coupling coefficient is substituted into the preset correction factor generation model to perform frequency response convolution inversion and generate impedance verification correction factor; the impedance verification correction factor is subjected to frequency domain weighted filtering based on the ratio between the amplitude phase lag and the actual impedance imaginary part amplitude to filter out error components introduced by high frequency noise or low frequency drift and generate charge transfer impedance verification correction factor.

5. The charging method based on an adjustable charging curve for battery health status sensing according to claim 1, characterized in that, The fitting and estimation of the corresponding battery health status assessment value in step S2 includes the following steps: Multiple groups of batteries of the same model were selected for accelerated aging tests. Each group of batteries was aged under the same ambient temperature and charge-discharge regime. The electrochemical impedance spectrum of the battery was collected every certain number of cycles and the characteristic parameters of the impedance spectrum were extracted. At the same time, the remaining capacity, capacity decay rate and internal resistance growth rate of each group of batteries were recorded as indicators of battery health status. The impedance spectrum characteristic parameters of each group of batteries were correlated with the battery health status characterization index, and the mapping relationship between the characteristic parameters and the battery health status was established by statistical methods. The established mapping relationship is stored in the Flash memory of the control chip to form a mapping table between feature parameters and battery health status. This table includes the threshold range of the feature parameters and the corresponding battery health status level, as well as the correspondence between the trend of feature parameter change and the trend of battery health status change. Once the impedance spectrum characteristic parameters of the battery are obtained, the corresponding battery health status level is found in the mapping table between characteristic parameters and battery health status. Based on the battery health status level, the battery health status assessment value corresponding to the current impedance spectrum characteristic parameters is calculated, including the remaining capacity estimate, the internal resistance growth rate estimate, and the capacity decay trend estimate.

6. The charging method based on an adjustable charging curve for battery health status sensing according to claim 1, characterized in that, Step S3 includes the following steps: A battery health status perception model is established based on a local lightweight learning algorithm, which includes a battery health status input interface, a state of charge input interface, a charging curve parameter output interface, and a model parameter update interface. The battery health status input interface receives the battery health status evaluation value, and the state of charge input interface receives the current state of charge value of the battery. The model sets mapping rules between battery health status and charging curve parameters, and sets corresponding upper limits for charging current, charging voltage, and charging cut-off conditions according to different levels of battery health status assessment values ​​and different ranges of battery state of charge. The adjustment step size and frequency of the charging curve parameters are adjusted in real time according to the changing trend of the battery health status assessment value and the changing rate of the battery state of charge, so as to generate corresponding dynamic adjustment rules. Based on the mapping rules and dynamic adjustment rules, the program code that can run on the control chip is compiled and loaded into the control logic of the battery health status perception model to generate corresponding charging curve parameters according to the input battery health status assessment value and battery state of charge value. Adjustable charging curves are generated based on charging curve parameters, including a continuous smooth curve of charging current changing with time, a continuous smooth curve of charging voltage changing with time, and an adaptive curve of charging cutoff conditions. The adjustable charging curves together define the charging strategy of the battery under different health and charge states.

7. The charging method based on an adjustable charging curve for battery health status perception according to claim 6, characterized in that, The process of generating an adjustable charging curve based on charging curve parameters includes the following steps: Obtain the set of output charging curve parameters, which includes charging current parameters, charging voltage parameters and charging cut-off condition parameters, and use this set of charging curve parameters as the initial charging curve control parameter vector, wherein the charging cut-off condition parameters include cut-off voltage, cut-off current and cut-off temperature threshold. Based on the current state of charge-internal resistance change gradient data of the battery, the control parameter vector of the initial charging curve is dynamically adjusted to generate a charging current instantaneous adjustment coefficient vector and a charging voltage instantaneous adjustment coefficient vector containing the time dimension. Based on the charging current instantaneous adjustment coefficient vector and the charging voltage instantaneous adjustment coefficient vector, combined with the entropy change rate data of the internal electrochemical reaction of the battery as the state constraint variable, a dynamic response constraint of charging current and charging voltage with time as the independent variable is constructed. The real-time polarization voltage data of the battery during the charging process is acquired, and the real-time polarization voltage data is fed back to the dynamic response constraint iterative calculation to generate a continuous smooth curve of charging current changing with time and a continuous smooth curve of charging voltage changing with time. Based on the changing trends of the state of charge-internal resistance gradient data and the entropy change rate data of the internal electrochemical reaction, the charging cutoff condition parameters are dynamically corrected to generate an adaptive charging cutoff condition curve that changes synchronously with the continuous smooth curve of charging current changing with time and the continuous smooth curve of charging voltage changing with time. The adaptive charging cutoff condition curve defines the charging termination threshold at different time points, and together with the continuous smooth curve of charging current changing with time and the continuous smooth curve of charging voltage changing with time, they constitute a complete adjustable charging curve.

8. The charging method based on an adjustable charging curve for battery health status perception according to claim 7, characterized in that, Generating the adaptive curve for the charging cutoff condition includes the following steps: The correlation coefficient between the state of charge-internal resistance change gradient data and the entropy change rate data of the internal electrochemical reaction in the battery is calculated within the sliding time window to generate a characteristic coupling coefficient characterizing the internal dynamic-thermodynamic coupling state of the battery. Based on the characteristic coupling coefficient, the charging cutoff condition parameters are normalized and corrected, and each corrected charging cutoff condition parameter is multiplied by the exponential decay constraint of the characteristic coupling coefficient to obtain the first round of cutoff condition parameter adjustment values. Obtain the first derivative of the continuous smooth curve of the charging current changing with time in the current charging stage within a set time interval, and perform a convolution operation on the first derivative with the feature coupling coefficient to generate a dynamic adjustment factor for the influence of current change on the cutoff condition. The dynamic adjustment factor is weighted and fused with the first round of cutoff condition parameter adjustment value, wherein the weight is dynamically determined by the second derivative of the entropy change rate data of the internal electrochemical reaction of the battery, to generate the second round of cutoff condition parameter adjustment value. Based on the second round of cutoff condition parameter adjustment values, and combined with the slope of the continuous smooth curve of the charging voltage changing with time at the current moment, a continuous relationship between the charging cutoff condition parameters and time is constructed. With time as the independent variable, the cutoff voltage threshold curve, cutoff current threshold curve, and cutoff temperature threshold curve are output, which together constitute the adaptive curve of the charging cutoff condition.

9. The charging method based on an adjustable charging curve for battery health status sensing according to claim 6, characterized in that, The local lightweight learning algorithm is trained using a decision tree algorithm or a small-scale neural network. The battery health status perception model is constructed using the small-scale neural network, including: A small-scale feedforward neural network model is constructed. Its input layer nodes receive battery health status assessment values, current state of charge, and current battery internal temperature. Its hidden layer adopts a gated unit structure with a self-attention mechanism to dynamically adjust the contribution weights of different input features in decision-making. Historical charging optimization datasets are loaded from the storage unit of the control chip. These datasets contain battery performance data obtained under different health status, state of charge, and temperature conditions using different charging curve parameters. These datasets serve as training samples and supervision labels for the neural network. During the charging interval, the feedforward neural network model is trained online using the historical charging optimization dataset to generate a battery health status perception model. Knowledge distillation loss is introduced during the training process, and the prior mapping rules in the battery health status perception model are used as soft labels to constrain the parameter update direction of the neural network, so as to obtain the pre-trained network weight parameters. The pre-trained network weight parameters are convolved and fused with the real-time collected battery state of charge-internal resistance change gradient data to generate a dynamic weight bias vector for the current battery state. This dynamic weight bias vector is then superimposed on the output layer of the battery health state perception model to complete the online fine-tuning of the model in the current charging scenario. The fine-tuned model is deployed in the real-time inference unit of the control chip. The current battery health status assessment value, state of charge and internal temperature are input into the model. Through the forward propagation calculation of the model, the charging curve parameter set consisting of the current optimal charging current, charging voltage and charging cut-off condition is directly output. The charging curve parameter set is then used to generate an adjustable charging curve.

Images