Electric two-wheeler extreme charging dynamic equalization control method based on multi-modal feature fusion

Abstract

The application provides a dynamic balance control method for electric two-wheeled vehicle extreme charging based on multi-modal feature fusion, relates to the technical field of electric two-wheeled vehicle extreme charging, and comprises the following steps: collecting multi-modal data of a battery pack of an electric two-wheeled vehicle in an extreme charging state in real time, and performing feature extraction; performing multi-modal feature fusion through a state estimation algorithm to obtain accurate SOC values of each electric core in the battery pack; calculating a maximum SOC difference value of the battery pack, and monitoring the maximum SOC difference value and the safety parameters of the electric core in real time. Through the calculation method combining the charging suspension time and the bypass shunt ratio, the application can intelligently calculate and control parameters according to the SOC difference value and the real-time current, realizes rapid and controllable active intervention, effectively suppresses the inconsistency among the electric cores, realizes rapid cut-off protection in combination with real-time safety parameter monitoring, enhances the charging safety of the battery system, prolongs the service life of the battery, and optimizes the overall performance of the extreme charging process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric two-wheeler extreme charging, and in particular to an electric two-wheeler extreme charging dynamic balancing control method based on multi-modal feature fusion. BACKGROUND

[0002] As a personal short-distance travel tool, the technical development of electric two-wheelers is rooted in the continuous evolution of battery technology, motor driving and control, and vehicle networking and other core technologies. Early products mainly used lead-acid batteries and brush hub motors, which had efficiency and performance bottlenecks. With the significant improvement of lithium-ion battery energy density, safety and cycle life, as well as the maturity of permanent magnet synchronous motor, brushless direct current motor and vector control technology, the vehicle's endurance, power and energy efficiency have been fundamentally improved. At the same time, the vehicle electrical architecture represented by CAN bus has gradually become popular, providing a foundation for the coordinated control of key components such as battery management system (BMS), motor controller, and charging system. In recent years, the integration of Internet of Things modules, sensors and positioning technology has further promoted the development of vehicle state remote monitoring, intelligent anti-theft, OTA upgrade and other intelligent functions, laying a technical foundation for building a digital riding ecosystem.

[0003] Extreme charging technology aims to achieve fast replenishment of electric energy, and its development relies on the coordinated progress of high-rate power batteries, high-power charging architecture and intelligent thermal management systems. This technology is based on high-performance cells that support high-rate discharge and have low internal resistance characteristics, and is combined with battery pack design and connection technology that can withstand high current. The charging system uses a scheme to increase the charging voltage or current, such as through real-time communication protocols between the charging pile and the BMS (battery management system), dynamically adjusting the charging curve to increase the input power as much as possible while ensuring safety. At the same time, efficient thermal management technologies such as liquid cooling or phase change materials are crucial for suppressing the rapid temperature rise of batteries and charging interfaces during high-power charging, and are key support for maintaining charging speed and battery life.

[0004] General electric two-wheeler extreme charging control methods are often triggered based on fixed voltage thresholds or use passive balancing, which have the problems of untimely triggering, slow balancing speed, and low energy efficiency, making it difficult to cope with the inconsistency risks caused by the rapid changes in cell state during extreme charging.

[0005] To solve the above-mentioned defects in the prior art, the present technical solution provides an electric two-wheeler extreme charging dynamic balancing control method based on multi-modal feature fusion. SUMMARY

[0006] The present application provides an electric two-wheeler extreme charging dynamic balancing control method based on multi-modal feature fusion to solve the defects in the prior art.

[0007] In one aspect, the application provides a dynamic equalization control method for electric two-wheeled vehicle extreme charging based on multi-modal feature fusion, comprising:

[0008] S1: Real-time acquisition of multi-modal data of the battery pack of the electric two-wheeled vehicle in the extreme charging state, and feature extraction of the multi-modal data to obtain a feature parameter set for representing the state of the battery cell;

[0009] S2: Based on the feature parameter set, multi-modal feature fusion is performed by a state estimation algorithm to obtain the accurate SOC value of each battery cell in the group;

[0010] S3: Based on the accurate SOC value, the maximum SOC difference value of the battery pack is calculated, and when the maximum SOC difference value is greater than a first set threshold, dynamic equalization is triggered; after the dynamic equalization is triggered, equalization control is performed; the step of performing equalization control comprises:

[0011] S31: The battery cell with a current SOC value higher than the first set threshold is determined as a target battery cell;

[0012] S32: According to the SOC difference value between the target battery cell and the low SOC battery cell and the extreme charging current value, the charging pause time and the bypass shunt ratio required by the target battery cell are calculated, and a control strategy is output;

[0013] S33: Based on the control strategy, the equalization circuit connected to the target battery cell is controlled to realize bypass shunting based on the bypass shunt ratio between the charging loop and the charging loop within the charging pause time, while maintaining normal extreme charging of non-target battery cells;

[0014] S4: Based on the equalization control, the maximum SOC difference value and the safety parameter of the battery cell are monitored in real time; when the maximum SOC difference value decreases to meet the preset safety cutoff condition, the bypass control of the target battery cell is exited, and all battery cells resume normal extreme charging.

[0015] According to the dynamic equalization control method for electric two-wheeled vehicle extreme charging based on multi-modal feature fusion provided by the application, in step S1, the step of obtaining the feature parameters for representing the state of the battery cell comprises:

[0016] S11: Based on the multi-modal data, a multi-modal original data sequence is generated;

[0017] S12: The multi-modal original data sequence is preprocessed, including using a moving average filter or a Kalman filter algorithm to denoise the voltage and current data, and removing outliers from the temperature data, and outputting the preprocessed data;

[0018] S13: Real-time feature extraction is performed on the preprocessed data, and the real-time features include the voltage instantaneous value, the voltage change rate of each battery cell, the average current value, the current direction of the battery pack, the temperature instantaneous value, the temperature change rate and the maximum temperature difference of each measurement point;

[0019] S14: Based on the average current value, the preliminary SOC value of each cell is calculated by the ampere-hour integration method, and the preliminary SOC value is used as an important internal state characteristic parameter, which together with other real-time characteristic parameters constitutes the characteristic parameter set.

[0020] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, step S14, which involves calculating the preliminary SOC value of each battery cell using the ampere-hour integration method, includes:

[0021] S141: Obtain the initial SOC value for each cell;

[0022] S142: Integrate the total current of the battery pack at a high sampling frequency, and calculate the change in SOC after the start of full charging by combining the rated capacity of the cells.

[0023] S143: Add the initial SOC value to the SOC change to obtain the initial SOC value of each cell at the current moment.

[0024] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, step S2, the step of obtaining the precise SOC value of each cell in the group, includes:

[0025] S21: Construct a battery state-space model with the SOC and internal resistance of each cell as state variables;

[0026] S22: Input the feature parameter set as the observation vector into the battery state space model, and perform recursive calculation using the extended Kalman filter algorithm;

[0027] S23: In each step of the recursive calculation, the predicted values ​​of the state variables are corrected using the voltage and temperature multimodal information in the observation vector. By minimizing the covariance between the predicted values ​​and the actual measured values, the optimal estimated accurate SOC value of each cell is output.

[0028] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, step S23, the correction step includes:

[0029] S231: In the correction stage of the Kalman filter algorithm, the actual measured voltage value in the observation vector is compared with the predicted value to obtain voltage information;

[0030] S232: The actual measured temperature value in the observation vector is used as the key constraint condition to adaptively adjust the covariance matrix of process noise and observation noise in the Kalman filter algorithm in real time;

[0031] S233: Based on the covariance matrix and voltage information, calculate the Kalman gain, and use this gain to perform weighted correction on the predicted values ​​of the state variables, outputting the accurate SOC value.

[0032] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, step S32, which calculates the charging pause time and bypass shunting ratio required for the target battery cell, includes:

[0033] S321: Set the equilibrium target and calculate the SOC difference that needs to be eliminated;

[0034] S322: Calculate the amount of charge that needs to be transferred or offset based on the SOC difference to be eliminated and the rated capacity of the target cell.

[0035] S323: Calculate the theoretical charging pause time based on the charge amount and the current charging current of the electrode;

[0036] S324: Correct the theoretical charging pause time based on the real-time average temperature of the battery pack to obtain the actual charging pause time, and calculate the bypass shunting ratio based on the actual charging pause time and the maximum shunting capacity of the equalization circuit.

[0037] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, in step S32, the control strategy includes: when there are multiple target cells, the strongest bypass shunt control is applied to the cell with the highest SOC value first; at the start of equalization, a first preset bypass shunt ratio is used for rapid suppression; when the maximum SOC difference reaches a preset ratio of the target threshold, the shunt ratio is gradually reduced, switching to fine-tuning; the voltage and temperature of the target cells are monitored in real time, and if their rate of increase exceeds the safety limit, the current equalization operation is paused.

[0038] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, step S324, the step of calculating the bypass diversion ratio, includes:

[0039] S3241: Obtain the maximum continuous shunt current value that the equalization circuit can withstand;

[0040] S3242: Based on the equalization target, calculate the theoretical value of the shunt current required to achieve an equivalent charging pause effect;

[0041] S3243: Combines the theoretical value of the shunt current with the maximum shunt capacity I of the equalization circuit. bm The comparison is performed, and the final determined bypass shunt ratio and adjusted charging pause time are output, including:

[0042] If I bn ≤I bmThe bypass shunting ratio is expressed as:

[0043]

[0044] If I bn >I bm The actual charging pause time will be extended according to bypass shunting ratio two, which is expressed as:

[0045]

[0046]

[0047] In the formula, I bn This is the theoretical value of the shunt current. For bypass diversion ratio 1, The bypass diversion ratio is 2. T represents the actual charging pause time. hold Theoretical charging pause time, This is the total electrode charging current.

[0048] According to the multimodal feature fusion-based dynamic equalization control method for electric two-wheeled vehicles provided by the present invention, in step S4, the preset safety cutoff condition includes:

[0049] During the equalization control process, the terminal voltage of the target cell is monitored in real time. If its voltage reaches or exceeds the safe upper limit of voltage determined by dynamic lookup table based on the current battery temperature, the safe cut-off condition is immediately met.

[0050] The rate of change of the terminal voltage of the target cell per unit time is calculated in real time. If the absolute value of the rate of change exceeds the preset gradient safety threshold for multiple consecutive control cycles, the safety cut-off condition is immediately met.

[0051] The surface temperature of the target battery cell is monitored in real time. If the rate of temperature rise per unit time exceeds the preset temperature rise safety threshold, the safety cut-off condition is immediately met.

[0052] The present invention provides a dynamic equalization control method for electric two-wheelers under extreme charging based on multimodal feature fusion. By collecting and fusing multimodal data such as voltage, current, and temperature in real time, and using an extended Kalman filter algorithm for high-precision SOC estimation, the method achieves dynamic equalization control of the battery pack under extreme charging conditions, thereby significantly improving the accuracy of SOC estimation and equalization efficiency. By adopting a calculation method that combines charging pause time and bypass shunting ratio, the method can intelligently calculate control parameters based on the SOC difference and real-time current, and consider the dynamic correction of equalization circuit capability and temperature influence, achieving rapid and controllable active intervention. This effectively suppresses inconsistencies between cells, avoids the risk of overcharging or undercharging, and achieves rapid cut-off protection by combining real-time safety parameter monitoring. Ultimately, this enhances the charging safety of the battery system, extends battery life, and optimizes the overall performance of the extreme charging process. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0054] Fig. 1 This is a flowchart of the dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion provided in Embodiment 1 of the present invention;

[0055] Fig. 2 This is a flowchart illustrating the calculation of the charging pause time and bypass shunting ratio required for the target battery cell in Embodiment 1 of the present invention.

[0056] Fig. 3 This is a flowchart illustrating how the predicted values ​​of state variables are corrected using multimodal information on voltage and temperature in the observation vector, as described in Embodiment 1 of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] Example 1:

[0059] The following is combined Figs. 1-3 This invention describes a dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion.

[0060] like Figs. 1-3 As shown in the figure, the dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion provided in this embodiment of the invention includes:

[0061] S1: Real-time acquisition of multimodal data of the electric two-wheeler battery pack under extreme charging conditions, and feature extraction of the multimodal data to obtain a set of feature parameters characterizing the cell state. Multimodal data includes, but is not limited to: total battery pack voltage, total current, terminal voltage of each cell directly acquired by the battery management system, and temperature data measured by temperature sensors placed on the cell surface or at key locations.

[0062] Step S1, the steps for obtaining characteristic parameters used to characterize the cell state, include:

[0063] S11: Generate a multimodal raw data sequence based on the multimodal data. The multimodal raw data sequence is a collection of multimodal data arranged in chronological order.

[0064] S12: Preprocess the multimodal raw data sequence, including denoising the voltage and current data using moving average filtering or Kalman filtering algorithms, removing outliers from the temperature data, and outputting the preprocessed data. The preprocessing steps also need to include data alignment to ensure consistency in timestamps across different modes, checking data integrity, and performing linear interpolation or context-based imputation for missing data.

[0065] S13: Perform real-time feature extraction on the preprocessed data. Real-time features include the instantaneous voltage value and voltage change rate of each cell, the average current value and current direction of the battery pack, and the instantaneous temperature value, temperature change rate, and maximum temperature difference at each measurement point. In addition, feature extraction also includes calculating the total power of the battery pack, the voltage fluctuation variance of each cell within a certain time window, the temperature distribution uniformity index, and determining whether the current is in a state of continuous rise, fall, or fluctuation based on historical data.

[0066] S14: Based on the average current value, the preliminary SOC value of each cell is calculated by the ampere-hour integration method, and the preliminary SOC value is used as an important internal state characteristic parameter, which together with other real-time characteristic parameters constitutes the characteristic parameter set.

[0067] Step S14, which involves calculating the preliminary SOC value of each cell using the ampere-hour integration method, includes:

[0068] S141: Obtain the initial SOC value for each cell. The initial SOC value can be obtained by looking up a table using the resting voltage at the end of the previous charge, or by estimating and calibrating it using the open-circuit voltage method before the start of this extreme charge to ensure the accuracy of the starting point.

[0069] S142: Integrate the total current of the battery pack at a high sampling frequency, and combine this with the rated capacity of the cells to calculate the change in SOC since the start of full charging. The high sampling frequency is typically no less than 1kHz to accurately capture instantaneous changes in current and reduce integration errors. The sign of the current (positive for charging, negative for discharging) must be considered during integration.

[0070] S143: Add the initial SOC value to the SOC change to obtain the preliminary SOC value of each cell at the current moment. At the same time, perform a reasonableness check on the preliminary SOC value. If the calculated value exceeds the theoretical range of 0% to 100%, clamp it to the boundary value and record the abnormality flag.

[0071] S2: Based on the feature parameter set, multimodal feature fusion is performed through a state estimation algorithm to obtain the accurate SOC value of each cell in the group.

[0072] Step S2, which involves obtaining the precise SOC value for each cell within the group, includes:

[0073] S21: Construct a battery state-space model with the SOC and internal resistance of each cell as state variables. This model needs to consider the ohmic polarization, electrochemical polarization, and concentration polarization effects of the cell, and its state equations and observation equations should reflect the nonlinear relationship between SOC, internal resistance, terminal voltage, and temperature. Model parameters can be obtained by performing HPPC tests on typical cells and identifying the parameters. The specific steps for constructing the battery state-space model include:

[0074] First, determine the state variables. State variables are the core parameters that describe the internal dynamic characteristics of the battery and need to be estimated. These include the state of charge (SOC), ohmic internal resistance, and polarization voltage / polarization resistance. SOC represents the remaining charge, and ohmic internal resistance reflects the instantaneous voltage drop of the cell, which is strongly correlated with its health and temperature. Polarization voltage / polarization resistance describes the dynamic characteristics of the cell and can be characterized by one or more resistor-capacitor networks to simulate relaxation effects. Therefore, the state vector can be defined as:

[0075]

[0076] Where, x k SOC represents the state vector at time k. k R0 represents the state of charge at time k. k This represents the ohmic internal resistance value at time k, where T is the transpose and k is the time index.

[0077] Secondly, state equations are established. State equations describe how state variables evolve over time. Specifically, they include:

[0078] The state equation of SOC (based on the ampere-hour integral method) is expressed as:

[0079]

[0080] Among them, i k Let represent the current at time k, η be the coulombic efficiency (typically less than 1 during charging and equal to 1 during discharging), and Δt represent the sampling time interval. Q rated This indicates the rated capacity of the battery cell.

[0081] The state equation for the ohmic internal resistance is expressed as:

[0082]

[0083] Among them, R0 k Let w be the ohmic internal resistance at time k. R0,k It is used to simulate process noise caused by slow changes in internal resistance.

[0084] Combining the state equations for SOC and ohmic resistance, we obtain the complete state equations:

[0085]

[0086] A more specific form is as follows:

[0087]

[0088] Where f is the state transition equation, w k For process noise, w SOC,k This represents process noise or process disturbance acting on the state variable SOC.

[0089] Furthermore, observation equations are established. These equations describe how the state variables are related to the external observations that we can actually measure. Specifically, they include:

[0090] Determine the observables, including at least the terminal voltage and temperature.

[0091] Based on the battery equivalent circuit model, the observation equation for the terminal voltage is determined, and the formula is expressed as:

[0092]

[0093] in, Let be the battery terminal voltage at time k. This is the open-circuit voltage, a function of SOC. The OCV-SOC correspondence needs to be obtained experimentally. k ·R0 k It is the instantaneous voltage drop caused by the ohmic internal resistance.

[0094] Temperature is introduced as an observation. Although temperature does not appear directly in the terminal voltage formula, it is a key factor affecting model parameters (especially internal resistance R0). In state estimation, temperature can be used as part of the observation vector to constrain and correct the state estimation process. Specifically, the relationship between internal resistance, temperature, and SOC is established. Therefore, the observation equation can be expressed as:

[0095]

[0096]

[0097]

[0098] Among them, y k Let h represent the observation vector, and v represent the observation function. k To observe the noise, L k Let T be the battery temperature measured at time k, and let T denote the transpose, i.e., it is a column vector.

[0099] Finally, model linearization is performed to extend the Kalman filter (EKF). This involves calculating the state transition matrix and the observation matrix, expressed by the following formula:

[0100]

[0101]

[0102] Where Fk is the state transition matrix, f is the state transition equation, and x is the state vector. Let f be the Jacobian matrix of f with respect to x, used to describe the rate of change of each output component with respect to each input component. The slope of the open-circuit voltage-state-of-charge (OCC) curve represents how much a small change in SOC will cause a change in the open-circuit voltage. It determines the "sensitivity" of the voltage observation to the SOC estimate. In the flat region of the OCV-SOC curve, this value is small, and the voltage's ability to correct for the SOC estimate is weak.

[0103] S22: Input the feature parameter set as the observation vector into the battery state space model, and perform recursive calculation using the extended Kalman filter algorithm.

[0104] S23: In each step of the recursive calculation, the predicted values ​​of the state variables are corrected using the voltage and temperature multimodal information in the observation vector. By minimizing the covariance between the predicted values ​​and the actual measured values, the optimal estimated accurate SOC value of each cell is output.

[0105] Step S23, which involves correcting the predicted values ​​of the state variables using the multimodal information of voltage and temperature in the observation vector, includes:

[0106] S231: In the correction phase of the Kalman filter algorithm, the actual measured voltage value in the observation vector is compared with the predicted value to obtain voltage information. Voltage information reflects the difference between the model-predicted voltage and the actual measured voltage, and is a key basis for state correction.

[0107] S232: Using the actual measured temperature value in the observation vector as a key constraint, the covariance matrix of process noise and observation noise in the Kalman filter algorithm is adaptively adjusted in real time. Specifically, when the temperature is high or the temperature rises rapidly, the process noise covariance is appropriately increased, indicating an increase in model uncertainty; when the temperature measurement value fluctuates greatly, the observation noise covariance is appropriately increased, reducing the confidence in the current temperature observation value.

[0108] S233: Based on the covariance matrix and voltage information, calculate the Kalman gain, and use this gain to weight and correct the predicted values ​​of the state variables, outputting the accurate SOC value. The Kalman gain determines the weighting of the predicted and observed values ​​in the final estimate. A larger gain indicates greater trust in the observed values; a smaller gain indicates greater trust in the model's predictions.

[0109] S3: Based on the precise SOC value, calculate the maximum SOC difference of the battery pack, and trigger dynamic balancing when the maximum SOC difference exceeds a first preset threshold. Once dynamic balancing is triggered, balancing control is executed. The first preset threshold can be dynamically adjusted according to the battery type and the extreme charging strategy; for example, it can be set to 5% for lithium iron phosphate batteries and 3% for ternary lithium batteries. The maximum SOC difference refers to the difference between the SOC values ​​of the highest and lowest SOC cells in the battery pack at the current moment.

[0110] Step S3, the steps for performing equalization control include:

[0111] S31: Identify the battery cell whose current SOC value is higher than a first set threshold as the target battery cell. The target battery cell is usually defined as a battery cell whose SOC value is significantly higher than the average SOC within the group or higher than a certain set upper limit. A target battery cell determination threshold can be set; for example, a battery cell whose SOC value is higher than the average SOC value within the group plus half of the first set threshold can be identified as the target battery cell.

[0112] S32: Based on the SOC difference between the target cell and the low SOC cell, and the peak charging current value, calculate the required charging pause time and bypass current ratio for the target cell, and output the control strategy. A low SOC cell typically refers to the cell with the lowest SOC value in the group, or a cell with an SOC value lower than the group's average. The calculation must consider the balance between balancing speed and cell safety.

[0113] In step S32, the control strategy includes: when multiple target cells exist, the strongest bypass shunt control is applied preferentially to the cell with the highest SOC value. At the start of equalization, a first preset bypass shunt ratio is used for rapid suppression; this first preset bypass shunt ratio represents the strongest bypass shunt control, which will not exceed the battery's rated charging current limit. When the maximum SOC difference approaches the target threshold, the shunt ratio is gradually reduced, transitioning to fine-tuning. The voltage and temperature of the target cells are monitored in real time; if their rate of increase exceeds a safety limit, the equalization operation is paused. This strategy aims to quickly converge the imbalance while avoiding overcharge and thermal runaway risks. Reaching the preset ratio of the target threshold can be defined as the maximum SOC difference entering within 20% of the first preset threshold.

[0114] Step S32, which involves calculating the required charging pause time and bypass shunting ratio for the target battery cell, includes:

[0115] S321: Set an equilibrium target and calculate the SOC difference that needs to be eliminated. The equilibrium target is usually set to reduce the maximum SOC difference to within a second set threshold (less than the first set threshold, for example, 1% to 2%), or to reach an acceptable level of imbalance. SOC difference to be eliminated = current maximum SOC difference - second set threshold.

[0116] S322: Calculate the amount of charge that needs to be transferred or offset based on the required SOC difference to be eliminated and the rated capacity of the target cell. The charge calculation is expressed as follows:

[0117]

[0118] in, For charge quantity, The SOC difference (per unit value) that needs to be eliminated. The rated capacity of the target battery cell.

[0119] S323: Calculate the theoretical charging pause time based on the charge amount and the current charging current at the current electrode. The theoretical charging pause time is expressed as:

[0120]

[0121] in, Theoretical charging pause time, The charging current is the polarity.

[0122] S324: The theoretical charging pause time is corrected based on the real-time average temperature of the battery pack to obtain the actual charging pause time. The bypass shunting ratio is then calculated based on the actual charging pause time and the maximum shunting capacity of the equalization circuit. The correction principle is as follows: In low-temperature environments, the pause time should be appropriately shortened to avoid excessive voltage rise due to increased internal resistance; in high-temperature environments, the pause time can be maintained or slightly extended, but temperature rise must be closely monitored. The correction coefficient can be found in the pre-calibrated temperature-time correction table.

[0123] Step S324, the step of calculating the bypass shunting ratio includes:

[0124] S3241: Obtains the maximum continuous shunt current that the equalization circuit can withstand. This value is determined by the rated current of the power devices (such as MOSFETs) in the equalization circuit, the heat dissipation conditions, and the system design; it is either a fixed parameter or a dynamic parameter that varies with temperature.

[0125] S3242: Based on the equalization objective, calculate the theoretical value of the shunt current required to achieve an equivalent charging pause effect. The theoretical value of the equivalent shunt current is expressed as:

[0126]

[0127] in, This is the theoretical value of the equivalent shunt current. This is the actual charging pause time after taking temperature correction into account.

[0128] S3243: Combines the theoretical value of the shunt current with the maximum shunt capacity I of the equalization circuit. bm The comparison is performed, and the final determined bypass shunt ratio and adjusted charging pause time are output, including:

[0129] If I bn ≤I bm The bypass shunting ratio is expressed as:

[0130]

[0131] If I bn >I bm The actual charging pause time will be extended according to bypass shunting ratio two, which is expressed as:

[0132]

[0133]

[0134] In the formula, I bn This is the theoretical value of the shunt current. For bypass diversion ratio 1, The bypass diversion ratio is 2. T represents the actual charging pause time. hold Theoretical charging pause time, This is the total shunt current. The bypass shunt ratio essentially controls the duty cycle of the switching devices in the equalization circuit during one switching cycle, thereby adjusting the average shunt current.

[0135] S33: Based on a control strategy, the equalization circuit connected to the target cell is controlled to achieve bypass shunting with the charging circuit based on the bypass shunting ratio during the charging pause time, while maintaining normal extreme charging for non-target cells. The equalization circuit is typically an active bypass circuit based on switching elements (such as MOSFETs) and shunt resistors. The control signal is a PWM wave, and its duty cycle is determined by the calculated bypass shunting ratio.

[0136] S4: Based on equalization control, it monitors the maximum SOC difference and cell safety parameters in real time. When the maximum SOC difference drops to meet the preset safety cutoff condition, it exits the bypass control of the target cell, allowing all cells to resume normal charging. The safety cutoff condition is to ensure that the cell will not enter a dangerous state due to equalization control under any circumstances.

[0137] In step S4, the preset safety cutoff conditions include:

[0138] During the equalization control process, the terminal voltage of the target cell is monitored in real time. If its voltage reaches or exceeds the safe voltage limit determined by a dynamic lookup table based on the current battery temperature, the safety cut-off condition is immediately met. The safe voltage limit table is formulated based on the voltage withstand characteristics of the cell's chemical system at different temperatures to prevent overcharging.

[0139] The system calculates the rate of change of the target cell's terminal voltage per unit time in real time. If the absolute value of the rate of change exceeds a preset gradient safety threshold for multiple consecutive control cycles, the safety cutoff condition is immediately met. The gradient safety threshold is used to detect abnormally sharp voltage rises, which may be a precursor to internal short circuits or other faults. Judging over multiple consecutive cycles (e.g., 3-5 cycles) improves anti-interference capability.

[0140] The surface temperature of the target battery cell is monitored in real time. If the rate of temperature rise per unit time exceeds the preset temperature rise safety threshold, the safety cutoff condition is immediately met. The temperature rise safety threshold is used to prevent local overheating and ensure battery thermal safety.

[0141] In summary, the dynamic balancing control method for electric two-wheelers based on multimodal feature fusion achieves dynamic balancing control of the battery pack during extreme charging by real-time acquisition and fusion of multimodal data such as voltage, current, and temperature, and using an extended Kalman filter algorithm for high-precision SOC estimation. This significantly improves the accuracy of SOC estimation and balancing efficiency. By employing a calculation method combining charging pause time and bypass shunting ratio, control parameters can be intelligently calculated based on the SOC difference and real-time current. Dynamic corrections considering the balancing circuit capability and temperature effects are also taken into account, enabling rapid and controllable active intervention. This effectively suppresses inconsistencies between cells, avoids the risks of overcharging or undercharging, and achieves rapid cut-off protection by combining real-time safety parameter monitoring. Ultimately, this enhances the charging safety of the battery system, extends battery life, and optimizes the overall performance of the extreme charging process.

[0142] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion, characterized in that, include: S1: Real-time acquisition of multimodal data of the electric two-wheeler battery pack under extreme charging state, and feature extraction of the multimodal data to obtain a set of feature parameters for characterizing the cell state; S2: Based on the set of feature parameters, multimodal feature fusion is performed using a state estimation algorithm to obtain the precise SOC value of each cell in the group; S3: Based on the accurate SOC value, calculate the maximum SOC difference of the battery pack, and trigger dynamic equalization when the maximum SOC difference is greater than a first set threshold; When dynamic balancing is triggered, balancing control is executed; The steps for performing equalization control include: S31: The battery cell whose current SOC value is higher than the first set threshold is identified as the target battery cell; S32: Based on the SOC difference between the target cell and the low SOC cell and the maximum charging current value, calculate the required charging pause time and bypass current ratio for the target cell, and output the control strategy. The steps for calculating the required charging pause time and bypass shunting ratio for the target battery cell include: S321: Set the equilibrium target and calculate the SOC difference that needs to be eliminated; S322: Calculate the amount of charge that needs to be transferred or offset based on the SOC difference to be eliminated and the rated capacity of the target cell; S323: Calculate the theoretical charging pause time based on the charge amount and the current charging current; S324: The theoretical charging pause time is corrected based on the real-time average temperature of the battery pack to obtain the actual charging pause time. The bypass shunting ratio is then calculated based on the actual charging pause time and the maximum shunting capacity of the equalization circuit. The steps include: S3241: Obtain the maximum continuous shunt current value that the equalization circuit can withstand; S3242: Based on the equalization target, calculate the theoretical value of the shunt current required to achieve an equivalent charging pause effect; S3243: Combine the theoretical value of the shunt current with the maximum shunt capacity of the equalization circuit. The comparison is performed, and the final determined bypass shunt ratio and adjusted charging pause time are output, including: like The bypass shunting ratio is then expressed as: like The actual charging pause time will be extended according to bypass shunting ratio two, which is expressed as: In the formula, This is the theoretical value of the shunt current. For bypass diversion ratio 1, The bypass diversion ratio is 2. This refers to the actual charging pause time. Theoretical charging pause time, This is the total electrode charging current; S33: Based on the control strategy, control the equalization circuit connected to the target cell so that it achieves bypass shunting with the charging circuit based on the bypass shunting ratio during the charging pause time, while maintaining normal extreme charging for non-target cells. S4: Based on the equalization control, monitor the maximum SOC difference and cell safety parameters in real time; when the maximum SOC difference drops to meet the preset safety cutoff condition, exit the bypass control of the target cell and restore all cells to normal extreme charging.

2. The dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion according to claim 1, characterized in that, Step S1, the steps for obtaining characteristic parameters used to characterize the cell state, include: S11: Based on the multimodal data, generate a multimodal raw data sequence; S12: Preprocess the multimodal raw data sequence, including using moving average filtering or Kalman filtering algorithms to denoise the voltage and current data, and removing outliers from the temperature data, and outputting the preprocessed data. S13: Perform real-time feature extraction on the preprocessed data. The real-time features include the instantaneous voltage value and voltage change rate of each cell, the average current value and current direction of the battery pack, and the instantaneous temperature value, temperature change rate, and maximum temperature difference of each measurement point. S14: Based on the average current value, calculate the preliminary SOC value of each cell using the ampere-hour integration method, and use the preliminary SOC value as an important internal state characteristic parameter, which together with the real-time feature constitutes the characteristic parameter set.

3. The dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion according to claim 2, characterized in that, Step S14, which involves calculating the preliminary SOC value of each cell using the ampere-hour integration method, includes: S141: Obtain the initial SOC value for each cell; S142: Integrate the total current of the battery pack at a high sampling frequency, and calculate the change in SOC after the start of full charging by combining the rated capacity of the cells. S143: Add the initial SOC value to the SOC change to obtain the preliminary SOC value of each cell at the current moment.

4. The dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion according to claim 1, characterized in that, Step S2, which involves obtaining the precise SOC value for each cell within the group, includes: S21: Construct a battery state-space model with the SOC and internal resistance of each cell as state variables; S22: Input the set of feature parameters as an observation vector into the battery state space model, and perform recursive calculation using the extended Kalman filter algorithm; S23: At each step of the recursive calculation, the predicted values ​​of the state variables are corrected using the voltage and temperature multimodal information in the observation vector. By minimizing the covariance between the predicted values ​​and the actual measured values, the optimal estimated precise SOC value for each cell is output.

5. The dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion according to claim 4, characterized in that, Step S23 includes the following calibration steps: S231: In the correction stage of the Kalman filter algorithm, the actual measured voltage value in the observation vector is compared with the predicted value to obtain voltage information; S232: Using the actual measured temperature value in the observation vector as a key constraint, the covariance matrix of process noise and observation noise in the Kalman filter algorithm is adaptively adjusted in real time. S233: Based on the covariance matrix and the voltage information, calculate the Kalman gain, and use the gain to perform weighted correction on the predicted values ​​of the state variables, and output the accurate SOC value.

6. The dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion according to claim 1, characterized in that, In step S32, the control strategy includes: when there are multiple target cells, the strongest bypass shunting control is applied to the cell with the highest SOC value first; at the start of equalization, a first preset bypass shunting ratio is used for rapid suppression; when the maximum SOC difference reaches a preset ratio of the target threshold, the shunting ratio is gradually reduced and fine-tuning is switched; the voltage and temperature of the target cells are monitored in real time, and if their rise rate exceeds the safety limit, the current equalization operation is paused.

7. The dynamic equalization control method for electric two-wheeled vehicles based on multimodal feature fusion according to claim 1, characterized in that, In step S4, the preset safety cutoff conditions include: During the equalization control process, the terminal voltage of the target cell is monitored in real time. If the terminal voltage reaches or exceeds the voltage safety upper limit determined by dynamic lookup table based on the current battery temperature, the safety cut-off condition is immediately met. The rate of change of the terminal voltage of the target cell per unit time is calculated in real time. If the absolute value of the rate of change of the terminal voltage exceeds the preset gradient safety threshold for multiple consecutive control cycles, the safety cut-off condition is immediately met. The surface temperature of the target battery cell is monitored in real time. If the rate of increase of the surface temperature within a unit time exceeds the preset temperature rise safety threshold, the safety cutoff condition is immediately met.

Images