Isolated battery cluster adaptive equalization control method and system

Abstract

The application discloses a heterogeneous battery cluster adaptive equalization control method and system, and belongs to the field of data processing, wherein the method comprises the following steps: establishing the topological connection of a heterogeneous battery cluster and a multi-port DAB converter, connecting the battery cluster to a low-voltage side full-bridge circuit, and performing phase shift control through full-bridge switching tubes; constructing a battery cluster comprehensive state vector, constructing a multi-dimensional energy difference function between any two battery clusters, and generating target power transmission demand between any two battery clusters; mapping the target power transmission demand to the corresponding initial phase shift angle difference value by calling a phase shift power transmission characteristic model; and performing multi-port coupling consistency optimization on the initial phase shift angle difference value, and controlling the equalization control difference value of the actual driving phase shift angle of each full-bridge switching tube. The application solves the technical problem that the direct parallel connection of the existing battery cluster to the DC bus will cause uncontrollable parallel circulating current due to the mismatch of terminal voltage, aggravate energy loss and device aging, and cause the risk of heat runaway.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically to an adaptive equalization control method and system for heterogeneous battery clusters. Background Technology

[0002] With the continuous expansion of application scenarios such as the reuse of cascaded batteries in energy storage systems and the hybrid networking of new and old batteries, the demand for parallel applications of heterogeneous battery clusters composed of different chemical systems, rated capacities, series-parallel configurations, and aging degradation degrees continues to increase. However, such battery clusters have inherent differences in open-circuit voltage, equivalent internal resistance, and charge-discharge voltage-current characteristics. Directly connecting them to the DC bus will generate uncontrollable parallel circulating currents due to terminal voltage mismatch, which will aggravate energy loss and device aging, and even lead to the risk of thermal runaway. Existing battery equalization control schemes are mostly designed for homogeneous battery clusters, and only use the difference in state of charge as the basis for equalization decisions. They do not incorporate multi-dimensional constraints such as internal resistance loss characteristics, temperature safety boundaries, and dynamic changes in the state of charge. At the same time, they do not consider the magnetomotive force coupling effect caused by the shared high-frequency magnetic circuit of multi-port DAB converters. The phase shift control quantities solved independently in pairs are prone to power transmission deviation, control oscillation, and transformer magnetic saturation problems, making it difficult to adapt to the high-precision and high-stability equalization control requirements in heterogeneous scenarios. Summary of the Invention

[0003] This application provides an adaptive equalization control method and system for heterogeneous battery clusters, aiming to solve the technical problem that existing battery clusters directly connected in parallel to a DC bus will generate uncontrollable parallel circulating currents due to terminal voltage mismatch, which will aggravate energy loss and device aging, and cause the risk of thermal runaway.

[0004] In view of the above problems, this application provides an adaptive equalization control method and system for heterogeneous battery clusters.

[0005] The first aspect disclosed in this application provides an adaptive equalization control method for heterogeneous battery clusters, the method comprising: A topology connection is established between heterogeneous battery clusters and a multi-port DAB converter. The multi-port DAB converter includes multiple low-voltage-side full-bridge circuits, with each battery cluster connected to one low-voltage-side full-bridge circuit. Each low-voltage-side full-bridge circuit is phase-shifted via full-bridge switches. Multiple battery cluster operating state parameters are acquired to construct a comprehensive battery cluster state vector. Based on this comprehensive state vector, a multidimensional energy difference function is constructed between any two battery clusters to generate the target power transfer requirement between them. The multi-port DAB converter maps the target power transfer requirement to a corresponding initial phase shift angle difference by calling a phase-shift power transfer characteristic model. Multi-port coupling consistency optimization is performed on the initial phase shift angle difference to generate a phase shift angle equalization control difference used to control the actual driving of each full-bridge switch.

[0006] Another aspect of this application discloses an adaptive equalization control system for heterogeneous battery clusters, the system comprising: The control module establishes a topology connection between the heterogeneous battery clusters and the multi-port DAB converter. The multi-port DAB converter includes multiple low-voltage-side full-bridge circuits, connecting each battery cluster to one low-voltage-side full-bridge circuit. Each low-voltage-side full-bridge circuit performs phase-shift control through full-bridge switches. The acquisition module acquires multiple battery cluster operating state parameters to construct a comprehensive battery cluster state vector. Based on the comprehensive battery cluster state vector, a multi-dimensional energy difference function between any two battery clusters is constructed to generate the target power transfer demand between any two battery clusters. The mapping module maps the target power transfer demand to the corresponding initial phase-shift angle difference by calling the phase-shift power transfer characteristic model. The optimization module performs multi-port coupling consistency optimization on the initial phase-shift angle difference to generate a phase-shift angle equalization control difference for controlling the actual driving of each full-bridge switch.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By constructing an isolated topology connection between heterogeneous battery clusters and a multi-port DAB converter, parallel circulating currents are physically blocked. A multi-dimensional energy difference function is established, which integrates the differences in state of charge, internal resistance loss weights, temperature constraints, and dynamic trends, to generate the target power transmission between clusters. Based on the phase-shift power transmission characteristic model that takes into account the voltage ratio imbalance, the power demand is accurately mapped to the phase-shift control quantity. The consistency correction of global phase-shift control is achieved through multi-port phase-shift coupling matrix and step-size feedforward iterative optimization. Ultimately, the control accuracy, operational stability, and overall service life of the heterogeneous battery cluster equalization system are improved.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0009] Figure 1 A flowchart illustrating the adaptive equalization control method for heterogeneous battery clusters is provided for embodiments of this application; Figure 2 A schematic diagram of the structure of the heterogeneous battery cluster adaptive equalization control system is provided for the embodiments of this application; Figure 3 A schematic diagram illustrating the multi-port optimization effect of the adaptive equalization control method for heterogeneous battery clusters is provided for embodiments of this application.

[0010] Explanation of reference numerals in the attached diagram: Control module 11, Acquisition module 12, Mapping module 13, Optimization module 14. Detailed Implementation

[0011] 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, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0012] The overall concept of the technical solution provided in this application is as follows: This application provides an adaptive equalization control method and system for heterogeneous battery clusters. By establishing an isolated topology connection between the heterogeneous battery cluster and the multi-port DAB converter to block parallel circulating current paths, a multi-dimensional energy difference function integrating SOC difference, internal resistance loss weights, temperature penalties, and dynamic trends is constructed to generate the target transmission power. A phase-shift power transmission characteristic model considering voltage ratio imbalance is used to achieve accurate mapping from power to phase angle. Furthermore, consistency optimization of the phase-shift control quantity is achieved through multi-port coupling matrices and stepped-step feedforward iterations, ultimately improving the control accuracy, operational stability, and overall lifespan of the heterogeneous battery cluster equalization system.

[0013] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0014] Example 1, as Figure 1 As shown in the embodiments of this application, an adaptive equalization control method for heterogeneous battery clusters is provided, the method comprising: S100: Establish a topological connection between the heterogeneous battery cluster and the multi-port DAB converter. The multi-port DAB converter includes multiple low-voltage side full-bridge circuits. Each battery cluster is connected to a low-voltage side full-bridge circuit. Each low-voltage side full-bridge circuit is phase-shifted through a full-bridge switch.

[0015] Specifically, heterogeneous battery clusters refer to multiple independent battery energy storage units with significant differences in cell chemistry, rated capacity, series-parallel configuration, and aging degradation. These battery clusters have inherent deviations in open-circuit voltage, equivalent internal resistance, and charge-discharge voltage-current characteristics. If directly connected in parallel to the same DC bus, uncontrollable parallel circulating currents will form due to the voltage difference, leading to device overheating, unnecessary energy loss, and even thermal runaway risks. Multi-port DAB converters, or multi-port bidirectional active bridge converters, employ an integrated winding with a shared high-frequency magnetic circuit, allowing simultaneous access to multiple low-voltage DC ports and supporting flexible bidirectional energy transfer between any two ports. It is the core power conversion carrier for realizing the energy balance regulation of multi-cluster heterogeneous batteries. Its hardware structure is configured with multiple low-voltage side full-bridge circuits corresponding to the number of battery clusters. The circuit is composed of four power semiconductor switching devices arranged in an H-shaped bridge arm topology. It can seamlessly switch between inverter and rectification modes. In inverter mode, it converts the DC power output from the battery cluster into a high-frequency AC square wave and sends it into the common magnetic circuit. In rectification mode, it converts the high-frequency AC energy transmitted by the magnetic circuit into DC power and charges the corresponding battery cluster. It is a dedicated power conversion interface for each battery cluster to access the equalization system. Moreover, the DC side of each low-voltage side full-bridge circuit is independent of each other and has no direct electrical connection.

[0016] When implementing the physical topology connection, the positive and negative DC output terminals of each heterogeneous battery cluster are respectively connected to the DC side port of an independent low-voltage full-bridge circuit. The DC loops of all battery clusters are electrically isolated from each other, and energy coupling is achieved only through the common high-frequency magnetic circuit inside the multi-port DAB converter. This directly blocks the parallel circulating current path caused by the voltage mismatch between the heterogeneous battery clusters at the physical structure level.

[0017] After completing the hardware wiring and electrical connections, an independent gate drive channel is configured for each low-voltage side full-bridge circuit. By adjusting the turn-on and turn-off timing of the full-bridge switching transistors, which are the core power semiconductor actuators constituting the full-bridge circuit, phase-shift control is performed. The controller sets a unified switching operating frequency and a global reference clock signal for all low-voltage side full-bridge circuits. The high-frequency AC square wave output by each full-bridge circuit uses this reference clock as the phase reference zero point. By adjusting the phase offset of a certain full-bridge drive signal relative to the reference signal or between any two different full-bridge drive signals, the phase difference of the voltage vector on both sides of the corresponding winding of the high-frequency transformer can be changed, thereby precisely controlling the amplitude and direction of power transmission between two corresponding battery clusters. The battery cluster with the leading phase outputs energy to the battery cluster with the lagging phase, and the absolute value of the phase difference is positively correlated with the magnitude of the transmitted power. This achieves continuous adjustment of the power flow direction and transmission intensity.

[0018] S200: Obtain multiple battery cluster operating state parameters of the heterogeneous battery cluster to construct a comprehensive battery cluster state vector, and construct a multidimensional energy difference function between any two battery clusters based on the comprehensive battery cluster state vector to generate the target power transmission requirement between any two battery clusters.

[0019] Specifically, firstly, the battery management system associated with each battery cluster collects real-time operational data for the entire cluster, constructing a comprehensive state vector for the battery cluster. This vector is a data structure that uniformly vectorizes the multi-dimensional energy storage, loss characteristics, safety boundaries, and dynamic trends of a single battery cluster, used to comprehensively depict the overall operation of the battery cluster. For the i-th battery cluster, its comprehensive state vector is denoted as: ; Among them, the state of charge The ratio of the remaining capacity to the rated capacity of the battery cluster directly reflects the existing energy level; equivalent internal resistance parameter. This represents the equivalent total value of the DC internal resistance and connection impedance of the battery cluster cells, reflecting the Joule loss characteristics during energy transfer; temperature parameters. The average operating temperature of the battery cluster cells characterizes the thermal safety boundary conditions; the rate of change of state of charge. It represents the change in SOC per unit time, reflecting the dynamic increase or decrease trend of the battery cluster's energy.

[0020] After constructing the state vector of a single cluster, a multidimensional energy difference function is constructed and data cleaned for any two battery clusters i and j. This function integrates four types of influencing factors: power difference, loss characteristics, safety constraints, and dynamic trends. It is used to quantify the mathematical model for calculating the target transmission power between the two clusters. The process consists of three steps: The first step is to perform state vector difference calculation, subtracting the corresponding dimensions of the combined state vectors of the two battery clusters one by one to obtain the basic state difference vector. ; The initial extraction of raw differences between the two clusters across four dimensions completes the preliminary extraction of the difference data. The second step involves normalization and cleaning. Since the physical dimensions and value ranges of the four-dimensional parameters are completely different, direct weighted calculation is not possible. Therefore, an extreme value normalization method is used to standardize and map each component of the basic state difference vector. The normalization formula is as follows: ; in To define the reasonable fluctuation range for the corresponding parameters across the entire range, for example, the SOC value range is 0~1, with a maximum fluctuation range of 1; the reasonable fluctuation range for the equivalent internal resistance is 0.5~2 times the rated internal resistance, with a maximum fluctuation range of 1.5 times the rated internal resistance; the reasonable operating range for temperature is -20℃~60℃, with a maximum fluctuation range of 80℃; the maximum range for the SOC change rate is the change rate corresponding to a 1C multiplier. Thus, all the differences across dimensions are uniformly mapped to the [0,1] interval, resulting in a standardized state difference vector that eliminates dimensional differences and possesses weighted scalability. ; The first step involves cleaning and standardizing the differential data. The second step extracts multi-dimensional energy difference terms and performs function fitting, which includes four core difference terms: First, the SOC difference term, which is the core benchmark driving energy balance, directly corresponding to the difference in remaining charge between the two clusters and contributing to the basic balance power. Second, the internal resistance loss weight term, used to characterize the difference in equivalent loss capacity of battery clusters during energy transmission. Battery clusters with larger internal resistance differences experience higher additional Joule losses due to high current transmission. Therefore, this term is used to correct the loss in the balance power, preventing excessive heating and aging of high-internal-resistance clusters. Third, the temperature constraint penalty term, used to characterize the limiting effect of battery cluster thermal safety boundary constraints on energy transmission. When the battery temperature exceeds the optimal operating range of 25℃~40℃, the balance power is gradually reduced according to the degree of temperature deviation. When the safety threshold of 55℃ or -10℃ is reached, the penalty coefficient is directly set to zero, and energy transmission is prohibited, ensuring thermal safety throughout the process. Fourth, the dynamic change trend term, used to characterize the predictive impact of the battery cluster state of charge change rate on future energy balance demand. This can predict the widening trend of SOC difference and adjust the balance power in advance to achieve proactive balance control.

[0021] The multidimensional energy difference function, obtained by fitting four types of difference terms, ultimately outputs the target power transmission demand, and its mathematical expression is as follows: ; in For the rated equalization power of a single cluster, These are the weight coefficients for the corresponding dimensions. This is a temperature penalty function, with a value range of [0,1]. The closer the temperature is to the safety boundary, the smaller the function value becomes. The final calculated value is... This represents the target power transfer requirement between the i-th cluster and the j-th cluster. Its positive and negative signs represent the power transfer direction, and its absolute value represents the magnitude of the power transferred. By traversing all battery cluster pairs, the complete inter-cluster target power transfer matrix can be obtained.

[0022] S300: The multi-port DAB converter maps the target power transmission requirement to the corresponding initial phase shift angle difference by calling the phase shift power transmission characteristic model.

[0023] Specifically, the core concept is first clarified: the phase-shifted power transmission characteristic model is a mathematical mapping model embedded in the multi-port DAB converter controller, describing the quantitative correspondence between the phase shift angle difference between any two low-voltage side full-bridge ports, the port voltage ratio, and the actual transmitted power. It serves as the core bridge connecting energy balance decision-making and hardware-driven control, specifically addressing the insufficient accuracy of traditional fixed-ratio DAB power analysis formulas in scenarios with inconsistent terminal voltages of heterogeneous battery clusters. Voltage ratio imbalance is a core parameter characterizing the impact of heterogeneous battery cluster terminal voltage differences on the transformer's magnetic circuit matching. Due to inherent differences in the terminal voltages of different battery clusters, the corresponding high-frequency square wave amplitudes of the full-bridge output differ. The degree to which the actual voltage ratio deviates from the rated design ratio when converted to the transformer winding side is the voltage ratio imbalance, a key interference variable affecting power transmission accuracy in heterogeneous scenarios. The model is built based on the classic analytical formula for dual active bridge power transmission as its underlying mechanism framework. For any two low-voltage side full-bridge ports i and j, the analytical expression for transmitted power under ideal operating conditions is: ; in These are the DC terminal voltages of the battery clusters corresponding to the two ports. To achieve a uniform switching frequency for all bridge switches, This is the equivalent series inductance of the transformer windings at the two ports. This represents the phase shift angle difference between the two full-bridge drive signals.

[0024] Based on this, a quantitative calculation of the voltage ratio imbalance is carried out. First, the AC voltage port turns ratio equations between each low-voltage side full-bridge circuit are established, and the number of turns of the transformer winding corresponding to the i-th port is set as follows. The number of turns in the high-voltage busbar winding is Then the rated transformation ratio of this port is In actual operation, the voltage ratio imbalance between the two ports is defined as: ; The larger this value, the worse the voltage matching between the two ports, and the higher the calculation error of the ideal analytical formula. To correct the model bias caused by heterogeneous imbalance, a hybrid modeling approach combining mechanistic framework and data correction is adopted. First, a training sample data acquisition and cleaning process is performed. The training sample data comes from offline simulation tests and online data acquisition during actual operation of the multi-port DAB converter. Each sample includes the state of charge (SOC), equivalent internal resistance parameters, port DC voltage, temperature parameters, applied phase shift control values, and corresponding measured actual transmission power for each battery cluster. Sample cleaning is performed in three steps: the first step is time series alignment, using the issuance time of the phase shift control command as a unified time reference. The first step involves linear interpolation, which aligns sensor data from different sampling frequencies (voltage, power, etc.) to the same timestamp, constructing a one-to-one training sample pair corresponding to voltage ratio imbalance, phase shift angle difference, and actual transmitted power. The second step involves outlier removal, using the 3σ criterion to eliminate transient jumps in power samples exceeding three standard deviations. Invalid operating condition samples with voltages exceeding the normal operating range of the battery cluster or temperatures exceeding safety thresholds are also filtered out to prevent outlier data from interfering with the convergence direction and accuracy of the model training. The third step involves hierarchical normalization, classifying samples into mild imbalance categories based on the magnitude of the voltage ratio imbalance. Moderate imbalance Severe imbalance Three levels are used to perform min-max normalization on the power and phase shift angle data in each level and map them to the [0,1] interval. This eliminates the difference in sample size between different imbalance intervals, avoids model training biased towards severely imbalanced conditions, and ensures consistent mapping accuracy across the entire operating range.

[0025] After sample cleaning, a nonlinear correction term is introduced based on the ideal analytical formula to construct the complete model structure to be trained. The model expression is: ; Where a and b are the nonlinear correction coefficients to be trained, used to compensate for nonlinear power deviations such as circulating current loss and soft-switching failure caused by voltage mismatch; then, an objective function characterizing the deviation between the target power transmission demand and the actual output power is established, and a loss function in the form of mean square error is adopted: ; In the formula, N is the total number of training samples. The predicted power output by the model based on the phase shift angle and imbalance degree of the k-th sample group. To correspond to the actual measured transmission power, the training process employs an adaptive gradient descent algorithm to iteratively optimize the correction coefficients a and b, as well as the equivalent inductance calibration value. After each full-sample iteration, the prediction error is calculated using an independent validation set. Model convergence is determined when the error decreases by less than 0.1% over 10 consecutive iterations, resulting in a phase-shifted power transmission characteristic model with fixed parameters. The converged model parameters are then downloaded to the local controller of the multi-port DAB converter for deployment. During actual operation and model invocation, the controller will use the previously generated target power transmission requirements. As a response target, the voltage ratio imbalance calculated in real time will be used Using the phase shift angle difference as the input variable and the solution variable as the solution variable, the ideal initial phase shift angle difference, without considering multi-port mutual coupling interference, is calculated by inversely solving the trained phase shift power transmission characteristic model. After traversing all battery cluster pairs, an initial set of phase shift angle differences is formed, which serves as the starting point for subsequent multi-port coupling optimization iterations.

[0026] S400: Perform multi-port coupling consistency optimization on the initial phase shift angle difference identification to generate phase shift angle equalization control difference for controlling the actual drive of each full-bridge switch.

[0027] Specifically, the optimization of multi-port coupling consistency is first clarified. Since all low-voltage side full-bridge circuits of the multi-port DAB converter share the same high-frequency transformer core magnetic circuit, any phase shift control action of the full bridge will induce additional voltage on all other full-bridge ports through the superposition effect of the magnetomotive force of the common magnetic circuit, thereby interfering with the actual transmission power of the corresponding battery cluster pair. As a result, the initial phase shift angle calculated independently in pairs cannot simultaneously meet the power target of all cluster pairs. This optimization process is to perform global coupling compensation correction on the initial phase shift angle so that the actual transmission power of all battery cluster pairs synchronously matches the target requirements, and achieves consistent convergence of the global equilibrium state.

[0028] Specifically, firstly, the initial phase shift angle difference set corresponding to all battery clusters is obtained. This set is obtained by independently solving the pairwise inverse of the preceding phase shift power transmission characteristic model. Directly putting it into operation will cause power deviation, power oscillation, or even transformer magnetic saturation problems due to coupling interference. Therefore, it is necessary to perform quantitative modeling of coupling characteristics and construct a multi-port phase shift coupling matrix. This matrix is ​​a mathematical carrier that quantitatively characterizes the degree of influence of each battery cluster's phase shift control action on the power transmission of all other battery clusters. For a multi-port DAB system containing n low-voltage side full-bridge ports, an n-order square matrix coupling matrix M is constructed, where the diagonal elements are... The power-phase shift main sensitivity of the i-th port is the change in power transmitted through the port caused by a unit change in phase shift angle. It is calculated based on the DAB power transfer mechanism using parameters such as port voltage, equivalent leakage inductance, and switching frequency. (Off-diagonal elements are not included in this definition.) The coupling influence coefficient of the phase shift action of the j-th port on the i-th port is characterized by the power deviation amplitude caused by the superposition of the magnetomotive force of the common magnetic circuit at the i-th port when the j-th port changes by a unit phase shift angle. Its value is derived from the magnetic circuit parameters such as the number of turns of the transformer winding, the excitation inductance, and the spatial arrangement of the winding, and fully reflects the magnetic circuit coupling strength between any two ports.

[0029] Subsequently, the coupling interference coefficient between each battery cluster pair was calculated based on the multi-port phase-shifting coupling matrix. The off-diagonal elements in the coupling matrix were normalized according to the principal sensitivity of the corresponding port to obtain the dimensionless coupling interference coefficient. This coefficient directly reflects the proportion of interference between the control action of port j and the power transmission of port i. The larger the coefficient, the stronger the cross-influence.

[0030] Next, all battery clusters are grouped according to the amplitude of the coupling interference coefficient. A first preset threshold is set as the convergence criterion and a second preset threshold is set as the strong coupling boundary criterion. Battery clusters with coupling interference coefficients greater than the second preset threshold are classified into the first group of strong coupling clusters. These clusters usually correspond to ports with adjacent winding positions and similar number of turns. Cross interference is the most significant, and improper control can easily cause power oscillations. Battery clusters with coupling interference coefficients between the first and second preset thresholds are classified into the second group of medium coupling clusters. These clusters have limited cross interference and have room for rapid adjustment.

[0031] After grouping, a feedforward iterative correction is performed under tiered step size constraints. The phase difference correction step size is the adjustment range of the phase angle difference value in each iteration. If the step size is too large, it can easily cause system oscillation or even magnetic saturation runaway. If the step size is too small, the equilibrium convergence speed will be too slow. Therefore, two sets of step size constraint intervals are configured to achieve adaptation. The first step size constraint interval is a small-amplitude correction interval, and its upper limit is less than the lower limit of the second step size constraint interval. It is specifically matched with the first group of strongly coupled clusters. By limiting the single correction amplitude, it avoids the strong coupled system from iterative divergence and ensures that the correction process is stable and controllable. The second step size constraint interval is a large-amplitude correction interval, which is matched with the second group of moderately coupled clusters. The larger correction step size accelerates the convergence speed and balances optimization efficiency and operational stability.

[0032] The iterative process employs a feedforward correction mechanism. First, based on the current set of phase angle difference values ​​and the coupling matrix, the total amount of coupling interference superimposed on each battery cluster is calculated and converted into the corresponding power deviation value. Then, the theoretical correction amount of the phase angle is derived in reverse through power-phase shift sensitivity. Subsequently, according to the step size constraint interval corresponding to the group to which the cluster pair belongs, the theoretical correction amount is limited to the corresponding interval range to generate the actual phase difference correction step size. The correction step size is then superimposed on the current phase angle difference value to complete one round of iterative correction. After that, the coupling interference coefficient under the new phase angle is recalculated, and it is verified whether the coupling interference coefficient of all cluster pairs is less than the first preset threshold. If the convergence condition is not met, the above correction process is repeated until the coupling interference coefficient of all cluster pairs is lower than the convergence threshold, indicating that the influence of multi-port cross-coupling has been suppressed to an acceptable accuracy range. At this time, the output phase angle difference value is the convergence effect. Figure 3 As shown.

[0033] Furthermore, in the method provided in the application embodiment, a multidimensional energy difference function is constructed between any two battery clusters based on the comprehensive state vector of the battery cluster. The method includes: performing state vector difference calculation on any two battery clusters according to the comprehensive state vector of the battery cluster to obtain a basic state difference vector, wherein the comprehensive state vector of the battery cluster includes the state of charge (SOC), equivalent internal resistance parameter, temperature parameter, and rate of change of state of charge; normalizing the basic state difference vector to obtain a standardized state difference vector; obtaining a multidimensional energy difference term based on the standardized state difference vector; fitting a multidimensional energy difference function with the multidimensional energy difference term, wherein the multidimensional energy difference function is used to output the target power transmission requirement of the target battery cluster, wherein the multidimensional energy difference term includes at least a SOC difference term, an internal resistance loss weight term, a temperature constraint penalty term, and a dynamic change trend term.

[0034] Specifically, the comprehensive state vector of a battery cluster is first defined as a multi-dimensional vectorized representation of the operating state of a single battery cluster. It integrates four core parameters: energy storage, loss characteristics, safety boundary, and dynamic trend. For the i-th battery cluster, its comprehensive state vector is denoted as: ; in State of charge reflects the proportion of the battery cluster's remaining charge to its rated capacity; The equivalent internal resistance parameter is the total equivalent value of the cell's DC internal resistance and the impedance of the connecting line, which determines the level of Joule loss during energy transmission. The average temperature parameter of the battery cluster characterizes the thermal safety boundary conditions. The rate of change of state of charge reflects the rate of increase or decrease of electrical charge per unit time, embodying the dynamic trend of energy change.

[0035] Based on this, state vector difference calculation is performed: For any two battery clusters i and j to be balanced, the combined state vectors of the two are subtracted bit by bit along the corresponding dimensions to obtain the basic state difference vector. This vector directly reflects the original difference between the two clusters in the four physical dimensions, and its calculation formula is as follows: ; in It represents the absolute difference between the two states of charge. This is the equivalent internal resistance difference. This is the temperature difference value. The difference in the rate of change of state of charge is used to initially extract the original differences between the two clusters in all dimensions after the differential calculation is completed. However, due to the great differences in the physical dimensions and value ranges of each parameter, such as the SOC value of 0~1, the internal resistance unit of milliohms, and the temperature unit of degrees Celsius, it is not possible to directly perform weighted summation calculation. Therefore, standardized cleaning must be performed.

[0036] The basic state difference vector is then normalized: an extreme value normalization method is used to perform dimensionless mapping on each component, resulting in a standardized state difference vector. All components of this vector are constrained within the interval [0,1], eliminating dimensional and magnitude differences and providing a comparable and weighted unified benchmark for differences in different dimensions. The normalization calculation formula is as follows: ; In the formula The original difference for the corresponding dimension is taken as the absolute value to ensure that the difference is non-negative. This represents the maximum reasonable fluctuation range of this parameter under all system operating conditions, where SOC dimension... This corresponds to a full-scale range difference of 0% to 100%; equivalent internal resistance dimension The difference between the maximum and minimum internal resistance of the battery clusters within the system is taken, typically 1.5 times the rated internal resistance; temperature dimension The allowable operating temperature range of the battery cluster is typically 80℃, corresponding to an operating range of -20℃ to 60℃; SOC change rate dimension. The SOC change rate corresponding to a 1C charge / discharge rate is taken, and after being normalized dimension by dimension, a standardized state difference vector is obtained: .

[0037] After standardization, four core multidimensional energy difference terms are extracted based on the standardized state difference vector. Each term corresponds to a constraint or objective of the equilibrium control. Finally, the multidimensional energy difference function is fitted through weighted and coupled operations. The output of this function is the target power transmission requirement between the two clusters. Its physical meaning is the magnitude and direction of the equilibrium power that should be transmitted between the two clusters after considering the power difference, loss cost, safety constraints, and trend prediction. The specific meanings and calculation methods of the four difference terms are as follows: First, the SOC difference term is the core benchmark term driving energy equilibrium, directly corresponding to the difference in the remaining power of the two clusters, with a value of... The first term determines the base amplitude of the equalization power; the greater the SOC difference, the higher the base equalization power. The second term is the internal resistance loss weighting term, which characterizes the difference in the equivalent loss capacity of the battery cluster during energy transfer. The greater the internal resistance difference, the higher the additional Joule loss brought by high-current equalization and the lower the energy utilization efficiency. Therefore, this term is used as a loss correction coefficient to reduce the base power. The calculation method is as follows: ; in The first is the internal resistance weighting coefficient, ranging from 0 to 1. The greater the difference in internal resistance, the smaller the reduction coefficient, preventing high-internal-resistance battery clusters from experiencing increased heating and aging due to high current equalization. The second is the temperature constraint penalty term, used to characterize the limiting effect of the battery cluster's thermal safety boundary constraints on energy transfer. The constraint benchmark is the side of the two clusters whose temperature deviates more from the optimal operating range. When the battery temperature is within the optimal range of 25℃ to 40℃, the penalty coefficient is 1, decreasing linearly with the degree of temperature deviation from the optimal range. When the safety threshold of -10℃ or 55℃ is reached, the penalty coefficient is set to 0, completely prohibiting energy transfer. This term is denoted as... The value range is [0,1], ensuring thermal safety throughout the equalization process; fourthly, the dynamic trend term is used to characterize the predictive impact of the battery cluster state of charge change rate on future energy balance demand. It predicts the expansion or contraction trend of future energy difference based on the current SOC change rate, achieving advanced equalization control. The value is... ,in This is a trend weighting coefficient. If the current trend will increase the SOC difference, this term will be added to increase the equalization power; otherwise, it will decrease the equalization power. Based on the above four types of difference terms, the final multidimensional energy difference function is obtained, and the formula for calculating the target power transmission demand is: ; in The rated equalization power of a single-port DAB is used to map the dimensionless difference value to the actual power amplitude. The positive and negative signs of the calculation result represent the power transmission direction. Positive values ​​indicate that energy flows from cluster i to cluster j, while negative values ​​indicate the opposite direction. The absolute value represents the magnitude of the equalization power. By repeating the above calculation for all battery cluster pairs, the global inter-cluster target power transmission matrix can be obtained.

[0038] Furthermore, in the method provided in the application embodiment, the target power transfer requirement between any two battery clusters is input to the multi-port DAB converter. The multi-port DAB converter maps the target power transfer requirement to the corresponding phase shift angle difference by calling the phase shift power transfer characteristic model. The phase shift power transfer characteristic model includes: establishing the AC voltage port ratio equation between each low-voltage side full-bridge circuit, and identifying the voltage ratio imbalance caused by the voltage mismatch at the terminals of the target battery clusters. The phase shift power transfer characteristic model takes the target power transfer requirement of the target battery clusters as the response target, the voltage ratio imbalance as the input variable, and controls the phase shift angle difference of the corresponding full-bridge switch of the target battery clusters as the solution variable for solving the problem, and outputs the phase shift angle difference for the target battery clusters.

[0039] Specifically, the phase-shifted power transmission characteristic model is first clarified as a mathematical mapping model that quantitatively maps the phase shift angle difference between any two low-voltage side full-bridge ports of a multi-port DAB converter, the port voltage ratio, and the actual transmitted power. It serves as the core bridge connecting energy balance decisions and hardware-driven control. The AC voltage port ratio equation is constructed first. This equation describes the proportional relationship between the high-frequency AC voltages output from the two low-voltage side full-bridges after being converted by the transformer windings. It forms the basis for quantifying the degree of heterogeneous voltage mismatch. For any two low-voltage side full-bridge ports corresponding to battery clusters i and j, let the number of turns in the low-voltage windings of the transformers corresponding to the two ports be respectively... The DC terminal voltages of the battery clusters, collected in real time by the voltage sampling circuit, are respectively Since the full-bridge circuit uses bipolar phase-shift control, the synchronous conduction of the diagonal switches directly applies DC voltage to both ends of the transformer winding. Therefore, the amplitude of the high-frequency square wave voltage across the winding is exactly equal to the DC terminal voltage of the corresponding battery cluster. Based on this, the equivalent excitation voltage referred to a single turn winding can be derived as follows: Under ideal voltage matching conditions, the single-turn excitation voltages at both ports are completely equal, which satisfies the condition. At this point, the port voltage ratio is exactly the same as the winding turns ratio, and the DAB power transfer characteristics conform to classical analytical laws. However, in heterogeneous scenarios, the terminal voltages of different battery clusters have inherent differences, which cannot meet the ideal matching conditions. Therefore, the voltage ratio imbalance is defined. The formula for quantifying the degree to which actual operating conditions deviate from the ideal matching state is as follows: ; A larger value indicates a more severe mismatch in the excitation voltage between the two ports, resulting in a higher calculation error in the traditional ideal power formula. It also leads to additional circulating current losses and a narrowing of the soft-switching range, making it a core interference variable affecting power transmission accuracy in heterogeneous scenarios. After completing the imbalance calculation, the model inverse solution stage is entered. The phase-shifted power transmission characteristic model is based on the classic dual active bridge power transmission mechanism, incorporating the nonlinear correction term for voltage ratio imbalance to construct a complete mapping relationship. Its forward power expression is: ; in To achieve a uniform switching frequency for all bridge switches, This is the equivalent series leakage inductance of the corresponding windings at the two ports. The phase shift angle difference between the two full-bridge drive signals, i.e. the phase offset of the two full-bridge output high-frequency square waves, is the core control quantity for DAB control of power transmission. The magnitude of the phase difference determines the magnitude of the transmitted power, and the phase lead-lag relationship directly determines the direction of power transmission. a and b are pre-calibrated nonlinear correction coefficients for imbalance, used to compensate for circulating current loss and power deviation caused by voltage mismatch, and to correct the model accuracy in heterogeneous scenarios.

[0040] The solution process is based on the target power transfer requirements of the target battery cluster pair. In response to the target, the voltage turns ratio imbalance is calculated in real time. As input variables, the phase shift angle difference As the only solution variable, the forward power formula is first rearranged to be about The standard form of a quadratic equation in one variable: ; The equation was then solved using the quadratic formula, yielding two theoretical roots. A value within this range was then selected. The reasonable solution within the interval is taken as the final output. The transmission power within this interval increases monotonically with the phase shift angle, and it can ensure that the full-bridge switch works in the soft switching range of zero voltage turn-on, avoiding excessive turn-on and turn-off losses of the device. The final output calculation result is the initial phase shift angle difference corresponding to the battery cluster pair. After repeating the above calculation for all battery cluster pairs, a complete set of initial phase shift angle differences is formed, which serves as the iterative starting point for multi-port coupling consistency optimization.

[0041] Furthermore, in the method provided in the application embodiment, the method for training the phase-shifted power transfer characteristic model includes: acquiring training sample data based on the operation process of the multi-port DAB converter, wherein the training sample data includes at least the state of charge (SOC), equivalent internal resistance parameters, port voltage parameters, temperature parameters, and corresponding actual phase-shifting angle control data of each battery cluster; performing time series alignment processing on the training sample data to construct training sample pairs reflecting the correspondence between voltage ratio imbalance and DAB power transfer behavior; establishing an objective function based on the training sample pairs, wherein the objective function is used to characterize the deviation between the target power transfer demand and the actual output power; updating the parameters of the phase-shifted power transfer characteristic model by minimizing the objective function until a converged phase-shifted power transfer characteristic model is obtained, and downloading the converged phase-shifted power transfer characteristic model to the multi-port DAB converter.

[0042] Specifically, firstly, multi-dimensional operational data is collected simultaneously during offline calibration testing and actual online operation of the multi-port DAB converter to form training sample data. This dataset covers full-scenario operational data under different SOC levels, different internal resistance decay degrees, different temperature conditions, and different voltage matching states. A single set of original samples contains five types of parameters: First, SOC, which reflects the current remaining charge level of the battery cluster and determines the open-circuit voltage and load-bearing terminal voltage characteristics of the battery cluster; second, equivalent internal resistance parameters, which characterize the ohmic loss and polarization loss levels inside the battery cluster, dynamically changing with temperature, SOC, and aging degree, and are key variables affecting terminal voltage drop and power loss; third, port voltage parameters, i.e., the real-time DC terminal voltage of the battery cluster connected to the low-voltage side full bridge, which is the direct input quantity that determines the port transformation ratio; fourth, temperature parameters, including the battery cluster cell temperature and the junction temperature of the power switch, which affect the internal resistance characteristics and device conduction voltage drop; and fifth, actual phase shift angle control data, i.e., the actual phase shift angle difference applied to the full bridge switch, which is the control input quantity for power transmission.

[0043] Simultaneously, the actual transmission power of the DAB port under the corresponding operating conditions is collected as the calibration true value of the model output. After the original data acquisition is completed, time series alignment processing is performed. This processing is a data cleaning process to solve the problems of inconsistent sampling frequencies and misaligned sampling times of multi-source sensor data. Since parameters such as voltage, current, temperature, and SOC come from different sampling channels with sampling frequencies ranging from 100Hz to 10kHz, and the phase shift control command is a discrete step signal, direct pairing would cause sample errors due to time misalignment. Therefore, the issuance time of the phase shift angle control command is used as a unified time reference, and all other sampled data are resampled using linear interpolation to align all parameters to the same timestamp, ensuring that the control input and state response in each sample are at the same operating time. Based on this, the voltage ratio imbalance of the corresponding battery cluster pair is calculated through port voltage and winding turns. Finally, a one-to-one training sample pair is constructed with voltage ratio imbalance, phase shift angle difference, and actual transmission power. This sample pair directly reflects the quantitative mapping relationship between phase shift control and power transmission under different voltage mismatch degrees and is the core data foundation for model training.

[0044] After completing the sample construction, an objective function for parameter optimization is established. This function quantifies the deviation between the model's predicted power and the actual output power, and is the core evaluation indicator guiding model parameter updates. The objective function is constructed using the mean squared error form, and its expression is: ; Where N is the total number of training samples. This represents the measured actual transmission power corresponding to the k-th sample group. The predicted power is calculated based on the imbalance and phase shift angle of the k-th sample in the phase-shifted power transmission characteristic model. The set of model parameters to be optimized includes unbalance nonlinearity correction coefficients a and b, and the equivalent series leakage inductance calibration value. The correction coefficient is used to compensate for the circulating current loss caused by voltage mismatch and the power deviation caused by soft switching offset, while the leakage inductance calibration value is used to correct the discrete error of magnetic circuit parameters caused by hardware production and assembly.

[0045] The model parameters are then iteratively updated by minimizing the objective function, that is, the parameter set is continuously adjusted through optimization algorithms. The goal is to continuously reduce the loss value of the objective function, meaning to make the model's predicted power constantly approach the measured power. The optimization process is as follows: First, calculate the partial derivative of the objective function with respect to each parameter to be optimized, i.e., the gradient value. The gradient direction represents the direction of the fastest increase in the loss function, so adjusting the parameters in the opposite direction of the gradient can reduce the loss. Then, adaptively adjust the learning step size of each parameter by combining the first and second moments. Use a larger step size for parameters with gentle fluctuations to speed up convergence, and use a smaller step size for parameters with drastic fluctuations to avoid oscillations. Each round of gradient calculation and parameter update for all samples constitutes one training iteration. During training, the sample set is divided into a training set and a validation set in an 8:2 ratio. The training set is used for parameter updates, and the validation set is used to verify the model's generalization ability in real time, avoiding the model becoming overly reliant on the training samples. This study addresses the issues of overfitting and inability to adapt to unknown operating conditions. During the iteration process, the loss value of the validation set is continuously monitored. When the decrease in the validation set loss is less than 0.1% for 10 consecutive iterations, and the maximum relative error of the model's predicted power is less than 2% across all operating conditions, the model is considered to have reached convergence in the phase-shift power transfer characteristic model state. At this point, the model parameters have been globally optimally calibrated, and the model can accurately predict the transmitted power under all voltage mismatch conditions, without showing significant accuracy improvements with increasing iteration count. Finally, the converged model parameters and mechanistic formulas are compiled and solidified, then downloaded to the local controller of the multi-port DAB converter, completing the model deployment. During online operation, no further parameter iteration is required; the mapping calculation from power to phase shift angle can be quickly completed simply through analytical inverse kinematics. Furthermore, in the method provided in the application embodiment, multi-port coupling consistency optimization is performed on the initial phase shift angle difference identification. The method includes: obtaining a set of initial phase shift angle differences corresponding to multiple battery cluster pairs; constructing a multi-port phase shift coupling matrix based on the initial phase shift angle difference set to characterize the mutual influence relationship between phase shift control and power transmission between battery cluster pairs; calculating the coupling interference coefficient between battery cluster pairs based on the multi-port phase shift coupling matrix; performing consistency correction on the initial phase shift angle difference according to the coupling interference coefficient; and generating a phase shift angle equalization control difference for controlling the actual driving of each full-bridge switch.

[0046] Specifically, firstly, the initial phase shift angle difference set corresponding to multiple battery cluster pairs is obtained. This set is calculated independently for each battery cluster pair by the phase shift power transfer characteristic model. The calculation process only considers the voltage state and power demand of the target cluster pair and does not include the indirect influence of the phase shift actions of other ports through the common magnetic circuit. Therefore, directly issuing control will result in the actual power deviating from the target and the power of multiple clusters interfering with each other and oscillating. Global correction is required through coupling optimization. Next, a multi-port phase shift coupling matrix is ​​constructed. This matrix is ​​a mathematical carrier that quantifies the degree of influence of the phase shift control action of each battery cluster pair on its own and the power transfer of all other battery cluster pairs. It fully describes the magnetic circuit coupling topology relationship between the multiple ports. For a multi-port DAB system with n low-voltage side full-bridge ports, an n-order square matrix coupling matrix is ​​constructed. The matrix elements are divided into two categories: diagonal elements and off-diagonal elements. Let be the power-phase shift main sensitivity of the i-th port, which characterizes the change in transmitted power of the corresponding battery cluster when the phase shift angle of the i-th port changes by one unit angle. Its value can be obtained by differentiating the DAB power transmission formula with respect to the phase shift angle. The specific expression is as follows: ; In the formula The equivalent voltage on the high-voltage bus side, For the equivalent leakage inductance of port i, The voltage transformation ratio imbalance correction coefficient; the main sensitivity is the core quantitative indicator of the port's own control capability; off-diagonal elements. Let be the cross-coupling coefficient between port j and port i. This coefficient characterizes the magnitude of the power transmission deviation at port i caused by the superposition effect of the magnetomotive force of the common core when the phase shift angle of port j changes by a unit angle. Its derivation is based on the transformer magnetomotive force balance principle: the vector sum of ampere-turns of all low-voltage windings equals the core excitation ampere-turns. The phase shift change at port j alters the phase and amplitude of its winding current, thus changing the total excitation magnetomotive force, inducing additional harmonic voltages across the i-th winding, ultimately causing a power offset at port i. The calculation formula can be expressed as: ; in For transformer magnetizing inductance, The larger the absolute value of this value, the stronger the magnetic circuit coupling between the two ports and the more significant the cross interference.

[0047] After constructing the coupling matrix, the coupling interference coefficient between each battery cluster pair is calculated based on the matrix elements. This coefficient is a dimensionless index normalized to the cross-coupling strength, used to intuitively characterize the relative interference degree of a single-port control action on the other ports, avoiding misjudgment of the interference degree due to differences in the main sensitivity of each port. For the coupling interference coefficient between port j and port i... The cross-coupling coefficient is calculated as the ratio of the cross-coupling coefficient to the main sensitivity of the interfered port. The formula is: ; This coefficient ranges from [0,1]. A value closer to 1 indicates stronger cross-interference. When the phase shift at port j exceeds a preset threshold, the phase shift will cause a non-negligible deviation in the power transmission at port i, which is the core grading basis for subsequent consistency correction. Based on this, consistency correction is performed on the initial phase shift angle difference. First, based on the current set of initial phase shift angle differences and the coupling matrix, the total deviation vector of coupled interference power experienced by all ports is calculated: ; in The initial phase angle difference vector is used, and the theoretical correction amount of the phase angle required to offset the power deviation is solved by the pseudo-inverse operation of the coupling matrix. Then, the step size constraint is applied to the correction amount based on the classification result of the coupling interference coefficient. For the strongly coupled port with a high coupling interference coefficient, a small step size constraint is used to avoid iterative oscillation. For the weakly coupled port with a low coupling interference coefficient, a large step size is used to speed up the convergence speed. The constrained correction amount is superimposed on the current phase angle difference to complete one round of iteration. After iteration, the coupling interference coefficient and the global power deviation are recalculated. The above correction process is repeated until the coupling interference coefficient of all ports is lower than the convergence threshold and the global power deviation meets the accuracy requirements. The final output phase angle difference is the phase angle equalization control difference used to control the actual drive of each full-bridge switch.

[0048] Furthermore, in the method provided in the application embodiment, the initial phase angle difference is corrected for consistency based on the coupling interference coefficient. The method includes: constructing a phase difference correction step size for each battery cluster pair based on the coupling interference coefficient; using the phase difference correction step size to perform feedforward iterative correction on the initial phase angle difference until the coupling interference coefficient is less than a first preset threshold to generate the phase angle equalization control difference.

[0049] Specifically, the phase shift difference correction step size is first defined as the adjustment range of the phase shift angle difference for a single battery cluster pair in each iteration. It is a parameter for the convergence speed and operational stability of the correction process. For each battery cluster pair, the correction step size is generated based on the comprehensive coupling interference coefficient it receives: First, based on the current set of phase shift angle differences and the multi-port phase shift coupling matrix, the total value of the coupling interference power superimposed from all other ports received by each battery cluster pair is calculated. Then, the power deviation is converted into the theoretical correction amount of the phase shift angle by the power-phase shift master sensitivity of the corresponding port. Subsequently, the theoretical correction amount is scaled according to the inverse proportional rule of the coupling interference coefficient. The higher the coupling interference coefficient, the stronger the cross influence of the other ports on the cluster pair. The lower the scaling ratio, the smaller the correction step size, so as to avoid system oscillation caused by excessive adjustment in a single strong coupling scenario. During the scaling process, the positive and negative signs of the correction amount are retained to ensure that the correction direction is opposite to the direction of coupling interference, so as to achieve the reverse cancellation effect of coupling interference. Finally, a unique phase shift difference correction step size is generated for each battery cluster pair.

[0050] After the step size is constructed, a feedforward iterative correction is performed, which is an active correction mechanism based on the coupling mechanism prediction. The interference generated by the current phase shift combination is predicted in advance through the coupling matrix, and the reverse compensation is directly applied to achieve deviation cancellation. It has the advantages of fast response speed and small adjustment overshoot. Specifically, the phase shift difference correction step size of each battery cluster pair is superimposed on the current initial phase shift angle difference value to complete the first round of feedforward correction and obtain the updated phase shift angle difference value set.

[0051] Subsequently, based on the updated phase angle difference, the coupling interference coefficients of all battery cluster pairs are recalculated. It is then verified whether the coupling interference coefficients of all cluster pairs are less than the first preset threshold. This threshold is the upper limit of coupling interference preset by the system, representing that the impact of cross-coupling on power transmission has been suppressed to an acceptable accuracy range and will not have a significant impact on the equalization control effect. In engineering, it is usually set to 0.02~0.05, corresponding to a power deviation caused by coupling of less than 2%~5%. If the coupling interference coefficients of some battery cluster pairs are still higher than this threshold, the process of calculating coupling interference, generating correction step size, and superimposing phase angle correction is repeated from the current updated phase angle difference as the starting point for iteration. Each iteration further reduces the coupling interference intensity until the coupling interference coefficients of all battery cluster pairs are all lower than the first preset threshold. At this point, the iteration is determined to have reached the convergence state, and the final output phase angle difference is the phase angle equalization control difference used to control the actual driving of each full-bridge switch.

[0052] Furthermore, in the method provided in the application embodiment, the multi-port DAB converter includes a first step length constraint interval and a second step length constraint interval; a first group of battery cluster pairs with a value greater than a second preset threshold and a second group of battery cluster pairs with a value less than the second preset threshold and a value greater than the first preset threshold are obtained according to the coupling interference coefficient; wherein, under the condition of the first step length constraint interval, a phase shift difference correction step size for each battery cluster pair in the first group of battery cluster pairs is generated, and under the condition of the second step length constraint interval, a phase shift difference correction step size for each battery cluster pair in the second group of battery cluster pairs is generated.

[0053] Specifically, the first and second long constraint intervals are defined as two non-overlapping phase angle correction ranges pre-defined within the multi-port DAB converter controller. The first long constraint interval is a small-amplitude correction interval, with its upper limit strictly lower than the lower limit of the second long constraint interval, used for small-amplitude smooth correction. The second long constraint interval is a large-amplitude correction interval, with an overall correction amplitude higher than the first long constraint interval, used for rapid convergence adjustment to reduce the number of iterations and improve response speed. The second preset threshold is a boundary judgment value that distinguishes between strong coupling and moderate coupling conditions, and its value is significantly greater than the first preset threshold, which serves as the convergence standard. Together, they constitute a two-level judgment boundary for coupling strength. The first group of battery clusters refers to battery cluster pairs with coupling interference coefficients greater than the second preset threshold. These cluster pairs typically correspond to ports with adjacent transformer winding positions and tight magnetic circuit coupling, resulting in strong cross-interference and being the main source of system power oscillations. The second group of battery clusters refers to battery cluster pairs with coupling interference coefficients less than the second preset threshold but greater than the first preset threshold. The magnetic circuit cross-influence of these cluster pairs is moderate, and large adjustments within a reasonable range will not cause system instability.

[0054] Specifically, the coupling strength of battery clusters is first graded and screened: the coupling interference coefficient of each battery cluster is compared with two preset thresholds. Clusters with a coupling interference coefficient greater than the second preset threshold are assigned to the first group of strong coupling sets, clusters with a coupling interference coefficient between the first and second preset thresholds are assigned to the second group of medium coupling sets, and clusters with a coupling interference coefficient less than the first preset threshold have met the convergence accuracy requirements and do not need to be further corrected.

[0055] After grouping, phase difference correction step sizes are generated for each of the two cluster pairs according to the corresponding step size constraint interval: For the first group of strongly coupled battery cluster pairs, a dedicated correction step size for each cluster pair is generated under the constraint of the first step size constraint interval. First, the theoretical phase shift correction amount required to cancel coupling interference is calculated through the multi-port phase shift coupling matrix. Then, the absolute value of the theoretical correction amount is compared with the upper and lower limits of the first step size interval. If it exceeds the upper limit, it is truncated to the upper limit value. If it is lower than the lower limit, the lower limit value is taken. By strictly limiting the single correction amplitude, the risk of iterative divergence, power oscillation, or even transformer magnetic saturation due to excessive adjustment in the strongly coupled system is avoided from the mechanism. For the second group of moderately coupled battery cluster pairs, a correction step size for each cluster pair is generated under the constraint of the second step size constraint interval. Similarly, the theoretical correction amount is calculated based on the coupling matrix and then limited within the second step size interval. Since the coupling interference of this type of cluster pair is relatively low, the iterative convergence speed can be accelerated by a larger correction amplitude, and the response delay of the equalization control can be reduced.

[0056] Furthermore, in the method provided in the application embodiment, the upper limit of the first step long constraint interval is less than the lower limit of the second step long constraint interval.

[0057] Specifically, the first step of the long constraint interval is used to limit the single phase shift adjustment amplitude under high coupling, so as to avoid system iterative divergence and power oscillation instability. The second step of the long constraint interval is used to amplify the single adjustment amplitude under weak coupling interference, reduce the number of iterations and improve the convergence speed. The upper limit of the first step of the long constraint interval is less than the lower limit of the second step of the long constraint interval. This is the core configuration rule of the two-level intervals. It means that the two-level intervals are completely separated in the numerical dimension of the phase shift angle and have no overlap. The maximum correction amplitude allowed in the first step of the long constraint interval is strictly less than the minimum correction amplitude allowed in the second step of the long constraint interval. This forms a two-level independent distribution with lower overall values ​​in the small step interval, corresponding to strong coupling conditions, and higher overall values ​​in the large step interval, corresponding to medium coupling conditions.

[0058] During the system parameter pre-configuration phase, engineers pre-calibrated four boundary values ​​for the two-stage step size interval based on the winding topology, port size, and magnetizing inductance and leakage inductance parameters of the multi-port DAB converter. The first step size constraint interval is set as follows: The second step size constraint interval is Forced to meet the requirements by fixing controller parameters The constraints, for example, in a typical engineering configuration, the first step length interval can be set to 0.05°~0.8°, the second step length interval can be set to 1.0°~5.0°, and a numerical buffer gap of 0.2° is retained between the two to eliminate the overlapping area of ​​the intervals.

[0059] During the group matching phase, when the coupling interference coefficient of a battery cluster pair exceeds the second preset threshold and it is assigned to the first strongly coupled cluster pair, the generation of its correction step size is limited by the first step size constraint interval. No matter how large the theoretical correction amount calculated through the coupling matrix is, the final output correction step size will not exceed [the threshold value]. The upper limit of the step size eliminates the risk of over-adjustment in strongly coupled scenarios from the boundary. When the coupling interference coefficient of the battery cluster is between the first preset threshold and the second preset threshold and is classified into the second group of moderately coupled clusters, its correction step size is limited by the second step size constraint interval. The minimum allowable correction amplitude is already higher than the upper limit of the first step size, and the iteration convergence speed can be accelerated by a larger adjustment amplitude. In the step size generation limiting stage, the independent boundaries of the two-level intervals can ensure a one-to-one correspondence between the coupling strength level and the step size level. There will be no control jitter problem where the step size jumps back and forth between the two intervals when the coupling coefficient fluctuates slightly near the level threshold due to sampling noise.

[0060] Furthermore, in the method provided in the application embodiment, the multi-port DAB converter further includes a high-frequency isolation transformer, through which the multi-port DAB converter exchanges energy with the DC bus terminal.

[0061] Specifically, it is first clarified that a high-frequency isolation transformer is a magnetically coupled energy transmission device operating in the switching frequency band of tens to hundreds of kHz. It is fabricated using high-permeability high-frequency magnetic cores such as ferrite and nanocrystalline materials, and possesses complete electrical insulation capabilities. It is both the core magnetic circuit carrier for multi-path energy convergence and voltage level matching in multi-port DAB converters and a core isolation component ensuring system electrical safety. The windings adopt a common core arrangement structure with a single high-voltage side winding and multiple low-voltage side windings. All windings are insulated from each other and wound on the same magnetic core circuit. The number of low-voltage side windings and... The number of heterogeneous battery clusters corresponds one-to-one. The two leads of each low-voltage side winding are connected to the AC output port of a low-voltage side full-bridge circuit, while the two leads of the high-voltage side winding are connected to the AC port of the high-voltage side full-bridge circuit on the DC bus side. The DC side of the high-voltage side full-bridge is directly connected to the centralized DC bus, i.e., the front-end DC bus of the energy storage converter PCS. This constructs a conduction link between the heterogeneous battery clusters, the low-voltage side full-bridge, the high-frequency isolation transformer, the high-voltage side full-bridge, and the DC bus. At the same time, all low-voltage side windings also form an inter-cluster energy coupling path that can directly interact through a common magnetic core.

[0062] In actual operation, bidirectional energy exchange is achieved through the coordinated switching of the full-bridge working mode. When the heterogeneous battery clusters feed energy to the DC bus, each low-voltage side full-bridge operates in inverter mode, converting the DC output of the corresponding battery cluster into a high-frequency AC square wave with the same switching frequency and applying it to the corresponding low-voltage winding. Through the electromagnetic induction effect of the common magnetic core, a high-frequency AC voltage of the same frequency is induced on the high-voltage side winding, and then rectified by the high-voltage side full-bridge into DC power that meets the bus voltage level and flows into the DC bus, completing the positive transmission of energy from the battery side to the bus side.

[0063] When the DC bus replenishes energy to the heterogeneous battery clusters, the high-voltage side full-bridge switches to inverter mode, inverting the DC power from the DC bus into a high-frequency AC square wave applied to the high-voltage winding. Through magnetic coupling, high-frequency AC power of the corresponding voltage level is induced in each low-voltage winding. This AC power is then rectified by the low-voltage side full-bridge into DC power matching the voltage of the corresponding battery cluster, thus charging the battery cluster and achieving reverse energy transfer from the bus side to the battery side. At the same time, the balanced energy transfer between heterogeneous battery clusters is also accomplished through the common magnetic circuit of the high-frequency isolation transformer. That is, the battery clusters with energy flowing out are injected into the common magnetic circuit through the low-voltage full-bridge inverter, and the battery clusters with energy flowing in are rectified by the low-voltage full-bridge to extract energy from the magnetic circuit. Direct and efficient balance between clusters can be achieved without passing through the DC bus.

[0064] In summary, the heterogeneous battery cluster adaptive equalization control method provided in this application has the following technical effects: By constructing an isolated topology connection between heterogeneous battery clusters and a multi-port DAB converter, parallel circulating currents are physically blocked. A multi-dimensional energy difference function is established, which integrates the differences in state of charge, internal resistance loss weights, temperature constraints, and dynamic trends, to generate the target power transmission between clusters. Based on the phase-shift power transmission characteristic model that takes into account the voltage ratio imbalance, the power demand is accurately mapped to the phase-shift control quantity. The consistency correction of global phase-shift control is achieved through multi-port phase-shift coupling matrix and step-size feedforward iterative optimization. Ultimately, the control accuracy, operational stability, and overall service life of the heterogeneous battery cluster equalization system are improved.

[0065] Example 2, based on the same inventive concept as the heterogeneous battery cluster adaptive equalization control method in the foregoing examples, such as... Figure 2 As shown in the embodiment of this application, an adaptive equalization control system for heterogeneous battery clusters is provided. The system includes: Control module 11 is used to establish a topology connection between the heterogeneous battery clusters and the multi-port DAB converter. The multi-port DAB converter includes multiple low-voltage side full-bridge circuits, connecting each battery cluster to one low-voltage side full-bridge circuit. Each low-voltage side full-bridge circuit is phase-shifted through full-bridge switches. Acquisition module 12 is used to acquire multiple battery cluster operating state parameters of the heterogeneous battery clusters to construct a comprehensive battery cluster state vector. Based on the comprehensive battery cluster state vector, a multi-dimensional energy difference function between any two battery clusters is constructed to generate the target power transfer requirement between any two battery clusters. Mapping module 13 is used by the multi-port DAB converter to map the target power transfer requirement to the corresponding initial phase shift angle difference by calling the phase shift power transfer characteristic model. Optimization module 14 is used to perform multi-port coupling consistency optimization on the initial phase shift angle difference identification to generate a phase shift angle equalization control difference for controlling the actual driving of each full-bridge switch.

[0066] Furthermore, the acquisition module 12 is also used to perform the following steps: perform state vector difference calculation on any two battery clusters according to the battery cluster comprehensive state vector to obtain a basic state difference vector, wherein the battery cluster comprehensive state vector includes state of charge (SOC), equivalent internal resistance parameter, temperature parameter, and state of charge change rate; normalize the basic state difference vector to obtain a standardized state difference vector; obtain a multidimensional energy difference term based on the standardized state difference vector, and fit a multidimensional energy difference function with the multidimensional energy difference term, wherein the multidimensional energy difference function is used to output the target power transmission requirement of the target battery cluster, wherein the multidimensional energy difference term includes at least a SOC difference term, an internal resistance loss weight term, a temperature constraint penalty term, and a dynamic change trend term.

[0067] Furthermore, the mapping module 13 is also used to perform the following steps: inputting the target power transfer demand between any two battery clusters into the multi-port DAB converter; the multi-port DAB converter maps the target power transfer demand to the corresponding phase shift angle difference by calling the phase shift power transfer characteristic model; the phase shift power transfer characteristic model includes: establishing the AC voltage port ratio equation between each low-voltage side full-bridge circuit, identifying the voltage ratio imbalance caused by the voltage mismatch at the ends of the heterogeneous battery clusters; the phase shift power transfer characteristic model takes the target power transfer demand of the target battery cluster pair as the response target, the voltage ratio imbalance as the input variable, controls the phase shift angle difference of the corresponding full-bridge switch of the target battery cluster as the solution variable for solving, and outputs the phase shift angle difference for the target battery cluster.

[0068] Furthermore, the mapping module 13 is also used to perform the following steps: acquiring training sample data based on the operation process of the multi-port DAB converter, wherein the training sample data includes at least the state of charge (SOC), equivalent internal resistance parameters, port voltage parameters, temperature parameters, and corresponding actual phase shift angle control data of each battery cluster; performing time series alignment processing on the training sample data to construct training sample pairs reflecting the correspondence between voltage ratio imbalance and DAB power transfer behavior; establishing an objective function based on the training sample pairs, wherein the objective function is used to characterize the deviation between the target power transfer demand and the actual output power; updating the parameters of the phase shift power transfer characteristic model by minimizing the objective function until a converged phase shift power transfer characteristic model is obtained, and downloading the converged phase shift power transfer characteristic model to the multi-port DAB converter.

[0069] Furthermore, the optimization module 14 is also used to perform the following steps: obtaining a set of initial phase shift angle differences corresponding to multiple battery cluster pairs; constructing a multi-port phase shift coupling matrix based on the set of initial phase shift angle differences to characterize the mutual influence relationship between phase shift control and power transmission between battery cluster pairs; calculating the coupling interference coefficient between battery cluster pairs based on the multi-port phase shift coupling matrix; performing consistency correction on the initial phase shift angle differences according to the coupling interference coefficient; and generating a phase shift angle equalization control difference for controlling the actual driving of each full-bridge switch.

[0070] Furthermore, the optimization module 14 is also used to perform the following steps: constructing a phase shift difference correction step size for each battery cluster pair based on the coupling interference coefficient; using the phase shift difference correction step size to perform feedforward iterative correction on the initial phase shift angle difference until the coupling interference coefficient is less than a first preset threshold to generate the phase shift angle equalization control difference.

[0071] Furthermore, the mapping module 13 is also used to perform the following steps: the multi-port DAB converter includes a first step length constraint interval and a second step length constraint interval; according to the coupling interference coefficient, a first group of battery cluster pairs with a value greater than a second preset threshold and a second group of battery cluster pairs with a value less than the second preset threshold and a value greater than the first preset threshold are obtained; wherein, under the condition of the first step length constraint interval, a phase shift difference correction step size for each battery cluster pair in the first group of battery cluster pairs is generated, and under the condition of the second step length constraint interval, a phase shift difference correction step size for each battery cluster pair in the second group of battery cluster pairs is generated.

[0072] Furthermore, the mapping module 13 is also used to perform the following steps: the upper limit of the first step long constraint interval is less than the lower limit of the second step long constraint interval.

[0073] Furthermore, the mapping module 13 is also used to perform the following steps: the multi-port DAB converter further includes a high-frequency isolation transformer, and the multi-port DAB converter exchanges energy with the DC bus terminal through the high-frequency isolation transformer.

[0074] 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 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. An adaptive equalization control method for heterogeneous battery clusters, characterized in that, The method includes: A topological connection is established between the heterogeneous battery cluster and the multi-port DAB converter. The multi-port DAB converter includes multiple low-voltage-side full-bridge circuits. Each battery cluster is connected to a low-voltage-side full-bridge circuit. Each low-voltage-side full-bridge circuit is phase-shifted through a full-bridge switch. Multiple battery cluster operating state parameters of the heterogeneous battery cluster are obtained to construct a comprehensive battery cluster state vector. Based on the comprehensive battery cluster state vector, a multidimensional energy difference function between any two battery clusters is constructed to generate the target power transmission requirement between any two battery clusters. The multi-port DAB converter maps the target power transmission requirement to the corresponding initial phase shift angle difference by calling the phase-shift power transmission characteristic model. Multi-port coupling consistency optimization is performed on the initial phase shift angle difference to generate a phase shift angle equalization control difference for controlling the actual drive of each full-bridge switch.

2. The method as described in claim 1, characterized in that, The method for constructing a multidimensional energy difference function between any two battery clusters based on the comprehensive state vector of the battery clusters includes: Based on the comprehensive state vector of the battery cluster, the state vector difference calculation is performed on any two battery clusters to obtain the basic state difference vector. The comprehensive state vector of the battery cluster includes the state of charge (SOC), equivalent internal resistance parameter, temperature parameter, and state of charge change rate. The basic state difference vector is normalized to obtain a standardized state difference vector; A multidimensional energy difference term is obtained based on the standardized state difference vector, and a multidimensional energy difference function is fitted using the multidimensional energy difference term. The multidimensional energy difference function is used to output the target power transmission requirement of the target battery cluster. The multidimensional energy difference term includes at least a SOC difference term, an internal resistance loss weight term, a temperature constraint penalty term, and a dynamic change trend term.

3. The method as described in claim 1, characterized in that, The target power transfer requirement between any two battery clusters is input to the multi-port DAB converter. The multi-port DAB converter maps the target power transfer requirement to a corresponding phase shift angle difference by calling a phase-shifted power transfer characteristic model. The phase-shifted power transfer characteristic model includes: Establish the AC voltage port transformation ratio equation between each low-voltage side full-bridge circuit, and identify the voltage transformation ratio imbalance caused by the voltage mismatch at the terminals of the target battery cluster. The phase-shifted power transmission characteristic model takes the target power transmission demand of the target battery cluster as the response target, the voltage ratio imbalance as the input variable, and the phase shift angle difference of the full-bridge switch corresponding to the target battery cluster as the solution variable to solve the problem, and outputs the phase shift angle difference value for the target battery cluster.

4. The method as described in claim 3, characterized in that, The method for training the phase-shifted power transfer characteristic model includes: Training sample data is obtained based on the operation process of the multi-port DAB converter. The training sample data includes at least the state of charge (SOC), equivalent internal resistance parameters, port voltage parameters, temperature parameters, and corresponding actual phase shift angle control data of each battery cluster. The training sample data is time-series aligned to construct training sample pairs that reflect the correspondence between voltage ratio imbalance and DAB power transfer behavior. A target function is established based on the training sample pairs. The target function is used to characterize the deviation between the target power transmission demand and the actual output power. The phase-shift power transfer characteristic model is updated by minimizing the objective function until a converged phase-shift power transfer characteristic model is obtained. The converged phase-shift power transfer characteristic model is then downloaded to the multi-port DAB converter.

5. The method as described in claim 1, characterized in that, The method for optimizing multi-port coupling consistency of the initial phase shift angle difference includes: Obtain the set of initial phase shift angle differences corresponding to multiple battery cluster pairs; Based on the initial set of phase shift angle differences, a multi-port phase shift coupling matrix is ​​constructed to characterize the mutual influence of phase shift control on power transmission between each battery cluster pair; The coupling interference coefficient between each battery cluster pair is calculated based on the multi-port phase-shifting coupling matrix. The initial phase-shifting angle difference is then corrected for consistency based on the coupling interference coefficient, generating a phase-shifting angle equalization control difference for controlling the actual driving of each full-bridge switch.

6. The method as described in claim 5, characterized in that, The method for performing consistency correction on the initial phase shift angle difference based on the coupling interference coefficient includes: Construct a phase shift correction step size for each battery cluster pair based on the coupling interference coefficient; The initial phase angle difference is iteratively corrected using the phase difference correction step size until the coupling interference coefficient is less than a first preset threshold, thereby generating the phase angle equalization control difference.

7. The method as described in claim 6, characterized in that, The multi-port DAB converter includes a first step long constraint interval and a second step long constraint interval. Based on the coupling interference coefficient, a first group of battery clusters with a value greater than the second preset threshold and a second group of battery clusters with a value less than the second preset threshold and a value greater than the first preset threshold are obtained; Specifically, under the condition of the first step length constraint interval, the phase shift difference correction step size of each battery cluster pair in the first group of battery cluster pairs is generated, and under the condition of the second step length constraint interval, the phase shift difference correction step size of each battery cluster pair in the second group of battery cluster pairs is generated.

8. The method as described in claim 7, characterized in that, The upper limit of the first step long constraint interval is less than the lower limit of the second step long constraint interval.

9. The method as described in claim 1, characterized in that, The multi-port DAB converter also includes a high-frequency isolation transformer, through which the multi-port DAB converter exchanges energy with the DC bus terminal.

10. A heterogeneous battery cluster adaptive equalization control system, characterized in that, The system is used for implementing the adaptive equalization control method for heterogeneous battery clusters according to any one of claims 1 to 9, wherein the system comprises: The control module is used to establish a topology connection between the heterogeneous battery clusters and the multi-port DAB converter. The multi-port DAB converter includes multiple low-voltage side full-bridge circuits, and each battery cluster is connected to a low-voltage side full-bridge circuit. Each low-voltage side full-bridge circuit is phase-shifted through a full-bridge switch. The acquisition module is used to acquire multiple battery cluster operating state parameters of the heterogeneous battery cluster to construct a comprehensive battery cluster state vector, and to construct a multidimensional energy difference function between any two battery clusters based on the comprehensive battery cluster state vector, which is used to generate the target power transmission demand between any two battery clusters. The mapping module is used by the multi-port DAB converter to map the target power transmission requirement to the corresponding initial phase shift angle difference by calling the phase shift power transmission characteristic model. The optimization module is used to perform multi-port coupling consistency optimization on the initial phase shift angle difference to generate a phase shift angle equalization control difference for controlling the actual drive of each full-bridge switch.

Images