Battery equalization control system

Abstract

The invention provides a battery equalization control system, and relates to the technical field of lithium ion battery pack equalization, and the method comprises the steps: firstly, employing terminal voltage, temperature and internal resistance data, carrying out the cooperative online estimation of the SOC and SOH of a single body through an extended Kalman filtering method and an internal resistance increment method, and introducing a temperature compensation function to carry out the real-time correction of an estimation error; and constructing a comprehensive consistency index to dynamically trigger the equalization demand. Carrying out dimensionality reduction on the high-dimensional state of the battery pack by adopting intrinsic orthogonal decomposition, extracting a dominant mode, sending the dominant mode into a lightweight LSTM network, and outputting an optimal current sequence and duration taking balance time and energy consumption into consideration through small sample training; during online operation, model input is updated in a rolling manner every preset time, a path is recalculated, and rapid convergence and fine adjustment are completed through double-stage self-adaptive current control in an offline scene. The invention aims to solve the problems of low equalization speed, low efficiency and strong dependence on big data samples in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lithium-ion battery pack equalization, and in particular to a battery equalization control system. BACKGROUND

[0002] With the wide application of lithium-ion batteries in portable electronic devices, electric vehicles and energy storage systems, their advantages of high energy density, low self-discharge rate and long cycle life have been fully recognized. However, in long-term practical use, due to the differences in key parameters such as self-discharge rate, internal resistance, temperature distribution and maximum available capacity, the inconsistency of capacity in each single battery within the battery pack is inevitable. This inconsistency not only reduces the overall available capacity and energy utilization efficiency of the battery pack, but also causes some batteries to overcharge or overdischarge during charging and discharging, which in turn leads to thermal runaway and other safety hazards, seriously affecting the safety and reliability of the battery system.

[0003] To address the above problems, the industry generally uses battery equalization technology to maintain the consistency of the battery pack. The current mainstream equalization methods mainly include energy consumption type equalization and energy transfer type equalization. Energy consumption type equalization consumes the power of high-capacity batteries to the same level as the lowest-capacity battery through resistance discharge. This method is simple to control, but energy is wasted and the equalization speed is slow. Energy transfer type equalization supplements low-capacity batteries by charging to the same capacity level as the highest-capacity battery. Although the energy utilization rate is relatively high, it still has the problems of long equalization time and low efficiency. In addition, with the continuous increase in the capacity of lithium-ion battery monomers, the limitations of traditional equalization methods become increasingly prominent, making it difficult to meet the urgent demand for fast and efficient equalization in modern high-capacity and high-power application scenarios.

[0004] In recent years, with the rapid development of machine learning technology, some research has attempted to introduce it into the field of battery equalization in order to optimize the equalization strategy through data-driven methods. For example, a neural network model is used to predict the optimal state of charge (SOC) equalization path, thereby improving the equalization efficiency. However, existing machine learning methods usually rely on a large number of historical data samples for model training, which has a high data acquisition cost, and it is difficult to cover all possible working condition scenarios in practical applications, resulting in limited generalization ability of the model and difficulty in large-scale application.

[0005] In summary, the existing battery equalization technology has different degrees of shortcomings in terms of equalization speed, energy efficiency, control accuracy and practical application adaptability, and there is an urgent need for a new method and supporting device that can quickly and accurately estimate the optimal equalization path under limited data samples, and achieve efficient and intelligent equalization control, in order to effectively solve the safety and performance problems caused by the inconsistency of the battery pack capacity, and thus ensure the long-term stable operation of the lithium-ion battery system.

[0006] In view of this, the present application is proposed. SUMMARY

[0007] The present application provides a battery equalization control system, which can at least partially improve the above problems.

[0008] To achieve the above object, the present application adopts the following technical solutions: A battery equalization control system, comprising: an auxiliary power supply module, an acquisition component, a battery equalization module, a central processing unit, and an interface component, wherein the data end of the central processing unit is electrically connected with the data end of the acquisition component and the data end of the battery equalization module, the battery equalization module is electrically connected with the battery to be repaired and the auxiliary power supply module through the interface component, and the acquisition component is electrically connected with the battery to be repaired and a sensor through the interface component. The central processing unit is configured to realize the following steps by executing the computer program stored therein: The terminal voltage of each battery collected by the acquisition component is obtained, the terminal voltage of each battery is estimated, the battery state of charge SOC estimation value of each battery is obtained, the battery state of charge SOC estimation value is dynamically calibrated according to the preset correction compensation data, and the dynamically calibrated battery SOC state value is obtained. The current SOH state of all batteries is estimated, the dynamically calibrated battery SOC state value is combined, the available capacity difference of all batteries is calculated, the comprehensive consistency score is calculated based on the consistency evaluation index between the batteries, and the available capacity difference and the consistency score are analyzed to determine whether to perform battery equalization processing. When it is determined that battery equalization processing is needed, an intrinsic orthogonal decomposition-long short-term memory network POD-LSTM model is called for prediction processing to generate an optimal equalization path.

[0009] In summary, the battery equalization control system aims to solve the problems of slow equalization speed, low efficiency and strong dependence on large data samples in the prior art. The state of charge and the state of health of each single battery are continuously updated by using the extended Kalman filter and the internal resistance increment method, and the estimation error is corrected by a temperature compensation function; then the voltage, SOC, SOH, temperature difference and internal resistance change are sent to the compressed LSTM network through the intrinsic orthogonal decomposition, and the optimal equalization current, duration and energy consumption can be output with only a few samples; in the online mode, the main circuit and the auxiliary power supply are refreshed every twenty minutes, and the path is recalculated, and in the offline mode, the dual strategy of large current fast convergence and small current fine adjustment is automatically switched. The device layer adopts an independent movable structure, integrates sensing, conversion, energy consumption, wireless communication and auxiliary power supply, can be plugged and used for any battery pack, and supports remote monitoring and parameter self-learning. The overall scheme reduces data dependence and energy waste, greatly shortens the equalization time and improves consistency, and is suitable for various high-capacity lithium battery scenes such as energy storage and electric transportation. The POD-LSTM method can quickly estimate the optimal equalization scheme with a small sample, thereby quickly improving the consistency of the battery pack, and the application scenarios are more extensive. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 is a flowchart of the battery equalization control system provided by the embodiment of the present application. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical scheme and advantages of the present application clearer and more apparent, the present application will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0012] Reference Figure 1 As shown in the figure, the embodiment of the present application discloses a battery equalization control system, which comprises an auxiliary power supply module, a collection component, a battery equalization module, a central processing unit and an interface component. The data end of the central processing unit is electrically connected with the data end of the collection component and the data end of the battery equalization module. The battery equalization module is electrically connected with the battery to be repaired and the auxiliary power supply module through the interface component. The collection component is electrically connected with the battery to be repaired and the sensor through the interface component. Preferably, the interface component comprises a voltage charge-discharge line interface, a detection line independent interface and a temperature measurement line interface. The central processing unit is built-in with a battery state of health SOH estimation module, which calculates the SOH value in real time by the internal resistance increment method and is used for SOC calibration.

[0013] Preferably, the acquisition component includes a voltage acquisition module and a temperature acquisition module. The voltage acquisition module is connected to the positive and negative terminals of the battery to be repaired through an independent interface of the detection line. The temperature acquisition module is connected to the battery temperature sensor through a temperature measurement line interface. The data terminal of the central processing unit is electrically connected to the data terminals of the voltage acquisition module and the temperature acquisition module.

[0014] Preferably, the battery balancing module includes a converter and a power dissipation resistor. One end of the converter is connected to the positive and negative terminals of the battery to be repaired through a voltage charging and discharging line interface, and the other end of the converter is connected to an auxiliary power supply module. The power dissipation resistor is connected in series with a control switch and is also connected to the positive and negative terminals of the battery to be repaired through a voltage charging and discharging line interface. The converter and the power dissipation resistor cannot be turned on simultaneously.

[0015] Preferably, it also includes a wireless communication module, through which the central processing unit communicates and interacts with an external host computer or cloud platform, wherein the wireless communication module is a Bluetooth or Wi-Fi module.

[0016] Specifically, in this embodiment, the battery balancing control system is presented as a portable, plug-and-play independent device. Its core idea is to allow any set of lithium-ion batteries already installed in a vehicle or energy storage cabinet to be connected to the system within seconds via three wiring harnesses—positive and negative power lines, positive and negative detection lines, and temperature sensing lines. Subsequently, the device autonomously completes the closed-loop process of diagnosis, modeling, decision-making, and energy correction without requiring any software or hardware modifications to the original battery management system.

[0017] The system casing houses, from bottom to top, an auxiliary power module, a data acquisition module, a battery balancing module, a central processing unit (CPU), and a wireless communication module. The auxiliary power module provides isolated power to the internal circuitry and acts as an energy "transfer station," replenishing energy to low-capacity cells when needed or absorbing energy released from high-capacity cells when necessary. The data acquisition module reads the terminal voltage, internal resistance, and temperature information (with 0.1°C accuracy) of each cell in real time through independent interfaces for detection lines and temperature measurement lines. This data is first fed into the CPU's built-in State of Health (SOH) estimation module. This module provides the SOH value within milliseconds based on the internal resistance increment method and immediately uses it to perform temperature-aging joint compensation on the SOC output of the extended Kalman filter, thus obtaining the most accurate current state of charge. Subsequently, the CPU calls a pre-defined POD-LSTM lightweight model, which can infer the optimal future balancing current sequence and duration locally using only historical curve segments from the past fifteen minutes. This process requires no cloud intervention, protecting data privacy and reducing dependence on network conditions.

[0018] Upon receiving the command, the battery balancing module immediately executes the following: If the target cell's SOC is higher than the average, the energy-dissipating resistor loop is closed first, rapidly discharging energy with a large current. Once the consistency index drops to a preset multiple of the initial value or the temperature rise rate exceeds the threshold, the switch immediately switches to a low-current fine-tuning mode. Energy is no longer wasted but is fed back to the auxiliary power module through the converter for use by nearby low-capacity cells. Conversely, if the target cell's SOC is lower than the average, the converter operates in reverse, injecting energy from the auxiliary power module into the cell with an optimal current, also undergoing a two-stage process of high current followed by low current. Throughout the process, the converter and the energy-dissipating resistor are interlocked to ensure that only one energy path is activated at any given time, eliminating the risk of circulating current and thermal runaway.

[0019] The system also integrates a Bluetooth / Wi-Fi wireless communication module, which periodically uploads key parameters, balancing progress, and predicted remaining time to a host computer or cloud platform. Maintenance personnel can view this information in real time on their mobile phones or remote monitoring screens. The weights and strategy coefficients of the POD-LSTM can be updated via OTA (Over-The-Air) to adapt to batteries with different chemical systems or aging stages. Thanks to this highly integrated and intelligent design, field personnel only need to plug the device into the battery pack. For example, users can complete a rapid balancing process within an average of 30 minutes, reducing capacity differences from 10% to less than 0.5%. Energy utilization is improved by more than 40%, and the number of balancing operations is reduced by half compared to traditional solutions, significantly extending the overall battery life while reducing maintenance costs and downtime.

[0020] The central processing unit is configured to perform the following steps by executing a computer program stored internally: S1, acquire the terminal voltage of each battery collected by the acquisition component, estimate the terminal voltage of each battery to obtain the estimated state of charge (SOC) value of each battery, and perform dynamic calibration processing on the estimated SOC value of the battery according to the preset correction and compensation data to obtain the dynamically calibrated SOC value of the battery. Specifically, step S1 further includes: acquiring the terminal voltage of each battery acquired by the voltage acquisition module, estimating the terminal voltage of each battery using the extended Kalman filter algorithm, and obtaining the estimated state of charge (SOC) value of each battery. The battery temperature T collected by the temperature acquisition module is obtained. Using battery temperature T as a compensation factor and the battery state of health (SOH) value as a correction factor, the estimated battery state of charge (SOC) is dynamically calibrated. The formula is as follows: ,in, This is the dynamically calibrated battery SOC state value. This is an estimated value for the battery's state of charge (SOC). An empirical coefficient related to battery type. This refers to the State of Health (SOH) value of the battery. This is a compensation function related to battery temperature, used to correct the effect of temperature on SOC estimation.

[0021] In this embodiment, after power-on and electrical connection with the battery pack to be balanced, the central processing unit immediately reads the terminal voltage of each cell through the independent interface of the detection line, and simultaneously acquires the battery temperature T (accuracy 0.1℃) output by the NTC thermistor array closely attached to the cell surface through the temperature measurement line interface. The above two sets of raw data first undergo analog front-end filtering and 16-bit ADC conversion inside the processor, and then enter the extended Kalman filter: this filter uses a second-order RC equivalent circuit as the state model and the terminal voltage as the observation value, and recursively updates the estimated state of charge (SOC) value of each cell in each millisecond cycle. In order to eliminate the systematic errors caused by temperature drift and aging degradation, the processor then calls the fixed SOH estimation module—this module uses the battery internal resistance increment method to calculate the battery health state SOH value in real time based on the internal resistance increment curve of the past few weeks, and the current battery temperature T, to jointly perform dynamic calibration on the SOC estimation value.

[0022] S2. Estimate the current SOH state of all batteries, combine the dynamically calibrated SOC state value of the batteries, calculate the available capacity difference of all batteries, calculate the comprehensive consistency score based on the consistency evaluation index between batteries, and analyze the available capacity difference and consistency score to determine whether battery equalization processing is required. Specifically, step S2 further includes: estimating the current SOH state of all batteries online using the battery internal resistance incremental method, and calculating the usable capacity difference of all batteries by combining the dynamically calibrated battery SOC state value, the formula of which is: , Due to the difference in usable battery capacity, The difference between the rated usable capacity of the battery, This represents the current SOH state of the battery. This is a capacity difference correction function related to battery temperature, which is used to correct the effect of temperature on the difference in usable battery capacity. A battery consistency evaluation index is introduced, and a weighted calculation is performed based on this index to obtain a comprehensive consistency score. The formula is as follows: ,in, For consistency scoring, , , , All are empirical coefficients. This represents the voltage difference between the battery terminals. This represents the SOC difference of the battery. This represents the SOH difference of the battery. This represents the temperature difference of the battery.

[0023] Analyze the difference in usable capacity of the battery and consistency score To determine the difference in usable battery capacity Does it exceed the predetermined value? And the SOH decay rate is greater than the threshold or the consistency score Does it exceed the threshold? If so, perform battery balancing. If not, battery balancing will not be performed.

[0024] In this embodiment, the past internal resistance-capacity database is first retrieved, and the battery internal resistance increment method is performed on each cell. The average resistance is measured under millisecond-level pulse current excitation, and then converted to the current SOH based on the empirical mapping curve. Since SOH, SOC, and SOH have all been calibrated through temperature compensation, the resulting difference in usable capacity directly reflects the true difference in "how much electricity can still be released" between cells. Next, a battery consistency evaluation index is introduced. This index comprehensively considers the battery terminal voltage difference, SOC difference, SOH difference, and temperature difference to calculate a comprehensive consistency score, used to comprehensively evaluate the consistency status between different batteries. Once the difference in usable capacity of the batteries is determined... Exceeding the predetermined value And the SOH decay rate is greater than the threshold or the consistency score If the threshold is exceeded, a balancing process is performed. Otherwise, no balancing process is required.

[0025] S3, when it is determined that battery balancing is required, the intrinsic orthogonal decomposition-long short-term memory network POD-LSTM model is called for prediction to generate the optimal balancing path.

[0026] Specifically, step S3 further includes: using the POD algorithm to perform POD decomposition processing on the historical dataset, calculating the energy contribution rate of each mode, removing secondary modes with contribution rates less than a preset value, extracting high-dimensional feature modes of the battery pack state, and selecting multiple dominant modes as inputs to the LSTM network. A time-sliding window mechanism is employed, inputting battery state data from multiple consecutive time points into an LSTM to predict the equalization parameters at future time points. Based on these predicted parameters, an optimal equalization path is generated. The input time-series data to the LSTM network includes: terminal voltage, dynamically calibrated battery SOC value, current battery SOH state, battery temperature T, and internal resistance. The output is the equalization current I. eq Duration Δt and energy consumption; The intrinsic orthogonal decomposition-long short-term memory network (POD-LSTM) model employs a multi-objective optimization strategy during training, with its objective function being: minimizing the equilibrium time t. eq and total energy consumption E eq The formula is: Its constraints are: , To balance time weights, As the weight of total energy consumption, Minimum allowable temperature, This is the maximum permissible temperature.

[0027] In this embodiment, the historical dataset is decomposed into POD (Power, Distance, and Occurrence) parameters to calculate the energy contribution rate of each orthogonal mode. Secondary components with contribution rates below 5% are automatically removed, retaining only the dominant modes with cumulative energy ≥ 95%. The selected battery state data for N consecutive time periods is input into a pre-trained LSTM network via a time sliding window to predict the equalization parameters for the next M time periods. The LSTM network input time-series data includes: terminal voltage, SOC (State of Charge), SOH (State of Health), temperature, and internal resistance; the output is equalization current, duration, and energy consumption. To balance speed and efficiency, a multi-objective optimization strategy is adopted during the training phase: the objective function is to minimize the equalization time T. eq and total energy consumption E eq .

[0028] Preferably, the equalization processing includes online equalization processing and offline equalization processing. In online equalization processing, the main circuit and auxiliary power supply work together, updating the input data of the intrinsic orthogonal decomposition-long short-term memory network (POD-LSTM) model once every preset cycle to regenerate the optimal equalization path. In offline equalization processing, an adaptive two-stage equalization is adopted. The initial stage uses a large current for rapid convergence, and the later stage switches to a small current for precise adjustment. The switching condition is as follows: or , For consistency scoring, This serves as a consistency metric for the initial equilibrium point. This is an empirical coefficient.

[0029] In this embodiment, after entering the equalization mode, it will determine in real time whether it is connected to the main power bus of the vehicle or energy storage: if the bus voltage is detected and the communication handshake is successful, it is identified as the online equalization processing mode. At this time, the DC / DC converter of the main circuit and the auxiliary power module cooperate in a "bidirectional energy pool" manner - every 20 minutes (preset cycle), the central processing unit automatically pulls the latest terminal voltage, SOC, SOH, temperature and internal resistance data, re-runs the POD-LSTM inference, and generates a new I eq-Δt sequence and immediately send it down; because the input data is kept "fresh", it can capture the capacity drift caused by load changes, ambient temperature changes and even sudden current surges, thus keeping the equalization error within 0.5%.

[0030] If the system disconnects from the main power bus or receives an "offline maintenance" instruction from maintenance personnel, it automatically switches to offline balancing mode. At this time, the strategy switches to an adaptive two-stage mode: the first stage uses a high-current pulse (≥1 C) to quickly discharge high-capacity cells or quickly replenish low-capacity cells, causing the consistency score to rapidly approach a preset multiple of the initial value within minutes; when it detects... or Upon switching, the system seamlessly transitions to the second stage, reducing the current to ≤0.2 C for millimeter-level fine-tuning until the index finally falls below the preset threshold. Because energy in the second stage is no longer dissipated as heat but is instead transferred between units via converters, it saves power compared to traditional high-current solutions and extends cycle life. The entire switching process is completed locally in a closed loop by the central processing unit, requiring no manual intervention. Maintenance personnel can simply view the progress bar on a mobile device via Bluetooth or Wi-Fi.

[0031] In summary, the battery balancing control system first obtains the terminal voltage of each battery through a voltage acquisition module, then uses an extended Kalman filter algorithm to estimate the state of charge (SOC) of each battery, and introduces the state of health (SOH) as a correction factor, while incorporating battery temperature as a compensation factor to dynamically calibrate the SOC estimate. The SOH of all batteries is estimated online using the incremental internal resistance method, and the usable capacity difference among all batteries is estimated based on their SOC states. Then, an optimal balancing path is generated based on an intrinsic orthogonal decomposition-long short-term memory (POD-LSTM) network. Under online balancing conditions, the main circuit and auxiliary power supply work together, updating the POD-LSTM input data and regenerating the path every 20 minutes. Under offline conditions, an adaptive two-stage balancing method is used. This method can quickly estimate the optimal balancing scheme using machine learning, thereby rapidly improving the consistency of the battery pack and broadening its application scenarios. This system improves the balancing speed and efficiency of lithium-ion battery packs and ensures the safe and stable operation of the battery system.

[0032] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A battery balancing control system, characterized in that, include: The system includes an auxiliary power module, a data acquisition component, a battery balancing module, a central processing unit, and an interface component. The data terminal of the central processing unit is electrically connected to the data terminals of the data acquisition component and the battery balancing module. The battery balancing module is electrically connected to the battery to be repaired and the auxiliary power module through the interface component. The data acquisition component is electrically connected to the battery to be repaired and the sensor through the interface component. The central processing unit is configured to perform the following steps by executing a computer program stored internally: The terminal voltage of each battery is acquired by the acquisition component, the terminal voltage of each battery is estimated, the estimated state of charge (SOC) value of each battery is obtained, and the estimated SOC value of the battery is dynamically calibrated according to the preset correction and compensation data to obtain the dynamically calibrated SOC value of the battery. Estimate the current SOH state of all batteries, combine it with the dynamically calibrated SOC state value of the batteries, calculate the usable capacity difference of all batteries, calculate the comprehensive consistency score based on the battery consistency evaluation index, and analyze the usable capacity difference and consistency score to determine whether battery equalization processing is required. When it is determined that battery balancing is required, the intrinsic orthogonal decomposition-long short-term memory network (POD-LSTM) model is invoked for prediction to generate the optimal balancing path.

2. The battery equalization control system according to claim 1, characterized in that, The interface components include a voltage charging / discharging line interface, a detection line independent interface, and a temperature measurement line interface. The central processing unit has a built-in battery health state (SOH) estimation module, which calculates the SOH value in real time using the internal resistance increment method and uses it for SOC calibration.

3. The battery equalization control system according to claim 2, characterized in that, The acquisition components include a voltage acquisition module and a temperature acquisition module. The voltage acquisition module is connected to the positive and negative terminals of the battery to be repaired through an independent interface of the detection line. The temperature acquisition module is connected to the battery temperature sensor through a temperature measurement line interface. The data terminal of the central processing unit is electrically connected to the data terminals of the voltage acquisition module and the temperature acquisition module.

4. The battery equalization control system according to claim 2, characterized in that, The battery balancing module includes a converter and a power dissipation resistor. One end of the converter is connected to the positive and negative terminals of the battery to be repaired through a voltage charging and discharging line interface, and the other end of the converter is connected to an auxiliary power module. The power dissipation resistor is connected in series with a control switch and is also connected to the positive and negative terminals of the battery to be repaired through a voltage charging and discharging line interface. The converter and the power dissipation resistor cannot be turned on simultaneously.

5. The battery equalization control system according to claim 1, characterized in that, It also includes a wireless communication module, through which the central processing unit communicates and interacts with an external host computer or cloud platform to exchange data. The wireless communication module is a Bluetooth or Wi-Fi module.

6. The battery equalization control system according to claim 3, characterized in that, The terminal voltage of each battery is acquired by the acquisition component. The terminal voltage of each battery is estimated to obtain the estimated state of charge (SOC) value of each battery. The estimated SOC value is then dynamically calibrated according to preset correction and compensation data to obtain the dynamically calibrated SOC value. Specifically: The terminal voltage of each battery is acquired by the voltage acquisition module, and the terminal voltage of each battery is estimated by the extended Kalman filter algorithm to obtain the estimated state of charge (SOC) value of each battery. The battery temperature T collected by the temperature acquisition module is obtained. Using battery temperature T as a compensation factor and the battery state of health (SOH) value as a correction factor, the estimated battery state of charge (SOC) is dynamically calibrated. The formula is as follows: ,in, This is the dynamically calibrated battery SOC state value. This is an estimated value for the battery's state of charge (SOC). An empirical coefficient related to battery type. This refers to the State of Health (SOH) value of the battery. This is a compensation function related to battery temperature, used to correct the effect of temperature on SOC estimation.

7. The battery equalization control system according to claim 1, characterized in that, Estimate the current State of Health (SOH) of all batteries, combine it with the dynamically calibrated State of Charge (SOC) values, calculate the difference in usable capacity among all batteries, and calculate a comprehensive consistency score based on the inter-battery consistency evaluation index, specifically: The current State of Health (SOH) of all batteries is estimated online using the incremental internal resistance method. Combined with the dynamically calibrated State of Charge (SOC) values, the difference in usable capacity among all batteries is calculated using the following formula: , Due to the difference in usable battery capacity, The difference between the rated usable capacity of the battery, This represents the current SOH state of the battery. This is a capacity difference correction function related to battery temperature, which is used to correct the effect of temperature on the difference in usable battery capacity. A battery consistency evaluation index is introduced, and a weighted calculation is performed based on this index to obtain a comprehensive consistency score. The formula is as follows: ,in, For consistency scoring, , , , All are empirical coefficients. This represents the voltage difference between the battery terminals. This represents the SOC difference of the battery. This represents the SOH difference of the battery. This represents the temperature difference of the battery.

8. The battery equalization control system according to claim 1, characterized in that, The available capacity difference and consistency score are analyzed to determine whether battery equalization processing is necessary. Specifically: Analyze the difference in usable capacity of the battery and consistency score To determine the difference in usable battery capacity Does it exceed the predetermined value? And the SOH decay rate is greater than the threshold or the consistency score Does it exceed the threshold? If so, perform battery balancing. If not, battery balancing will not be performed.

9. The battery equalization control system according to claim 1, characterized in that, The intrinsic orthogonal decomposition-long short-term memory network (POD-LSTM) model is used for prediction processing to generate the optimal equilibrium path, specifically: The POD algorithm is used to decompose the historical dataset, calculate the energy contribution rate of each mode, remove secondary modes with contribution rates less than the preset value, extract the high-dimensional feature modes of the battery pack state, and select multiple dominant modes as inputs to the LSTM network. A time-sliding window mechanism is employed, inputting battery state data from multiple consecutive time points into an LSTM to predict the equalization parameters at future time points. Based on these predicted parameters, an optimal equalization path is generated. The input time-series data to the LSTM network includes: terminal voltage, dynamically calibrated battery SOC value, current battery SOH state, battery temperature T, and internal resistance. The output is the equalization current I. eq Duration Δt and energy consumption; The intrinsic orthogonal decomposition-long short-term memory network (POD-LSTM) model employs a multi-objective optimization strategy during training, with its objective function being: minimizing the equilibrium time t. eq and total energy consumption E eq The formula is: Its constraints are: , To balance time weights, As the weight of total energy consumption, Minimum allowable temperature, This is the maximum permissible temperature.

10. The battery equalization control system according to claim 1, characterized in that, The equalization process includes online equalization and offline equalization. In online equalization, the main circuit and auxiliary power supply work together, updating the input data of the intrinsic orthogonal decomposition-long short-term memory network (POD-LSTM) model once every preset cycle to regenerate the optimal equalization path. In offline equalization, an adaptive two-stage equalization is employed. The initial stage uses a large current for rapid convergence, while the later stage switches to a small current for precise adjustment. The switching condition is as follows: , For consistency scoring, This serves as a consistency metric for the initial equilibrium point. This is an empirical coefficient.

Images