Low electric quantity balance protection method based on battery health degree and charging and storage integrated power supply equipment

Abstract

The invention relates to the technical field of battery management, in particular to a low-electric-quantity balance protection method based on the health degree of a battery and charging and storage integrated power supply equipment. According to the scheme, a three-dimensional continuous risk probability field fusing the state of charge, the single voltage difference and the health degree of the battery is constructed, and the risk gradient abrupt change feature is used as a criterion; and dynamically determining a safety voltage critical value matched with the current health state, and guiding risk adaptive equalization execution according to the safety voltage critical value. The invention solves or at least alleviates the problems of over-discharge risk identification lagging and equalization strategy misalignment caused by the fact that the battery health degree is not brought into a low-electric-quantity safety boundary dynamic correction system in the prior art, and provides a low-electric-quantity equalization protection method based on the battery health degree and charging and storage integrated power supply equipment.

Description

Technical Field

[0001] This invention belongs to the field of battery management technology, and particularly relates to a low-charge equalization protection method based on battery health and an integrated charging and storage power supply device. Background Technology

[0002] Lithium-ion batteries, as the core carrier of current electrochemical energy storage systems, have been deeply integrated into key application scenarios such as electric vehicles, large-scale energy storage power stations, and high-reliability portable power devices. In these systems, the operational safety and service life of battery packs, composed of multiple individual cells connected in series and parallel, highly depend on the battery management system's ability to maintain the consistent state of each individual cell, with equalization protection technology being particularly critical. Especially in the low-charge operating range (state of charge below 10%), the electrochemical reaction kinetics within the battery deteriorate significantly, and even small voltage deviations can induce irreversible lithium plating or structural collapse, leading to the risk of thermal runaway. Therefore, accurately identifying and suppressing the risk of over-discharge under this sensitive operating condition has become a core issue in ensuring the safe operation of the system throughout its entire lifecycle.

[0003] In existing technologies, most mainstream solutions use fixed thresholds or dynamic adjustment mechanisms based on internal resistance compensation to set over-discharge protection boundaries. For example, patent CN107231010B proposes using a formula... Linear correction of the over-discharge threshold involves the internal resistance of individual cells. This method considers the impact of current and internal resistance on terminal voltage to some extent and can compensate for ohmic voltage drop. However, its model is essentially a static linear framework and does not incorporate battery health, a core state variable reflecting the cumulative effects of long-term aging, into the risk assessment system. More importantly, this scheme does not consider the strong nonlinear coupling relationship between voltage, state of charge, and health in the low-charge region. When the health drops to 70%, to avoid over-discharge, the critical safe voltage value should shift upward by at least 0.05 volts, and this nonlinear shift cannot be captured by simple linear compensation of internal resistance. Correspondingly, while another scheme, CN119051221B, introduces the concept of performance degradation gradient and uses it to optimize the timing and intensity of equalization intervention, its risk judgment logic is still based on preset thresholds or empirical rules. It lacks the ability to quantitatively model the continuous evolution of risk and does not construct a multidimensional risk space with health as the key dimension. This makes it difficult to achieve adaptive correction of the safety boundary when dealing with battery packs with large health dispersion operating at low charge levels.

[0004] However, with the increasing demands for safety redundancy in high-energy-density battery systems, and the growing inconsistency in aging faced by battery packs in scenarios such as secondary use and long-term service, the aforementioned technical approaches reveal deep-seated contradictions at the principle level: On the one hand, the traditional binary safety judgment mechanism (i.e., discrete judgment of "safe" or "dangerous") cannot characterize the physical nature of risk evolving continuously with state, causing frequent misjudgments in the critical region and a high measured misjudgment rate; on the other hand, even if parameters such as internal resistance are introduced for threshold compensation, if health is not used as an independent and dominant risk adjustment factor, and a three-dimensional risk probability field is not constructed by combining state of charge and voltage difference, it is impossible to accurately identify the critical point where the risk probability gradient changes abruptly, thus making it difficult to achieve the adaptive protection logic of "the lower the health, the higher the safety threshold". The reason for this is that the aging effect of batteries in the low-charge region does not act uniformly on all electrochemical parameters, but rather nonlinearly compresses its safe voltage window by changing mechanisms such as active material utilization, interface impedance, and lithium inventory loss. This compression effect varies significantly among individuals with different levels of health and interacts in a complex way with the current state of charge, making it fundamentally impossible for compensation strategies based on single-dimensional or linear models to cover the dynamic topology of the true risk distribution.

[0005] Based on this, the fundamental challenge facing existing technological systems can be summarized as follows: how to overcome the limitations of traditional threshold protection and binary judgment without increasing hardware complexity, and construct a three-dimensional continuous risk probability field that integrates state of charge, individual cell voltage difference, and battery health. Furthermore, based on the abrupt changes in the risk gradient within this field, a safe voltage threshold that precisely matches the current health state should be dynamically and nonlinearly determined, thereby achieving proactive suppression of over-discharge risk and precise guidance of balancing strategies under low-charge operating conditions. Therefore, establishing a low-charge balancing protection mechanism driven by health status, criterion-based on abrupt changes in risk probability gradients, and possessing self-learning and self-correction capabilities has become a key technical challenge urgently needing to be addressed by those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art, solve or at least alleviate the technical problems in the prior art caused by the failure to incorporate battery health as a core variable into the dynamic correction system of the safety boundary of the low-power operation range, resulting in lagging over-discharge risk identification, high misjudgment rate and inaccurate balancing strategy. This invention provides a low-power balancing protection method and an integrated charging and storage power supply device based on battery health.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a low-charge equalization protection method based on battery health, comprising the following steps: S1. Risk Probability Field Construction and Dynamic Update: Construct a three-dimensional parameter space for state of charge, individual cell voltage difference, and battery health, and preset an initial risk probability distribution; continuously collect battery pack state data and extract feature parameters related to over-discharge risk; calculate the contribution weight of each feature parameter to the risk probability and incorporate it into the model to update the probability distribution of low-charge areas; calculate the risk probability gradient and mark high-risk areas. S2. Risk probability correction associated with health status: When the risk probability of the marked area changes significantly, a health status assessment is triggered; health status data is acquired and aging characteristics are extracted, the influence coefficient of health status on the risk threshold is determined by associating the risk gradient, and a mapping relationship is generated to correct the risk probability field; S3. Determination of dynamic safety threshold: Analyze the corrected risk probability field and locate the risk critical point; map the critical point to the safety voltage critical value, integrate real-time characteristic parameters for fine-tuning, and generate a dynamic safety voltage critical value with applicable conditions and timeliness. S4. Risk-guided equilibrium execution: Map the current battery state to the corrected risk probability field to determine the risk level; combine the dynamic safety threshold to determine the urgency of equilibrium, configure equilibrium parameters and execute, and monitor risk changes during the process.

[0008] To further realize the present invention, the following technical solutions may be preferred: Preferably, in step S1, the three-dimensional parameter space is discretized into a segmented non-uniform grid, with a denser grid in the low-charge region; the extracted feature parameters include the extreme value of the single-cell voltage, voltage dispersion, discharge current characteristics, temperature change rate, and state of charge change rate; the contribution weight is dynamically adjusted according to the state of charge, and the weight of voltage-related feature parameters is increased in the low-charge region; the update intensity is adaptively adjusted according to the state of charge change rate.

[0009] Preferably, in step S2, a health assessment is triggered when the risk probability of the marked low-charge area changes more than a threshold over multiple consecutive cycles; an electrochemical impedance spectroscopy measurement is performed by prioritizing the selection of a time window when the state of charge and discharge current are stable, applying sinusoidal voltage excitations of different frequencies in sequence, collecting the current response, and extracting the real part impedance increment in the mid-frequency band as a health characterization index; the battery health is calculated by the normalized difference between the impedance increment and the new battery reference value and the life end threshold.

[0010] Preferably, step S2 further includes: simultaneously assessing the confidence level of the health data, taking into account the temperature stability, current stability, and state of charge range during measurement; when the confidence level is lower than the threshold, reducing the risk probability field correction intensity and initiating the data verification mechanism; the mapping relationship is reflected in the fact that the lower the health level, the higher the corrected risk probability, and the risk critical region shifts towards the high voltage direction.

[0011] Preferably, in step S3, the accuracy of the risk critical point positioning is dynamically adjusted based on the confidence level of the health data. When the confidence level is high, an interpolation method is used for precise positioning, and when the confidence level is low, an approximation method is used. The risk critical point is mapped to a safe voltage critical value, and real-time temperature and current fine-tuning are integrated to generate a dynamic safe voltage critical value with applicable conditions and a time-sensitive confidence interval.

[0012] Preferably, in step S4, risk levels are divided according to risk probability; the urgency of balancing is determined by the difference between the current lowest single-cell voltage and the dynamic safety threshold; the balancing current intensity is configured according to the risk level, with high risk corresponding to low current and low risk corresponding to high current; if the remaining validity period of the dynamic safety threshold is insufficient, the duration of a single balancing is shortened.

[0013] Preferably, the method further includes the following steps; S5. Risk Probability Field Self-Learning Optimization: Compare the predicted risk threshold with the actual over-release event location, calculate the error and analyze the related factors; adjust the risk probability field model parameters and the mapping relationship between health and risk probability, and optimize the health assessment strategy. In step S5, the correlation between error and health status, single-unit voltage difference, and temperature is analyzed; the weights of the risk probability field model parameters are adjusted to enhance the nonlinear response in the low health status region; the mapping relationship between health status and risk probability is corrected to a piecewise exponential form; and the health status assessment strategy is optimized by increasing temperature compensation and the density of low-frequency measurement points in the low power region.

[0014] A charging and storage integrated power device for implementing the above method includes a battery pack, a central collaborative controller, a multi-channel voltage and current acquisition unit, a health assessment module, a risk probability field modeling engine, a dynamic threshold calculation unit, and an equalization execution unit. The battery pack is composed of multiple lithium-ion cells connected in series and parallel; the multi-channel voltage and current acquisition unit synchronously acquires the voltage, branch current, and temperature signals of each cell; the health assessment module integrates an electrochemical impedance spectroscopy measurement circuit to perform battery health assessment; the risk probability field modeling engine constructs and updates a three-dimensional risk probability field; the dynamic threshold calculation unit generates a dynamic safe voltage threshold; the equalization execution unit controls the energy transfer between cells; and the central coordination controller schedules the operation of each module.

[0015] Preferably, the risk probability field modeling engine discretizes and stores the three-dimensional parameter space; the electrochemical impedance spectroscopy measurement circuit of the health assessment module covers a wide frequency range and has adjustable amplitude; the bidirectional energy transfer channel of the equalization execution unit adopts an inductor-switch topology structure, and the equalization current is continuously adjustable.

[0016] Preferably, the central collaborative controller adopts a dual-core heterogeneous architecture, with the main core responsible for task scheduling and logical decision-making, and the co-core responsible for real-time control instruction generation; it is internally configured with a dual buffer mechanism to isolate data access conflicts between risk probability field updates and balanced execution; data writing is triggered by the direct memory access controller, and reading is driven by hardware timer interrupts.

[0017] The beneficial effects of this invention are: This invention overcomes the limitations of traditional binary safety judgment and linear threshold compensation by constructing a three-dimensional continuous risk probability field with health as the core dimension; it utilizes the risk gradient mutation characteristics as a safety critical criterion to achieve nonlinear adaptive correction of the safety voltage boundary; combined with health confidence management and a closed-loop self-learning mechanism, it ensures the long-term robustness and prediction accuracy of the system under scenarios of inconsistent battery aging and complex and variable operating conditions; and it achieves forward-looking suppression of over-discharge risk and precise distribution of balanced energy in the low-charge sensitive range, thereby improving the operational safety and service life of the battery pack. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the composition of the method of the present invention; Figure 2 This is a timing diagram of the method of the present invention; Figure 3 A schematic diagram of the three-dimensional risk probability field constructed for this invention and its gradient distribution in the low-electricity region; Figure 4 This is a schematic diagram illustrating the configuration of the balanced current intensity according to the risk level in this invention. Figure 5 This is a closed-loop feedback flowchart of the risk probability field self-learning optimization mechanism of the present invention; Figure 6 This is a block diagram of the integrated charging and storage power supply device of the present invention; Figure 7 This is a comparative schematic diagram of Embodiment 3 of the present invention. Detailed Implementation

[0019] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1

[0022] This embodiment discloses a charging and storage integrated power supply device, referring to... Figure 6 The system comprises a battery pack, a central co-controller, a multi-channel voltage and current acquisition unit, a health assessment module, a risk probability field modeling engine, a dynamic threshold calculation unit, and an equalization execution unit. The battery pack consists of multiple lithium-ion cells connected in series and parallel, typically in a 16-series, 4-parallel configuration, with a total nominal voltage of 51.2 volts and a rated capacity of 100 amp-hours. The multi-channel voltage and current acquisition unit uses a high-precision analog front-end chip, featuring 16 synchronous voltage sampling channels and 4 branch current sampling channels, with a 16-bit sampling resolution and a 10-millisecond sampling period. It also integrates a digital temperature sensor, with each pair of cells sharing a temperature measurement node; the temperature sampling period is synchronized with the voltage sampling. The health assessment module incorporates a dedicated electrochemical impedance spectroscopy measurement circuit, including a programmable AC excitation source, differential signal conditioning circuitry, and a high-speed analog-to-digital converter. The risk probability field modeling engine consists of an embedded coprocessor and a dedicated memory array. The coprocessor supports floating-point matrix operations and real-time gradient field solving, while the memory array caches discretized grid data in the three-dimensional parameter space. The dynamic threshold calculation unit, centered on a microcontroller, runs a safety voltage critical value mapping algorithm and receives data input from the risk probability field modeling engine and the health assessment module. The balancing execution unit employs a distributed topology, with each battery string configured with a bidirectional energy transfer channel. Each channel is driven by a high-speed optocoupler-isolated MOSFET switch, enabling directional energy flow between individual cells. All functional modules are interconnected via a CAN FD high-speed serial bus, with a central coordinating controller centrally scheduling the runtime sequence and data interaction of each module.

[0023] Furthermore, the energy transfer circuit of the equalization execution unit adopts an inductor-switch topology. Each equalization channel includes a pair of complementary N-channel MOSFET switches and a shared energy storage inductor. The inductance value is 10 microhenries, and the MOSFET on-resistance is less than 5 milliohms. The equalization current intensity is continuously adjustable by adjusting the switch duty cycle through a pulse width modulation signal, with a duty cycle resolution of no less than 0.5%, corresponding to a current adjustment step of approximately 10 mA. The central coordinating controller generates a duty cycle signal based on the risk level command and prevents shoot-through between the upper and lower bridge arms through dead-time control.

[0024] Furthermore, the central coordinating controller adopts a dual-core heterogeneous architecture. The main core is responsible for task scheduling and logical decision-making, while the coordinating core is dedicated to real-time control command generation. Internally, it features a dual-buffer mechanism to isolate data access conflicts between the risk probability field update process and the balancing execution process. Write operations are automatically triggered by the direct memory access controller at the end of each sampling period, writing newly acquired state data to a backup buffer. Read operations are driven by hardware timer interrupts, reading data from the currently valid buffer at the beginning of each control cycle, ensuring the consistency and timeliness of the state information upon which control commands are based.

[0025] As a preferred embodiment of the present invention, the risk probability field modeling engine analyzes the state of charge (… ), individual unit voltage difference ( ) and battery health ( The three-dimensional parameter space, consisting of three dimensions, is discretized into a segmented non-uniform grid. The charge state dimension is divided into 100 equally spaced sub-intervals within the 0% to 15% range, each with a width of 0.15 percentage points; and into 20 sub-intervals above 15%, each with a width of 4.25 percentage points. The individual cell voltage difference dimension uses the maximum voltage deviation between adjacent cells as a benchmark, dynamically scaling the grid size: when the maximum deviation is less than 10 mV, the grid step size is set to 1 mV; when the deviation is between 10 and 50 mV, the step size is set to 5 mV; and when the deviation exceeds 50 mV, the step size expands to 10 mV. The health dimension has discrete nodes set at 5% steps, covering a health range of 70% to 100%, with a total of seven working nodes. Each grid node stores the risk probability value for the corresponding state combination. and its gradient vector The gradient calculation uses the central difference method, and the boundary region uses forward or backward difference correction.

[0026] Example 2

[0027] Based on the hardware of Embodiment 1, this embodiment further discloses a low-power balancing protection method based on battery health, referring to... Figures 1-5 Specifically, it includes five stages: risk probability field construction and dynamic updating, risk probability correction related to health, dynamic safety threshold determination, balanced execution of risk guidance, and self-learning optimization of the risk probability field.

[0028] First, during the risk probability field construction and dynamic update phase, a preset initial risk probability distribution function is loaded during system power-on initialization. This function is trained based on a historical over-discharge event database and focuses on characterizing the nonlinear risk gradient characteristics in the region where the state of charge is below 10%. The initial function form is as follows:

[0029]

[0030] in It is the Sigmoid activation function. For each dimension, the basis functions are... This is a coupling term between the state of charge and the voltage difference, with coefficients... The maximum likelihood estimation fitting was used to determine the value. Subsequently, the system continuously collected data on individual cell voltages, branch currents, and temperatures at a 10-millisecond sampling period, and extracted feature parameters, including the lowest individual cell voltage, using sliding window midpoint filtering and wavelet denoising algorithms. Highest single-unit voltage Voltage standard deviation , RMS value of discharge current Temperature change rate Based on this, the system calculates the contribution weight of each feature parameter to the risk probability. The weight update rule is as follows:

[0031]

[0032] in For learning rate, For the first One feature parameter, This represents the change in state of charge (SOC) within the current sampling period. Next, the system substitutes the updated feature parameters and their dynamic weights into the risk probability field model to locally recalculate the probability distribution for the low charge region (SOC < 15%); and updates the intensity. According to the rate of change of state of charge Adaptive adjustment:

[0033]

[0034] in This is a scaling factor, typically 200. After the update, the system will calculate the gradient distribution of the risk probability field in the low-power region and identify those that meet the criteria. spatial points, among which To preset a safety threshold, typically 0.8 per percentage point, these points are then marked as high-risk critical areas for subsequent health correlation analysis.

[0035] Next, in the health-related risk probability correction phase, if the system detects a significant change in the risk probability within the low power area (i.e., within three consecutive sampling periods)... , When a threshold value (typically 0.05) is set, a high-precision health assessment process is immediately triggered. This process prioritizes time windows where the rate of change of state of charge is less than 0.5% per minute and the standard deviation of discharge current fluctuation is less than 0.5 amperes for electrochemical impedance spectroscopy (EIS) measurement. During measurement, the excitation source sequentially applies sinusoidal voltage signals of different frequencies, acquires the corresponding current responses, obtains the frequency domain impedance spectrum after Fast Fourier Transform (FFT), and then reconstructs the time domain response through Inverse Fast Fourier Transform (IFFT) to extract the real part impedance increment in the mid-frequency band (typically 10 Hz). As an indicator of health. Health Calculate using the following formula:

[0036]

[0037] in This is the reference value for the real part impedance of the new battery at intermediate frequency. This is the end-of-life threshold. Afterwards, the system will analyze the acquired data. Data is used to extract aging characteristic parameters (such as impedance growth rate and capacity decay rate), which are then compared with the previously calculated risk gradient. Pearson correlation analysis was conducted to determine the impact coefficient of health status on the risk threshold. Then establish a mapping relationship between health status and risk probability:

[0038]

[0039] This correction logic manifests as follows: the lower the health level, the more the boundary of the risk critical region shifts towards higher voltage, ultimately generating a risk probability field model after health level correction. Additionally, the system simultaneously assesses the confidence level of the health level data. This evaluation will take into account the temperature stability at the time of measurement. ), current stability ( ) and SOC interval ( Factors such as ); if satisfied It will automatically reduce the correction intensity to 60% of the original value and activate the data verification mechanism to remeasure in the next stable window.

[0040] Third, at the stage of determining the dynamic safety threshold, the system will analyze the risk probability field after health correction and locate the conditions that meet the requirements. spatial point Combined with the confidence level of health Dynamically adjust the critical point positioning accuracy: when the following conditions are met When, use cubic spline interpolation for precise positioning; when satisfying At this point, the nearest neighbor grid point is used as an approximation. After determining the precise risk critical point, the system maps it to the safe voltage critical value. The mapping formula is:

[0041]

[0042] in This is the nominal open-circuit voltage. This is a safety margin factor, fine-tuned by real-time characteristic parameters (such as temperature and current). The final generated safe voltage threshold includes applicable conditions (such as temperature range and current limit) and a time-based confidence interval (typically 300 seconds). The system will verify the consistency of this threshold with the historical safe operation database. If the actual lowest single-unit voltage and... If the deviation exceeds 50 millivolts, the process will backtrack to the risk probability field construction stage for reassessment. After successful verification, a timestamped safety threshold packet will be generated and then passed to the load balancing module.

[0043] Fourth, during the risk-guided equilibrium execution phase, the system will update the current battery status. Mapped to the health-corrected risk probability field, the current risk probability is obtained by looking up a table. Then, risk levels are determined based on this: Low risk Medium risk. This is considered high-risk. (Combined with dynamic safety voltage threshold) and the current lowest single-cell voltage The difference To determine the urgency of the balancing operation: if the following conditions are met... Furthermore, if the risk level is high, it is marked as an emergency equilibrium; if the following conditions are met... Furthermore, if the risk level is medium risk, it is marked as regular balancing; otherwise, balancing is temporarily suspended. The balancing requirement report includes a risk level identifier, a suggested balancing current, and a maximum allowable duration. The system then configures the balancing current intensity according to the risk level: 30% of the rated value for high risk, 60% for medium risk, and 100% for low risk. Simultaneously, considering the timeliness of the safety threshold, if the remaining validity period is less than 60 seconds, the single balancing duration is shortened to less than 10 seconds. The balancing execution command is sent to the balancing execution unit via the CAN FD bus. During the balancing process, the system monitors the risk probability field every 50 milliseconds; if the risk level jumps one level, the balancing current is immediately reduced by 20%; if a change is detected... If the condition is not met, the equalization process is paused and a recalibration request is sent to the safety threshold module. After the equalization process is complete, the system records the entire process. , Equalizing current Data is used to calculate the decrease in risk probability before and after equilibrium. The effect data is then fed back to the risk probability field construction module to optimize the feature parameter weight update strategy.

[0044] Fifth, during the risk probability field self-learning optimization phase, the system periodically (every hundred charge-discharge cycles or 500 hours of cumulative operation) compares the predicted risk critical points. The actual point of over-discharge event (Defined as the first time the single-cell voltage falls below 2.5 volts at the state of charge (SOC) value), calculate the prediction error. .like Then initiate error attribution analysis: identify errors through a multiple regression model. and , The correlation coefficients of variables such as mean temperature are used to form a list of key areas for learning and optimization. Based on this list, the system adjusts the parameter weight distribution in the risk probability field model, for example, for low health (…). ) area The basis function introduces a quadratic term to enhance the nonlinear response; simultaneously, it modifies the mapping relationship between health status and risk probability, changing the linear correction to a piecewise exponential form:

[0045]

[0046] in The aging sensitivity coefficient is determined by minimizing the sum of squared historical errors. Furthermore, the system optimizes the triggering logic for health assessment based on error analysis results: if high-risk false alarms are found to occur frequently in low-temperature environments, a temperature compensation term is added; simultaneously, the excitation strategy for electrochemical impedance spectroscopy is adjusted. The measurement density of low-frequency points (0.5Hz) is increased in the interval to improve the resolution in the early stages of aging. After the new model and strategy are generated, the system is cross-validated in a simulation environment: the prediction accuracy is tested using fifty sets of historical operating condition data that were not used in the training. If the accuracy improves by more than 5% and there are no new missed reports, it is solidified as the standard operating parameters of the system, and a structured learning report is generated and stored in the system knowledge base.

[0047] Example 3

[0048] To verify the technical effects of the present invention, the following reference examples and comparative examples are designed in this embodiment.

[0049] The method and equipment described in this invention are used to manage the low-charge capacity of a group of sixteen lithium iron phosphate cells (initial capacity 100 AH, initial SOH 100%). After the battery pack undergoes accelerated aging cycles, the SOH drops to 78%, at which point it enters the low-charge operating range (SOC = 8%). The system constructs a three-dimensional risk probability field and detects a risk gradient abrupt change point at SOC = 6.2%, corresponding to a critical safe voltage value of 2.93 V. The health assessment module completes impedance measurement at SOC = 7.5%, confirming SOH = 78.3% with a confidence level of 0.85. After health correction, the critical safe voltage value shifts upward to 2.97 V. The balancing execution unit initiates balancing at a medium risk level, with the balancing current set to 1.8 A (60% of the rated value). After 45 seconds, the lowest single-cell voltage rises back to 3.02 V, and the risk probability drops from 0.68 to 0.41, without triggering over-discharge protection.

[0050] In the comparative study, a traditional fixed threshold method (with a constant safety voltage threshold of 2.85 volts) was used to manage the aging batteries in the same group. When the SOC dropped to 7.1%, the lowest single-cell voltage had fallen to 2.83 volts, triggering over-discharge protection. The system forcibly disconnected the load, resulting in a loss of approximately 13% of usable energy. The balancing operation failed to be completed due to the over-discharge interruption, exacerbating voltage inconsistency.

[0051] The table below summarizes the key performance indicators of the reference and comparative examples under the same aging conditions:

[0052] project Reference example (of this invention) Comparative example (fixed threshold) SOC when over-discharge protection is triggered Not triggered 7.1% Minimum single-cell voltage maintenance level ≥ 2.97 V 2.83 V Low power range available energy utilization rate 98.5% 86.2% Balanced completion rate 100% 0% Individual cell voltage standard deviation (after equalization) 8 mV 22 mV

[0053] Reference Figure 7 The data above shows that by integrating battery health to dynamically correct safety boundaries, this invention effectively avoids premature load disconnection and improves energy utilization efficiency and balance effectiveness in low-power ranges.

[0054] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A low-charge equalization protection method based on battery health, characterized in that, Includes the following steps: S1. Risk Probability Field Construction and Dynamic Update: Construct a three-dimensional parameter space for state of charge, individual cell voltage difference, and battery health, and preset an initial risk probability distribution; continuously collect battery pack state data and extract feature parameters related to over-discharge risk; calculate the contribution weight of each feature parameter to the risk probability and incorporate it into the model to update the probability distribution of low-charge areas; calculate the risk probability gradient and mark high-risk areas. S2. Risk probability correction associated with health status: When the risk probability of the marked area changes significantly, a health status assessment is triggered; health status data is acquired and aging characteristics are extracted, the influence coefficient of health status on the risk threshold is determined by associating the risk gradient, and a mapping relationship is generated to correct the risk probability field; S3. Determination of dynamic safety threshold: Analyze the corrected risk probability field and locate the risk critical point; The critical point is mapped to the critical value of safe voltage, and real-time characteristic parameters are integrated for fine-tuning to generate a dynamic critical value of safe voltage with applicable conditions and timeliness. S4. Risk-guided balanced execution: Map the current battery state to the corrected risk probability field to determine the risk level; The urgency level of the load balance is determined by combining dynamic safety thresholds, load balance parameters are configured and executed, and risk changes are monitored during the process.

2. The method according to claim 1, characterized in that, In step S1, the three-dimensional parameter space is discretized into a segmented non-uniform grid, with a denser grid in the low-charge region; the extracted feature parameters include the extreme value of the single-cell voltage, voltage dispersion, discharge current characteristics, temperature change rate, and state of charge change rate; the contribution weight is dynamically adjusted with the state of charge, and the weight of voltage-related feature parameters is increased in the low-charge region. The update intensity is adaptively adjusted according to the rate of change of state of charge.

3. The method according to claim 1, characterized in that, In step S2, a health assessment is triggered when the risk probability of the marked low-charge area changes beyond a threshold over multiple consecutive cycles. A time window with stable state of charge and discharge current is selected first, and electrochemical impedance spectroscopy is performed. Different frequency sinusoidal voltage excitations are applied sequentially, and the current response is collected. The real impedance increment in the mid-frequency band is extracted as a health characterization index. Battery health is calculated using the normalized difference between the impedance increment and the new battery baseline value and the lifespan termination threshold.

4. The method according to claim 3, characterized in that, Step S2 further includes: synchronously assessing the confidence level of the health data, taking into account the temperature stability, current stability, and state of charge range during measurement; when the confidence level is lower than the threshold, reducing the risk probability field correction intensity and initiating the data verification mechanism; the mapping relationship is reflected in the fact that the lower the health level, the higher the corrected risk probability, and the risk critical region shifts towards the high voltage direction.

5. The method according to claim 4, characterized in that, In step S3, the accuracy of risk critical point positioning is dynamically adjusted based on the confidence level of health data. When the confidence level is high, interpolation is used for precise positioning, and when the confidence level is low, an approximation method is used. The risk critical point is mapped to a safe voltage critical value, and real-time temperature and current fine-tuning are integrated to generate a dynamic safe voltage critical value with applicable conditions and a time-sensitive confidence interval.

6. The method according to claim 1, characterized in that, In step S4, risk levels are classified according to risk probability; The urgency of the balancing process is determined by combining the difference between the current lowest unit voltage and the dynamic safety threshold. The current intensity is configured according to the risk level, with low current corresponding to high risk and high current corresponding to low risk. If the remaining validity period of the dynamic safety threshold is insufficient, shorten the duration of a single equilibrium.

7. The method according to claim 1, characterized in that, It also includes the following steps; S5. Risk Probability Field Self-Learning Optimization: Compare the predicted risk critical point with the actual over-release event location, calculate the error and analyze the related factors; Adjust the parameters of the risk probability field model and the mapping relationship between health status and risk probability to optimize the health status assessment strategy; In step S5, the correlation between error and health status, individual unit voltage difference, and temperature is analyzed. Adjust the parameter weights of the risk probability field model to enhance the nonlinear response in low-health regions; The mapping relationship between health status and risk probability was corrected to a piecewise exponential form; the health status assessment strategy was optimized by increasing the density of low-frequency measurement points in temperature compensation and low-power areas.

8. A charging and storage integrated power supply device, used to implement the method according to any one of claims 1 to 7, characterized in that, It includes a battery pack, a central coordinating controller, a multi-channel voltage and current acquisition unit, a health assessment module, a risk probability field modeling engine, a dynamic threshold calculation unit, and an equalization execution unit; The battery pack is composed of multiple lithium-ion cells connected in series and parallel; the multi-channel voltage and current acquisition unit synchronously acquires the voltage, branch current and temperature signals of each cell; the health assessment module integrates an electrochemical impedance spectroscopy measurement circuit to perform battery health assessment; the risk probability field modeling engine constructs and updates a three-dimensional risk probability field; the dynamic threshold calculation unit generates a dynamic safe voltage critical value. The equalization execution unit controls the transfer of energy between individual units; the central coordination controller schedules the operation of each module.

9. The device according to claim 8, characterized in that, The risk probability field modeling engine discretizes and stores the three-dimensional parameter space; the electrochemical impedance spectroscopy measurement circuit of the health assessment module covers a wide frequency range and has adjustable amplitude; the bidirectional energy transfer channel of the equalization execution unit adopts an inductor-switch topology structure, and the equalization current is continuously adjustable.

10. The device according to claim 8, characterized in that, The central collaborative controller adopts a dual-core heterogeneous architecture, with the main core responsible for task scheduling and logical decision-making, and the co-core responsible for real-time control instruction generation. It is internally configured with a dual buffer mechanism to isolate data access conflicts between risk probability field updates and balanced execution. Data writing is triggered by the direct memory access controller, and reading is driven by hardware timer interrupts.

Images