A method for formation and capacity control of a gallium nitride-based low-ripple lithium battery

Abstract

This invention discloses a low-ripple lithium battery formation and capacity control method based on gallium nitride (GaN), relating to the fields of lithium battery manufacturing equipment and power electronic control technology. The invention establishes an improved state-space prediction model incorporating the nonlinear characteristics of GaN bidirectional converters and the electrochemical characteristics of lithium batteries. It collects battery voltage, current, and GaN switching transistor operating parameters in real time. Based on the prediction model and a rolling optimization strategy, it generates a finite number of candidate switching transistor action timing sequences in each control cycle, calculates the output current ripple value for each combination, and selects the combination with the smallest ripple as the optimal control sequence for execution. Simultaneously, it introduces an error correction mechanism, adaptive prediction cycle adjustment, and online calibration compensation for device aging, ensuring that the output current ripple accuracy remains below 0.1% throughout the entire lifespan. In a multi-channel system, high-precision clock synchronization causes the carrier phases of each channel's switches to be interleaved sequentially, reducing the total output ripple to 1 / 4 of that of a single channel.

Description

Technical Field

[0001] This invention relates to the field of lithium battery manufacturing equipment and power electronic control technology, specifically to a method for low-ripple lithium battery formation and capacity control based on gallium nitride. Background Technology

[0002] Formation and capacity testing of lithium-ion batteries is a crucial step in lithium-ion battery manufacturing. Its purpose is to activate new batteries through the first charge-discharge cycle, forming a stable solid electrolyte interphase (SEI) film inside the battery and grading and screening the battery capacity. During formation and capacity testing, the ripple of the charging current directly affects the formation quality of the SEI film and the battery's cycle life. Traditional formation and capacity testing equipment often uses silicon-based MOSFETs or IGBTs as power switching devices, controlling the charging current through PWM modulation. However, with the increasing demands for current ripple in lithium-ion battery manufacturing, and the increasingly stringent requirements for formation efficiency and consistency, traditional silicon-based switching devices, due to their slow switching speed, high conduction losses, and significant nonlinear characteristics, are no longer sufficient to meet the requirements for low-ripple, high-precision formation and capacity testing. In recent years, gallium nitride (GaN) devices, with their advantages of high electron mobility, low on-resistance, and ultra-fast switching speed, have been gradually introduced into the field of lithium-ion battery formation and capacity testing, providing a new technological path for achieving ultra-low ripple control.

[0003] However, existing gallium nitride (GaN)-based lithium-ion battery formation and capacity control technologies still have many shortcomings. First, traditional control methods often employ fixed duty cycle PWM modulation or simple PI closed-loop control, failing to fully consider the nonlinear characteristics of GaN bidirectional converters and the electrochemical characteristics of lithium batteries. This results in unsatisfactory current ripple suppression under different operating conditions, making it difficult to maintain stable low ripple accuracy. Second, existing technologies lack effective compensation mechanisms for GaN device aging. As the cumulative operating time and switching frequency increase, parameters such as on-resistance, gate threshold voltage, and parasitic capacitance drift, leading to control model mismatch and gradual degradation of ripple accuracy over time. Furthermore, in multi-channel formation and capacity systems, there is a lack of effective ripple collaborative suppression mechanisms between channels. Current ripples between channels overlap, resulting in a large overall system output ripple and difficulty in ensuring current consistency across channels. In addition, traditional control methods lack predictive ability when battery states change abruptly, failing to adaptively adjust the control strategy, which can easily cause a sudden increase in ripple, affecting formation quality.

[0004] In summary, existing lithium-ion battery composition and capacity control technologies have significant shortcomings in ripple suppression accuracy, device aging compensation, multi-channel collaborative control, and dynamic operating condition adaptability, making it difficult to meet the stringent requirements of current lithium-ion battery manufacturing for ultra-low ripple, high consistency, and long-life operation. Therefore, there is an urgent need for a low-ripple lithium-ion battery composition and capacity control method that can fully integrate the characteristics of gallium nitride devices with the electrochemical characteristics of lithium-ion batteries, and possess online aging compensation and multi-channel collaborative ripple suppression capabilities. This method aims to achieve the control objective of maintaining ripple accuracy below 0.1% throughout the entire lifespan, while reducing the total output ripple of the multi-channel system to one-quarter of that of a single channel, and controlling the current consistency error of each channel to within 0.05%. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a gallium nitride (GaN)-based low-ripple lithium battery formation and capacity control method. This method establishes an improved state-space prediction model incorporating the nonlinear characteristics of the GaN bidirectional converter and the electrochemical characteristics of the lithium battery. It collects battery voltage, current, and GaN switching transistor operating parameters in real time. Based on the prediction model and a rolling optimization strategy, it generates a finite number of candidate switching transistor action timing sequences in each control cycle, calculates the output current ripple value for each combination, and selects the combination with the smallest ripple as the optimal control sequence for execution. Simultaneously, it introduces an error correction mechanism, adaptive prediction cycle adjustment, and online calibration compensation for device aging, ensuring that the output current ripple accuracy remains below 0.1% throughout the entire lifespan. In multi-channel systems, high-precision clock synchronization causes the switching carrier phases of each channel to be interleaved sequentially, reducing the total output ripple to 1 / 4 of that of a single channel, and the current consistency error of each channel is less than 0.05%.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for classifying and controlling the capacity of gallium nitride-based low-ripple lithium batteries, the method comprising the following components:

[0007] Step 1: Establish a predictive model that includes the nonlinear characteristics of gallium nitride bidirectional converters and the electrochemical characteristics of lithium batteries;

[0008] Step 2: Real-time acquisition of voltage and current signals of the lithium battery to be processed, as well as the operating status parameters of the gallium nitride switch.

[0009] Step 3: Based on the prediction model and the parameters collected in real time, predict the voltage and current change trends of the lithium battery in the next N control cycles, and introduce an error correction mechanism in the prediction process;

[0010] Step 4: Based on the prediction model and the parameters collected in real time, a rolling optimization strategy is adopted to generate a finite number of candidate gallium nitride switch timing combinations, and the output current ripple value under each combination is calculated.

[0011] Step 5: Select the timing combination with the minimum output current ripple as the optimal control sequence;

[0012] Step 6: Execute the switching action of the first control cycle in the optimal control sequence, and repeat steps 3 to 6 above in the next control cycle for rolling optimization.

[0013] Furthermore, in step 1, the prediction model adopts an improved state-space model, with the following formula:

[0014] ,

[0015] in For the first The state vector for each control cycle includes four state variables: gallium nitride converter output inductor current, output capacitor voltage, lithium battery ohmic polarization voltage, and concentration polarization voltage. This is a state matrix, whose elements follow the current state vector. and the junction temperature of gallium nitride switches The on-resistance, parasitic capacitance, and parasitic inductance parameters of gallium nitride devices are dynamically changing and are obtained through factory calibration and online real-time correction. The input matrix has elements that vary with the input voltage. and switch status change; This is the output matrix; The direct transmission matrix is ​​used; the polarization parameters of the lithium battery are obtained through electrochemical impedance spectroscopy before formation and are updated online during formation based on real-time voltage and current data; among them, the state matrix... At least one element is a state vector and junction temperature The nonlinear function is used to characterize the nonlinear change in the on-resistance of gallium nitride devices with junction temperature; at least one element of the input matrix B varies with the switching state. The changes are used to characterize the nonlinear behavior of converter topology switching; the polarization parameters of the lithium battery are introduced into the model through electrochemical impedance spectroscopy testing and online updates.

[0016] Furthermore, in step 1, the aging status of the gallium nitride device is monitored in real time, and the cumulative operating time and switching count of the device are recorded. For every 1000 hours of cumulative operating time or an increase of 10 switching counts compared to the previous calibration, the aging status is recorded. 9Next, an online calibration of device parameters is performed. The specific calibration method is as follows: During the break in the formation and capacity testing process, the bidirectional converter is placed in test mode, and a test pulse with a duty cycle of 5% and a frequency of 1kHz is applied. The on-resistance, gate threshold voltage, and parasitic capacitance parameters of the device are measured. The on-resistance is obtained by measuring the ratio of drain-source voltage to current, and the gate threshold voltage is obtained by scanning the gate voltage and monitoring the abrupt change point of the drain current. The measured parameters are updated into the prediction model to compensate for parameter changes caused by device aging. At the same time, an aging warning threshold is set. When the on-resistance of the device increases by more than 20% or the gate threshold voltage changes by more than 10%, a warning signal is issued to remind maintenance personnel to replace the device, ensuring that the ripple accuracy of the system remains below 0.1% throughout its entire life cycle.

[0017] Furthermore, in step 2, the terminal voltage, charging / discharging current, and surface temperature signals of the lithium battery to be processed are acquired in real time. The acquisition frequency is set to 20 times the switching frequency. Differential sampling is used to suppress common-mode interference, achieving a voltage sampling accuracy of 1mV, a current sampling accuracy of 0.1A, and a temperature sampling accuracy of 0.1℃. Simultaneously, the gate voltage, drain-source voltage, and on-state voltage drop signals of the gallium nitride switch are acquired. The acquisition frequency is synchronized with the battery parameter acquisition frequency. All acquired signals are digitally filtered and then input into the digital controller. The digital filtering uses a combination of moving average filtering and median filtering, with the filtering window size set to 5 sampling points to effectively remove high-frequency noise and spike interference, ensuring that the parameters input to the prediction model are accurate and reliable.

[0018] Furthermore, in step 3, the voltage and current variation trends of the lithium battery over the next N control cycles are predicted based on the prediction model and real-time acquired parameters. The parameters acquired in the current control cycle are used as the initial conditions for prediction. The initial conditions are substituted into the prediction model, and the state vector and output vector for each future control cycle are calculated sequentially to obtain the predicted values ​​of the lithium battery voltage and current over the next N control cycles. An error correction mechanism is introduced during the prediction process. The predicted value of the previous control cycle is compared with the actual acquired value to calculate the prediction error. The prediction error is multiplied by a correction coefficient and added to the prediction result of the current control cycle. The correction coefficient is adaptively adjusted according to the magnitude of the prediction error. Specifically, when the prediction error is less than 0.5%, the correction coefficient is set to 0.1; when the prediction error is between 0.5% and 2%, the correction coefficient is set to 0.3; and when the prediction error is greater than 2%, the correction coefficient is set to 0.5, ensuring that the prediction accuracy remains within 1% across the entire operating range.

[0019] Furthermore, in step 3, the changes in battery state are detected in real time, and the rate of change of battery voltage and current is calculated. When the rate of change of voltage exceeds 0.1V / s or the rate of change of current exceeds 10A / s, for a battery with a nominal capacity of 50Ah, this threshold can be adjusted according to the battery type and capacity configuration, indicating a sudden change in battery state. At this time, the prediction cycle number N is automatically increased from 5 to 10 to improve the anticipation of control. At the same time, the weight of the objective function is adjusted, and the weight of ripple suppression is increased to twice the original value to prioritize ripple accuracy. When the battery state returns to stability, that is, the rate of change of voltage is less than 0.02V / s and the rate of change of current is less than 2A / s for 10 control cycles, the prediction cycle number N is automatically restored to 5, and the weight of the objective function is also restored to the initial value, ensuring that the system can obtain good control performance under both steady-state and dynamic conditions.

[0020] Furthermore, in step 4, the output current ripple value for each switching timing combination is calculated using the following formula: ,in This represents the total output current ripple value of the switching timing combination over the next N control cycles. For the first The formula for calculating the single-cycle current ripple value of each control cycle is as follows: ,in Input voltage, This is the predicted value of the battery terminal voltage. For the output inductance value, and These represent the on and off times of the switch within that cycle; For the first The formula for calculating the battery polarization coefficient for each control cycle is as follows: ,in This is the battery polarization resistor. For ohmic internal resistance, and It is obtained through online identification using the battery equivalent circuit model and real-time collected voltage and current. The maximum ripple amplitude over N periods. The formula, with weighting coefficients, fully considers the influence of battery electrochemical characteristics on ripple, thus improving the accuracy of ripple calculation.

[0021] Furthermore, in step 5, the timing combination with the minimum output current ripple is selected as the optimal control sequence. An optimization objective function is constructed with the goal of minimizing the ripple cost J. At the same time, multiple constraints are set, including the maximum junction temperature constraint of the gallium nitride switch, the maximum current stress constraint, the maximum charge and discharge current constraint of the battery, and the maximum voltage change rate constraint. The candidate finite number of switching timing combinations are substituted into the objective function for calculation. Combinations that satisfy all constraints are selected. From the combinations that satisfy the constraints, the combination with the minimum objective function value is selected as the optimal control sequence. When there are multiple combinations with equal objective function values ​​or differences within a preset minimum tolerance range, the combination with the fewest switching operations is selected as the optimal control sequence to reduce the switching losses of the gallium nitride device.

[0022] Furthermore, in step 6, the switching action of the first control cycle in the optimal control sequence is executed, and the above steps are repeated in the next control cycle for rolling optimization. The rolling optimization cycle is consistent with the control cycle. After each control cycle, the latest battery and device parameters are immediately collected, the initial conditions of the prediction model are updated, and prediction and optimization are performed again to generate a new optimal control sequence. During the rolling optimization process, the historical data of the first 3 control cycles are retained to correct the parameters of the prediction model and improve the prediction accuracy. When the system is running stably, the parameters of the prediction model are updated once every 100 control cycles. When the system state changes, the parameters of the prediction model are updated immediately to ensure the real-time performance and accuracy of the control.

[0023] Furthermore, when applied to multi-channel batch capacity systems, all channels execute steps 1 to 6, employing a multi-channel collaborative ripple suppression mechanism. The digital controllers of all channels are synchronized using a high-precision clock synchronization signal, achieving a synchronization accuracy of 10ns. For n channels, the switching carrier phases of each channel are sequentially interleaved by 360° / n, causing the current ripples of adjacent channels to cancel each other out. Simultaneously, the ripple influence term of adjacent channels is added to the prediction model of each channel, adjusting the switching timing of this channel based on the predicted ripple values ​​of adjacent channels, further reducing the total output ripple of the system. In a two-channel system, the theoretical maximum total output ripple can be reduced to 1 / 2 of that of a single channel. Through this method's multi-channel collaborative prediction, the actual total output ripple can be further reduced, with tests showing it can be as low as 1 / 4 of that of a single channel, while the current consistency error between channels is less than 0.05%.

[0024] Compared with existing technologies, this gallium nitride-based low-ripple lithium battery formation and capacity control method has the following advantages:

[0025] I. This invention establishes an improved state-space prediction model that incorporates the nonlinear characteristics of gallium nitride bidirectional converters and the electrochemical characteristics of lithium batteries. Combined with a rolling optimization strategy, it generates a finite number of candidate switching timing combinations in each control cycle. The combination with the smallest output current ripple is selected as the optimal control sequence for execution. At the same time, an error correction mechanism and adaptive prediction cycle adjustment are introduced, so that the system can suppress the output current ripple to an extremely low level under both steady-state and dynamic operating conditions, ensuring that the ripple accuracy remains below 0.1% throughout the entire life cycle.

[0026] II. In a multi-channel batching and capacity-breaking system, this invention uses high-precision clock synchronization to sequentially interleave the carrier phases of the switches in each channel. Utilizing the principle of ripple cancellation, the total output ripple of the system is reduced to 1 / 4 of that of a single channel, and the current consistency error of each channel is less than 0.05%. Simultaneously, the aging status of the gallium nitride devices is monitored in real time, with the ripple increasing by 10 units for every 1000 hours of accumulated operation or every 10 units of accumulated switching frequency compared to the previous calibration. 9 The system performs online parameter calibration and issues an early warning when the on-resistance increases by more than 20% or the gate threshold voltage changes by more than 10%, ensuring long-term stable operation of the system.

[0027] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0029] Figure 1 A flowchart of a gallium nitride-based low-ripple lithium battery formation and capacity control method;

[0030] Figure 2 This is a flowchart of the online calibration and compensation process for gallium nitride device aging in a gallium nitride-based low-ripple lithium battery formation and capacity control method.

[0031] Figure 3 This is a flowchart of a multi-channel collaborative ripple suppression control method for a gallium nitride-based low-ripple lithium battery formation and capacity control method. Detailed Implementation

[0032] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0033] Example 1

[0034] This embodiment applies to a single-channel lithium battery formation equipment, performing constant current and constant voltage formation on a 3.2V lithium iron phosphate battery with a nominal capacity of 50Ah. First, a predictive model incorporating the nonlinear characteristics of the gallium nitride bidirectional converter and the electrochemical characteristics of the lithium battery is established. The predictive model employs an improved state-space model; the formula is:

[0035] ,

[0036] in For the first The state vector of each control cycle; The state matrix; The input matrix; This is the output matrix; This is a direct transmission matrix. During system operation, the aging status of the gallium nitride devices is monitored in real time, and the cumulative operating time and switching count of the devices are recorded. For every 1000 hours of cumulative operating time or an increase of 10 switching counts compared to the previous calibration, the data is recorded. 9 Next, an online calibration of the device parameters is performed. Specifically, with the capacity gap reduced, the bidirectional converter is placed in test mode, and a test pulse with a duty cycle of 5% and a frequency of 1kHz is applied. The on-resistance, gate threshold voltage, and parasitic capacitance of the device are measured, and the measured parameters are updated into the prediction model. At the same time, an aging warning threshold is set. When the on-resistance of the device increases by more than 20% or the gate threshold voltage changes by more than 10%, a warning signal is issued.

[0037] The terminal voltage, charging / discharging current, and surface temperature signals of the lithium battery to be processed are acquired in real time. At the same time, the gate voltage, drain-source voltage, and on-state voltage drop signals of the gallium nitride switch are also acquired. All acquired signals are input to the digital controller after digital filtering. The digital filtering adopts a combination of moving average filtering and median filtering, and the filtering window size is set to 5 sampling points.

[0038] The prediction model and real-time acquired parameters are used to predict the voltage and current trends of the lithium battery over the next five control cycles. The parameters acquired in the current control cycle are used as the initial conditions for prediction. The initial conditions are substituted into the prediction model, and the state vector and output vector for each future control cycle are calculated sequentially to obtain the predicted values ​​of the lithium battery voltage and current over the next five control cycles. An error correction mechanism is introduced during the prediction process. The predicted value of the previous control cycle is compared with the actual acquired value to calculate the prediction error. The prediction error is multiplied by a correction coefficient and added to the prediction result of the current control cycle. The correction coefficient is adaptively adjusted according to the magnitude of the prediction error. When the prediction error is less than 0.5%, the correction coefficient is 0.1; when it is between 0.5% and 2%, it is 0.3; and when it is greater than 2%, it is 0.5. The system monitors changes in battery status in real time and calculates the rate of change of battery voltage and current. When the rate of change of voltage exceeds 0.1V / s or the rate of change of current exceeds 10A / s, it is determined that a sudden change in battery status has occurred. At this time, the number of prediction cycles is automatically increased from 5 to 10, and the weight of the objective function is adjusted, with the weight of ripple suppression being increased to twice the original value to prioritize ripple accuracy. When the battery status returns to stability, i.e., the rate of change of voltage is less than 0.02V / s and the rate of change of current is less than 2A / s for 10 control cycles, the number of prediction cycles is automatically restored to 5, and the weight of the objective function is also restored to the initial value.

[0039] Based on the prediction results, a finite number of candidate gallium nitride switch timing combinations are generated. The output current ripple value for each combination is calculated using the total output current ripple value calculation formula; the formula is: ,in This represents the total output current ripple value of the switching timing combination over the next N control cycles. For the first Single-cycle current ripple value for each control cycle; For the first Battery polarization coefficient per control cycle The maximum ripple amplitude over N periods. This is the weighting coefficient, with a value of 0.1.

[0040] The method for generating a finite number of candidate switching timing combinations is as follows: Based on the optimal solution of the duty cycle in the current control cycle, five candidate duty cycles are uniformly selected in the neighborhood of the duty cycle ±10%, and mapped to the switching time of each switch to form five candidate timing combinations; or the branch and bound method is used, limiting the search depth to N=5, and retaining only the top three branches with the minimum ripple cost in each control cycle.

[0041] The optimal control sequence is selected by choosing the action timing combination with the minimum output current ripple. An optimization objective function is constructed with the goal of minimizing the ripple cost J. Multiple constraints are set, including the maximum junction temperature constraint of the gallium nitride switch, the maximum current stress constraint, the maximum charge / discharge current constraint of the battery, and the maximum voltage change rate constraint. The candidate finite number of switching timing combinations are substituted into the objective function for calculation. Combinations that satisfy all constraints are selected. The combination with the minimum objective function value among the combinations that satisfy the constraints is selected as the optimal control sequence. When multiple combinations have the same objective function value or the difference is within a preset minimum tolerance range, the combination with the fewest switching actions is selected as the optimal control sequence.

[0042] The switching action in the first control cycle of the optimal control sequence is executed, and the above steps are repeated in the next control cycle for rolling optimization. Immediately after each control cycle, the latest battery and device parameters are re-acquired, the initial conditions of the prediction model are updated, and prediction and optimization are performed again to generate a new optimal control sequence. During rolling optimization, historical data from the first three control cycles is retained to correct the parameters of the prediction model. When the system is running stably, the parameters of the prediction model are updated every 100 control cycles; when the system state changes, the parameters of the prediction model are updated immediately. In this embodiment, the output current ripple is controlled within 0.1% throughout the entire formation process, significantly improving the consistency and safety of lithium battery formation.

[0043] Example 2

[0044] This embodiment is applied to an industrial-grade four-channel lithium battery formation and capacity testing equipment. It simultaneously performs parallel constant current formation on four 3.2V lithium iron phosphate batteries with a nominal capacity of 100Ah. The rated charge and discharge current of a single channel is 50A. The gallium nitride bidirectional converter adopts a half-bridge topology and the switching frequency is set to 100kHz.

[0045] First, an improved state-space prediction model is independently established for each channel, incorporating the nonlinear characteristics of the gallium nitride converter and the electrochemical characteristics of the lithium battery. The state vector contains 16 state variables, including the output inductor current, output capacitor voltage, and the ohmic polarization voltage and concentration polarization voltage of the corresponding battery for each channel. The digital controllers of all channels are synchronized via a high-precision clock signal transmitted through optical fiber, with the synchronization accuracy controlled within 8ns. The switching carrier phases of the four channels are set to be interleaved by 90°, i.e., channel 1 phase 0°, channel 2 phase 90°, channel 3 phase 180°, and channel 4 phase 270°.

[0046] The ripple influence terms of the three adjacent channels are added to the prediction model of each channel to construct a multi-channel coupled prediction model. The predicted ripple values ​​and switching timing of adjacent channels are obtained in real time. During the switching timing optimization process of this channel, the superposition effect of the ripple of adjacent channels is included in the calculation of the total output ripple, and the weighting coefficient λ of ripple suppression in the objective function is adjusted to 0.15. Each channel independently executes the single-channel low-ripple control process, and the terminal voltage, charging and discharging current, surface temperature of the battery, and the gate voltage, drain-source voltage and on-state voltage drop of the gallium nitride switch are collected at a frequency of 2MHz. The signal is processed by a combination of moving average and median filtering, and the filtering window size is 5 sampling points.

[0047] The voltage and current trends of each channel and the total system output are predicted based on a multi-channel coupled prediction model over the next five control cycles. An error correction mechanism is introduced, with the correction coefficient adaptively adjusted according to the single-channel prediction error. When generating candidate switching timing combinations for each channel, six candidate duty cycles are selected within a neighborhood of ±8% of the current optimal duty cycle. The single-channel ripple and the total system ripple cost J under each combination are calculated. Combinations that satisfy all constraints, including a maximum junction temperature of 125℃ for the gallium nitride switch, a maximum current stress of 80A, a maximum battery charge / discharge current of 50A, and a maximum voltage change rate of 0.05V / s, are selected as the optimal control sequence.

[0048] Throughout the formation process, the current consistency of each channel is continuously monitored, and the current deviation between channels is calibrated every 100 control cycles. Experimental results show that, using the four-channel collaborative control method of this embodiment, the total output current ripple of the system is stabilized below 0.06%, which is only 1 / 5.2 of the total ripple when using single-channel independent control. The current consistency error between channels is less than 0.04%. Compared with the traditional multi-channel independent control method, the total ripple is reduced by 78%, and the channel consistency is improved by 65%.

[0049] Example 3

[0050] This embodiment is applied to a single-channel high-rate lithium battery formation equipment, which performs 3C rate constant current formation on a 4.2V ternary lithium battery with a nominal capacity of 20Ah. The rated charge and discharge current is 60A. The gallium nitride bidirectional converter adopts a full-bridge topology and the switching frequency is set to 150kHz.

[0051] First, an improved state-space prediction model adapted to the electrochemical characteristics of ternary lithium batteries is established. In view of the characteristics of fast polarization response and high voltage change rate of ternary lithium batteries, a dynamic polarization parameter term for the SEI film formation process is added to the model. The initial polarization resistance and capacitance parameters are obtained by electrochemical impedance spectroscopy test before formation, and the polarization parameters are updated online every 50 control cycles during the formation process.

[0052] The system collects real-time data on battery terminal voltage, charging / discharging current, surface temperature, and gallium nitride (GaN) switch operating parameters at a frequency of 3 MHz. Voltage sampling accuracy is 0.5 mV, current sampling accuracy is 0.05 A, and temperature sampling accuracy is 0.05 °C. Based on a predictive model, the system forecasts battery state changes over the next five control cycles. An error correction mechanism is introduced: a correction coefficient of 0.1 for prediction errors less than 0.5%, 0.3 for errors between 0.5% and 2%, and 0.5 for errors greater than 2%.

[0053] The system calculates the battery voltage and current change rates in real time. During the rapid SEI film formation phase in the early stages of formation, when the battery voltage change rate reaches 0.12V / s, the system automatically identifies this as a sudden change in battery state, increases the prediction cycle number N from 5 to 10, and doubles the weight of ripple suppression in the objective function to prioritize ripple accuracy. When the voltage change rate drops to 0.015V / s and the current change rate is less than 1.5A / s for 10 consecutive control cycles, the system automatically restores the prediction cycle number to 5 and the weighting coefficients to their initial values.

[0054] When generating candidate switch timing combinations, a branch-and-bound method is used, limiting the search depth to N=10. The top four branches with the lowest ripple cost are retained for each control cycle. The ripple cost J for each combination is calculated, and the optimal combination satisfying all constraints is selected for execution. Experimental results show that in this embodiment, the output current ripple is consistently controlled below 0.08% throughout the entire 3C high-rate formation process. Even in the early stages of formation when battery state changes abruptly, the ripple peak does not exceed 0.09%. Compared to the traditional PI control method, the ripple suppression effect under high-rate conditions is improved by 85%, effectively avoiding damage to the SEI film of the ternary lithium battery caused by high-current ripple.

[0055] Example 4

[0056] This embodiment is for a cumulative working time of 1500 hours and a cumulative number of on / off cycles of 1.2 × 10⁻⁶. 9 The gallium nitride switch was calibrated and compensated online and applied to a single-channel 50Ah lithium iron phosphate battery formation device with a switching frequency of 100kHz.

[0057] The system records the cumulative operating time and switching count of gallium nitride devices in real time. When the cumulative operating time reaches 1000 hours and the cumulative switching count increases by 10 compared to the last calibration, the system will take action. 9At this time, during the interval after the current battery formation process is completed, the bidirectional converter is automatically switched to test mode. A standard test pulse with a duty cycle of 5% and a frequency of 1kHz is applied, and the on-resistance, gate threshold voltage, and parasitic capacitance parameters of the device are measured through a high-precision sampling circuit: the on-resistance is calculated by measuring the ratio of drain-source voltage to drain current, and the measured value is 18.5mΩ, which is 23.3% higher than the factory calibration value of 15mΩ; the measured value of gate threshold voltage is 1.35V, which is 10% different from the factory value of 1.5V; and the measured value of parasitic capacitance is 220pF, which is 10% higher than the factory value of 200pF.

[0058] The system updates the measured parameters in real time to the state matrix A and input matrix B of the prediction model, compensating for parameter drift caused by device aging. Simultaneously, if the on-resistance increases by more than 20% and the gate threshold voltage changes by 10%, the system automatically issues a device aging warning signal, reminding maintenance personnel to replace the device via the equipment display and remote monitoring platform.

[0059] Battery formation tests were conducted under the same conditions before and after calibration compensation. The results showed that before calibration compensation, due to model parameter mismatch, the output current ripple deteriorated to 0.21%, exceeding the control target of 0.1%. After online calibration compensation, the output current ripple immediately recovered to 0.09%, once again meeting the full life-cycle ripple accuracy requirements. Without aging compensation, after 500 hours of continued operation, the ripple will further deteriorate to 0.35%, severely affecting the formation quality.

[0060] Example 5

[0061] This embodiment simulates the typical abrupt change from constant current to constant voltage during the lithium battery formation process. It is applied to a single-channel 50Ah lithium iron phosphate battery formation equipment, with a constant current charging current of 25A and a constant voltage charging voltage of 3.65V.

[0062] During the constant current charging phase, the battery voltage rises slowly, with the voltage change rate stabilizing below 0.01V / s. The system operates in steady-state control mode, with a prediction cycle number N=5, a ripple suppression weighting coefficient λ=0.1, and the output current ripple stabilizing at approximately 0.07%. When the battery terminal voltage rises to 3.65V, the system switches to constant voltage charging mode. At this point, the current rapidly decreases from 25A to 2A, with a current change rate reaching 12A / s. The system immediately identifies this as a sudden change in battery state.

[0063] The system automatically executes an adaptive adjustment strategy: increasing the number of prediction cycles N from 5 to 10 to improve control anticipation; increasing the weight λ for ripple suppression in the objective function to 0.2 to prioritize ripple accuracy; and simultaneously narrowing the search range of candidate duty cycles to ±5% of the current optimal duty cycle to accelerate optimization calculations. In the first control cycle after mode switching, the system generates the optimal switching sequence based on the updated prediction model and weighting coefficients, effectively suppressing ripple impacts caused by sudden current changes.

[0064] When the current drops below 1A, the voltage change rate is less than 0.02V / s, and the current change rate is less than 2A / s for 10 consecutive control cycles, the system determines that the battery state has stabilized, automatically restores the predicted cycle number N to 5, and the weighting coefficient λ to 0.1, entering steady-state constant voltage charging control. Experimental results show that, using the adaptive control method of this embodiment, the peak current ripple during the constant current to constant voltage transition is only 0.095%, not exceeding the control target of 0.1%; while using the traditional fixed parameter control method, the peak ripple during the transition reaches 0.32%, which is 3.4 times that of the method of this invention, fully verifying the ripple suppression capability of this invention under dynamic operating conditions.

[0065] Comparative analysis of the technical effects of different embodiments:

[0066] one Conventional formation of single-channel lithium iron phosphate Basic scrolling optimization + error correction ≤0.1% - High accuracy of steady-state ripple control two Four-channel lithium iron phosphate parallel formation Multi-channel clock synchronization + phase interleaving + coupling prediction Total ripple ≤ 0.06% <0.04% Significantly reduces total system ripple and improves channel consistency. three High-rate formation of ternary lithium 3C Ternary lithium battery chemical model + dynamic polarization update ≤0.08% - Ripple stability under high magnification conditions Four Gallium nitride device aging compensation Online parameter calibration + aging warning After compensation, ≤0.09% - Achieve ripple accuracy guarantee throughout the entire life cycle five Constant current to constant voltage sudden change working condition Adaptive prediction cycle + dynamic weight adjustment Peak value ≤ 0.095% - Small ripple impact under dynamic operating conditions

[0067] The above embodiments comprehensively cover the implementation of the present invention under different battery types, different system architectures, different device states and different operating conditions, and verify that the present invention can achieve the control target of output current ripple accuracy ≤0.1% in all scenarios and throughout the entire life cycle. The total ripple of the multi-channel system can be reduced to less than 1 / 5 of that of the single channel, and the channel consistency error is less than 0.05%, which is significantly better than the prior art.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for classifying and controlling the capacity of a low-ripple lithium battery based on gallium nitride, characterized in that, The method comprises the following components: Step 1: Establish a predictive model that includes the nonlinear characteristics of gallium nitride bidirectional converters and the electrochemical characteristics of lithium batteries; Step 2: Real-time acquisition of voltage and current signals of the lithium battery to be processed, as well as the operating status parameters of the gallium nitride switch. Step 3: Based on the prediction model and the parameters collected in real time, predict the voltage and current change trends of the lithium battery in the next N control cycles, and introduce an error correction mechanism in the prediction process; Step 4: Based on the prediction model and the parameters collected in real time, a rolling optimization strategy is adopted to generate a finite number of candidate gallium nitride switch timing combinations, and the output current ripple value under each combination is calculated. Step 5: Select the timing combination with the minimum output current ripple as the optimal control sequence; Step 6: Execute the switching action of the first control cycle in the optimal control sequence, and repeat steps 3 to 6 above in the next control cycle for rolling optimization; In step 1, the prediction model adopts an improved state-space model, and the formula is: ， in For the first The state vector of each control cycle; The state matrix; The input matrix; This is the output matrix; For direct transmission matrix; In step 1, the aging status of the gallium nitride device is monitored in real time, and the cumulative operating time and switching count are recorded. For every 1000 hours of cumulative operating time or an increase of 10 switching counts compared to the previous calibration, the aging status is recorded. 9 Next, an online calibration of device parameters is performed. By applying a test signal, the on-resistance, gate threshold voltage, and parasitic capacitance parameters of the device are measured. The measured parameters are then updated into the prediction model to compensate for parameter changes caused by device aging. At the same time, an aging warning threshold is set. When the on-resistance of the device increases by more than 20% or the gate threshold voltage changes by more than 10%, a warning signal is issued to remind maintenance personnel to replace the device. In step 2, the terminal voltage, charging and discharging current and surface temperature signals of the lithium battery to be processed are collected in real time. At the same time, the gate voltage, drain-source voltage and on-state voltage drop signals of the gallium nitride switch are collected. All collected signals are input into the digital controller after digital filtering. The digital filtering adopts a combination of moving average filtering and median filtering, and the filtering window size is set to 5 sampling points. In step 3, the voltage and current trends of the lithium battery over the next N control cycles are predicted based on the prediction model and real-time collected parameters. The parameters collected in the current control cycle are used as the initial conditions for prediction. The initial conditions are substituted into the prediction model, and the state vector and output vector for each future control cycle are calculated sequentially to obtain the predicted values ​​of the lithium battery voltage and current over the next N control cycles. An error correction mechanism is introduced during the prediction process. The predicted value of the previous control cycle is compared with the actual collected value to calculate the prediction error. The prediction error is multiplied by a correction coefficient and added to the prediction result of the current control cycle. The correction coefficient is adaptively adjusted according to the magnitude of the prediction error. When the prediction error is less than 0.5%, the correction coefficient is set to 0.

1. When the prediction error is between 0.5% and 2%, the correction coefficient is set to 0.

3. When the prediction error is greater than 2%, the correction coefficient is set to 0.

5.

2. The method for classifying and controlling the capacity of a low-ripple lithium battery based on gallium nitride according to claim 1, characterized in that, In step 3, the changes in battery state are detected in real time, and the rate of change of battery voltage and current is calculated. When the rate of change of voltage exceeds 0.1V / s or the rate of change of current exceeds 10A / s, it is determined that a sudden change in battery state has occurred. At this time, the prediction period N is automatically increased from 5 to 10, and the weight of the objective function is adjusted, increasing the weight of ripple suppression to twice the original value to prioritize ripple accuracy. When the battery state returns to stability, that is, the rate of change of voltage is less than 0.02V / s and the rate of change of current is less than 2A / s and lasts for 10 control cycles, the prediction period N is automatically restored to 5, and the weight of the objective function is also restored to the initial value.

3. The method for classifying and controlling the capacity of a low-ripple lithium battery based on gallium nitride according to claim 1, characterized in that, In step 4, the output current ripple value for each switching timing combination is calculated using the following formula: ,in This represents the total output current ripple value of the switching timing combination over the next N control cycles. For the first Single-cycle current ripple value for each control cycle; For the first Battery polarization coefficient per control cycle The maximum ripple amplitude over N periods. These are the weighting coefficients.

4. The method for classifying and controlling the capacity of a low-ripple lithium battery based on gallium nitride according to claim 1, characterized in that, In step 5, the timing combination with the smallest output current ripple is selected as the optimal control sequence. An optimization objective function is constructed with the goal of minimizing the total output current ripple value. At the same time, multiple constraints are set, including the maximum junction temperature constraint of the gallium nitride switch, the maximum current stress constraint, the maximum charge and discharge current constraint of the battery, and the maximum voltage change rate constraint. The candidate finite number of switching timing combinations are substituted into the objective function for calculation. Combinations that satisfy all constraints are selected. The combination with the smallest objective function value among the combinations that satisfy the constraints is selected as the optimal control sequence. When there are multiple combinations with equal objective function values ​​or differences within a preset minimum tolerance range, the combination with the fewest switching actions is selected as the optimal control sequence.

5. The method for classifying and controlling the capacity of a low-ripple lithium battery based on gallium nitride according to claim 1, characterized in that, In step 6, the switching action of the first control cycle in the optimal control sequence is executed, and the above steps are repeated in the next control cycle for rolling optimization. After each control cycle, the latest battery and device parameters are immediately collected, the initial conditions of the prediction model are updated, and prediction and optimization are performed again to generate a new optimal control sequence. During the rolling optimization process, the historical data of the first 3 control cycles are retained to correct the parameters of the prediction model. When the system is running stably, the parameters of the prediction model are updated once every 100 control cycles. When the system state changes, the parameters of the prediction model are updated immediately.

6. The method for classifying and controlling the capacity of a low-ripple lithium battery based on gallium nitride according to claim 1, characterized in that, When applied to a multi-channel batch capacity system, all channels execute steps 1 to 6, employing a multi-channel collaborative ripple suppression mechanism. The digital controllers of all channels are synchronized using a high-precision clock synchronization signal, achieving a synchronization accuracy of 10ns. For n channels, the switching carrier phases of each channel are sequentially interleaved by 360° / n, causing the current ripples of adjacent channels to cancel each other out. Simultaneously, the prediction model for each channel incorporates the ripple influence term of adjacent channels, adjusting the switching timing of this channel based on the predicted ripple values ​​of adjacent channels, further reducing the total output ripple of the system. In a two-channel system, through multi-channel collaborative prediction, the total output ripple can be reduced to as low as 1 / 4 of that of a single channel; for an n-channel system, the total output ripple is further reduced, while the current consistency error between channels is less than 0.05%.

Images