A safety protection method for lithium battery

Abstract

This application discloses a safety protection method for lithium batteries, comprising: constructing a dynamic battery network topology through graph theory modeling, mapping energy transfer paths and updating weights in real time; combining differential flatness theory and state observer to predict key parameter trends and identify key nodes and risk propagation paths in the network; further employing feedforward control to proactively reduce the energy of key nodes through preventive micro-discharge before parameter anomalies occur, forming a local buffer; simultaneously embedding passive waveguide suppression elements to block electromagnetic energy transfer, and dynamically adjusting circuit parameters based on real-time energy flow data to compensate for unsteady-state losses; and finally, continuously correcting the model and control strategy through global monitoring and closed-loop optimization to achieve full closed-loop collaborative control, thereby significantly improving the safety and robustness of the system under dynamic operating conditions.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and in particular to a safety protection method for lithium batteries. Background Technology

[0002] With the rapid development of new energy vehicles and large-scale energy storage systems, the scale of lithium battery packs is constantly expanding, posing severe challenges to their safe operation. Under non-steady-state conditions, battery parameters (such as voltage and temperature) are prone to rapid transient changes, exacerbating the complexity of energy transfer between battery packs. Traditional protection systems mostly rely on fixed thresholds for alarms or power cuts, which are difficult to dynamically adapt to changes in operating conditions, are prone to response lag, and cannot effectively block long-distance energy transfer caused by thermal radiation and electromagnetic coupling due to individual cell failure, thus significantly increasing the risk of cascading failures. Existing solutions are mostly focused on local protection at the individual cell level, lacking real-time perception and coordinated control of the dynamic energy flow of the entire system, making it difficult to cope with complex disturbances in actual operation such as dynamic loads and sudden changes in ambient temperature, and the system-level safety redundancy is clearly insufficient.

[0003] Currently, lithium battery safety protection mainly relies on monitoring parameters such as voltage and temperature based on fixed thresholds and passive cut-off protection. These methods are effective under steady-state conditions. However, when large-scale battery arrays face non-steady-state conditions such as dynamic loads and sudden changes in ambient temperature, battery parameters change rapidly and energy transfer paths become complex. Traditional methods, due to response lag, are unable to effectively block the spread of thermal runaway, especially failing to suppress long-distance cascade failures caused by electromagnetic coupling and thermal radiation. Existing solutions mostly focus on the independent protection of individual cells, lacking real-time modeling and collaborative control of the overall energy flow dynamics of the battery pack. This makes it difficult to accurately predict and actively intervene before risks occur, resulting in insufficient overall system safety redundancy and difficulty in adapting to highly dynamic and complex real-world application scenarios. Summary of the Invention

[0004] This application provides a safety protection method for lithium batteries, which improves lithium battery safety by using dynamic network modeling and intelligent feedforward control to achieve early prediction, proactive intervention and continuous optimization of battery risks.

[0005] This application provides a safety protection method for lithium batteries, including: S1 uses a high-frequency sensor array to collect multi-dimensional data and environmental data of each individual cell in the battery pack in real time, and uses graph theory to model the battery pack as a dynamic network topology graph, where nodes represent individual cells, edges represent energy transfer paths, and the weights of the edges are updated in real time according to the operating conditions, thereby constructing a battery network model that can reflect the dynamics of energy flow in real time; the target parameters in the multi-dimensional data include the voltage, current, temperature, pressure, and gas concentration of each individual cell in the battery pack; S2, based on a dynamic battery network model, uses an electrothermal model and state observer constructed with differential flatness theory to predict the changing trends of key parameters of each individual battery cell, and combines graph theory algorithms to identify key nodes in the network and potential thermal runaway risk propagation paths. S3, based on the prediction and identification results of step S2, before the predicted value of a certain parameter in the multi-dimensional data exceeds the safety threshold, the battery management unit drives the self-discharge circuit to implement preventive micro-discharge on the key node, actively reducing its energy to form a local energy buffer. S4, passive suppression elements are embedded between each individual cell in the battery pack to block the transfer of electromagnetic energy. At the same time, the battery management unit dynamically adjusts the parameters of the self-discharge circuit to compensate for unsteady losses based on real-time energy flow data. S5 continuously monitors the overall status of the battery pack and compares the actual data with the predicted values. Based on the comparison error, it dynamically corrects the prediction model and the control parameters of the self-discharge circuit through optimization algorithms.

[0006] Preferably, the identification of key nodes and potential thermal runaway risk propagation paths in the network specifically includes: constructing an electrothermal state-space model using differential flatness theory, and using a state observer to estimate the internal state of the battery and predict parameters based on real-time data; in the dynamic battery network model, calculating the comprehensive node influence degree for each node, which integrates its topological importance, predicted temperature rise trend, and real-time energy state indicators, and identifying key nodes accordingly; taking the cell that is predicted to reach the thermal runaway temperature earliest as the risk source, and determining the most likely risk propagation path based on the dynamic coupling strength and failure probability in the network.

[0007] Preferably, the node influence degree is calculated using the following formula: , It is a node importance score based on the current graph topology. It is the predicted rate of temperature rise. It is the future The amount of temperature change within the time window, It is the length of the predicted future time interval. Related to energy density is the product of the state of charge of the i-th individual cell at the current moment and its terminal voltage. It refers to the state of charge of a single battery cell. It is the real-time terminal voltage of a single cell under the current load. , , These are the weighting coefficients, and their sum equals 1.

[0008] Preferably, the local energy buffer specifically includes: based on the parameter prediction curve of a key node, when the predicted parameter value of any key node is detected to reach or exceed the preventive action threshold, and the prediction trend indicates that it will reach the final safety threshold within a set critical time, a preventive micro-discharge command is generated for that node; according to the degree and risk level of the key node exceeding the action threshold, the target parameters for the preventive micro-discharge are calculated, and a discharge operation window is set; according to the target parameters, a controllable discharge is performed by adjusting the resistance value of the load element in the corresponding discharge branch until the discharge cutoff condition is reached; during and after the discharge, the actual measured value of the node's parameters is compared with the predicted value to verify the protection effect; when it is confirmed that the predicted parameter value has stabilized below the preventive action threshold and the rate of change meets the requirements, it is determined that the local energy buffer has been established and the discharge is terminated.

[0009] Preferably, the active reduction of its energy to form a local energy buffer further includes: S31 deploys a high-density array of miniature temperature sensors at sensitive nodes inside the battery pack to generate a real-time temperature field thermogram and calculate regional gradients, thereby identifying and marking high-risk temperature zones. S32, based on multi-source data, constructs an electro-thermal coupling model to predict the current concentration and temperature evolution trend in high-risk temperature zones, and dynamically adjusts the safety parameter thresholds for each zone accordingly. S33, based on the output of the electro-thermal coupling model, identifies specific high-risk individual cells or modules that require energy intervention and activates an independently controlled segmented equalization circuit; S34 The battery management unit continuously monitors the instantaneous temperature change rate of each area. If the temperature rise rate of any area exceeds the preset limit, the charging and discharging current of the relevant circuit or the entire battery pack will be forcibly limited.

[0010] Preferably, the high-risk temperature zone includes: a miniature temperature sensor array deployed in a grid-like layout at thermally sensitive locations inside the battery pack, forming a dense sensor network and connected to the battery management unit; the battery management unit synchronously reads the temperature measurements of all sensors at fixed intervals and generates a digital heat map of the continuous temperature distribution inside the battery pack using a spatial interpolation algorithm; based on the digital heat map, the battery management unit calculates the temperature gradient between specific key areas; the calculated temperature gradient values ​​of each area are compared in real time with a preset gradient safety threshold, and when the temperature gradient of a certain area continuously exceeds the threshold, the area is marked as a high-risk zone and its location information is recorded.

[0011] Preferably, the forced current limiting of the charging and discharging current of the relevant circuit or the entire battery pack includes: the battery management unit continuously acquiring temperature data from the micro temperature sensor array and calculating the instantaneous temperature change rate of each monitoring area; comparing the real-time temperature change rate of each area with a preset safety limit; when the temperature change rate of any area exceeds the safety limit, immediately generating and executing a protection command with the highest priority; the protection command sends a forced control signal to the main control power circuit to limit the charging and discharging current of the relevant affected circuit or the entire battery pack to below a preset safety value; after the current limiting protection action is executed, continuously monitoring the temperature change rate of the relevant area until it falls back to within the safety limit and stabilizes, and then gradually relaxing the current limit according to a preset recovery logic.

[0012] Preferably, the prediction of current concentration and temperature evolution trends in high-risk temperature zones specifically includes: S321: During the constant current charging phase of the battery pack, the terminal voltage of each individual cell is collected. By calculating the voltage difference between adjacent SOC intervals, the differential voltage curve of each individual cell is generated. The characteristic parameters of this curve are extracted and compared with the characteristic benchmark library established by historical normal battery data to identify abnormal shift patterns that characterize the degradation of electrochemical performance. S322: Based on the extracted differential voltage features, calculate the comprehensive health score for each individual cell, compare the health scores of all individual cells in the battery pack, calculate the standard deviation of each individual cell's score from the group average, define this value as the health deviation index of that individual cell, and combine real-time temperature data to correct the index to eliminate the interference of ambient temperature. S323 maps the health deviation index of each individual battery cell to the current adjustment coefficient of the corresponding charging channel, and instructs the charger with multi-channel independent output capability to perform differentiated current distribution.

[0013] Preferably, the current adjustment coefficient includes: determining the corresponding current adjustment coefficient according to a preset rule based on the temperature-corrected single-cell health deviation index; multiplying the current adjustment coefficient by a preset reference charging current to obtain the target charging current value of each single-cell; and verifying that each target charging current value does not exceed the maximum safe charging current limit allowed by the corresponding single-cell.

[0014] Preferably, the differentiated current allocation specifically includes: quantifying the electrical and thermal coupling strength between individual cells in the battery pack dynamic graph model; acquiring the differential voltage characteristic sequence and total current fluctuation sequence of each individual cell in real time, and quantitatively assessing the potential oscillation risk between individual cells, including the oscillation origin module, main propagation path, oscillation frequency, and estimated amplitude. When an oscillation mode with a frequency exceeding a set threshold is identified, the path is marked as a high-risk oscillation propagation path; deploying a distributed damping controller between individual cells, actively offsetting the detected oscillation energy by injecting reverse-phase compensation current and coordinating control; and globally redistributing the charging current based on the oscillation source tracing results, limiting the current of the oscillation source cell and compensating the affected modules.

[0015] One or more technical solutions provided in this application have at least the following technical effects or advantages: By employing dynamic battery network modeling and differential flatness theory prediction, global dynamic adaptation and proactive risk perception under unsteady conditions are achieved, fundamentally solving the problem of lag in traditional solutions. Combining feedforward control and preventative micro-discharge, it proactively establishes an energy buffer before thermal runaway occurs, effectively compensating for system delays. Simultaneously, the synergistic suppression of physical blocking and adaptive circuitry precisely cuts off remote energy transfer paths, significantly reducing the risk of cascade failure. A closed-loop optimization mechanism continuously improves decision-making accuracy, giving the system high robustness and long-term reliability, thereby significantly enhancing the intrinsic safety level of large-scale battery arrays.

[0016] By deploying a high-density temperature sensor array inside the battery, early and refined detection of precursors to thermal runaway is achieved. The electro-thermal coupling model built upon this model accurately predicts local current concentration and temperature rise trends, and dynamically sets region-specific safety thresholds accordingly, achieving a leap from comprehensive global protection to precise local early warning. Combined with a segmented equalization circuit, targeted energy fine-tuning can be performed on high-risk areas, and in conjunction with global protection measures, effectively suppressing the initiation and spread of thermal runaway. Real-time monitoring and instantaneous current limiting mechanisms provide millisecond-level rapid protection, while closed-loop learning continuously optimizes the prediction model, enabling the entire protection scheme to be adaptive and self-evolving, significantly improving the intrinsic safety and operational reliability of large-scale lithium battery arrays.

[0017] By analyzing the differential voltage curve of the battery online, characteristics reflecting the microscopic electrochemical states such as phase transitions of electrode materials are directly extracted, enabling early quantitative assessment of latent performance inconsistencies. Based on the health deviation index generated by this assessment, the system can perform differentiated dynamic allocation of charging current with milliampere precision, proactively balancing electrochemical stress at its source and significantly delaying battery pack degradation. This method shifts safety protection from macroscopic thermal management to microscopic electrochemical state management, effectively suppressing the risk of thermal runaway caused by uneven material degradation, and fundamentally improving the consistency, safety, and cycle life of the battery pack.

[0018] By introducing a graph neural network for global coupling analysis of the battery array, this method achieves, for the first time, real-time detection and precise localization of system-level oscillations caused by the transmission of implicit inconsistencies between modules. Based on this, distributed damping control actively injects counter-phase compensation current to precisely cancel out oscillations along their propagation path, fundamentally suppressing the chain reaction of risk. Simultaneously, the system can dynamically reconstruct the charging strategy based on oscillation source tracing results, maximizing charging efficiency while eliminating oscillations. This method effectively solves the problem of global instability caused by local faults in large-scale battery systems, achieving a leap from individual cell protection to system-level stability, and significantly improving the overall robustness and safety of the array. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart of a safety protection method for lithium batteries according to an embodiment of the present invention. Detailed Implementation

[0020] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Example 1: Figure 1 This is a flowchart illustrating a safety protection method for lithium batteries according to an embodiment of the present invention.

[0023] like Figure 1 As shown, a safety protection method for lithium batteries includes the following steps: S1 uses a high-frequency sensor array to collect multi-dimensional data and environmental data of each individual cell in the battery pack in real time, and uses graph theory to model the battery pack as a dynamic network topology graph, where nodes represent individual cells, edges represent energy transfer paths, and the weights of the edges are updated in real time according to the operating conditions, thereby constructing a battery network model that can reflect the dynamics of energy flow in real time.

[0024] The target parameters in the multi-dimensional data include the voltage, current, temperature, pressure, and gas concentration of each individual cell in the battery pack. Environmental data includes the load current change rate and the ambient temperature gradient.

[0025] S2, based on a dynamic battery network model, uses an electrothermal model and state observer constructed with differential flatness theory to predict the changing trends of key parameters of each individual battery cell, and combines graph theory algorithms to identify key nodes in the network and potential thermal runaway risk propagation paths.

[0026] Specifically, based on the electrochemical-thermal coupling principle of batteries, a state-space model with current as the control variable is constructed for each individual cell, and the individual cell voltage and temperature are selected as flat outputs. For this model, an extended Kalman filter is used as the state observer. This observer continuously receives real-time voltage, current, and temperature measurements from S1, and outputs the optimal estimate of the unmeasurable states such as the core temperature inside the individual cell, along with its error covariance, through recursive calculations of prediction and update. Using the current state estimate provided by the state observer as initial conditions, combined with known future load conditions, a multi-step forward integration calculation is performed on the future trajectory of the flat output within a fixed time window, thereby obtaining the predicted temperature and voltage change trend curves of each individual cell over a future period.

[0027] In the dynamic battery network model established in S1, the battery management unit calculates a comprehensive influence score for each node in the graph. This calculation first applies the PageRank algorithm to the current network topology to obtain the topological importance score of each node; then, this score is linearly weighted and fused with the node's temperature rise rate extracted from the prediction curve, and the product of the node's current state of charge and voltage. After completing the influence calculation for all nodes, the battery management unit sorts and filters the nodes according to a preset threshold, identifying nodes with influence scores higher than the threshold as key nodes.

[0028] Among them, node influence The calculation formula can be expressed as: It is a node importance score based on the current graph topology. It is the predicted rate of temperature rise. It is the future The amount of temperature change within a time window (usually the difference between the predicted temperature and the current estimated temperature). It is the length of the predicted future time interval. Related to energy density is the product of the state of charge of the i-th individual cell at the current moment and its terminal voltage. It refers to the state of charge of a single battery cell. It is the real-time terminal voltage of a single cell under the current load. , , These are the weighting coefficients, and their sum equals 1.

[0029] Finally, the battery management unit selects the cell predicted to be the first to reach the thermal runaway critical temperature from all nodes as the starting point for risk propagation. Centered on this source point, an improved Dijkstra graph search algorithm is used. Based on the time-varying edge weights representing the thermal / electrical coupling strength in the network and the edge failure probability calculated from the difference between the predicted temperature and the safe temperature of the source point, the most probable risk propagation path from the risk source point to other nodes in the network is searched and determined, thereby generating a complete risk propagation path topology graph.

[0030] S3. Based on the prediction and identification results of step S2, before the predicted value of a certain parameter in the multi-dimensional data exceeds the safety threshold, the battery management unit drives the self-discharge circuit to implement preventive micro-discharge on the key node, actively reducing its energy to form a local energy buffer.

[0031] Specifically, the battery management unit (BMU) acquires and monitors the parameter prediction curves of each key node generated in step S2 in real time and compares them with preset safety thresholds. To implement proactive intervention, a preventative action threshold lower than the final safety threshold is set. When the parameter prediction value of any key node reaches or exceeds this preventative action threshold, and its prediction trend indicates that the final safety threshold will be reached within a set critical time window, the BMU immediately determines that the triggering condition is met and generates a preventative micro-discharge command for that node. The triggering time is determined by the intersection of the prediction curve and the preventative action threshold. After the discharge is triggered, the BMU calculates the target current or target discharge amount of the micro-discharge based on the degree to which the node exceeds the action threshold and its risk level. Specifically, it calculates the target discharge current value required for the preventative micro-discharge and sets a maximum discharge duration based on the target energy reduction to limit the discharge operation window.

[0032] Subsequently, based on the calculated target discharge current, the battery management unit adjusts the resistance of the load element in the dedicated discharge branch connected to the critical node via a pulse-width modulation signal or a digital potentiometer. During this process, the setpoint of the load resistor is adjusted in real time according to the ratio of the node's real-time voltage to the target discharge current. This controlled discharge process continues until the preset maximum discharge duration is reached, or until real-time monitoring and reassessment confirm that the predicted risk of the node has been reduced to an acceptable level.

[0033] During and after the discharge operation, the battery management unit continuously compares the actual measured values ​​of the node's parameters with the updated predicted values. The effectiveness of the buffer formation is verified by evaluating the rate of decrease in the node's state of energy and the slope of the temperature prediction curve. When it is confirmed that the predicted value of the node's parameters has stabilized below the preventive action threshold and its rate of change tends to level off or become negative, the local energy buffer is determined to have been effectively established. The discharge command is then terminated, and the node's state is marked as protected, completing this preventive micro-discharge control cycle.

[0034] S4 embeds passive suppression elements between each individual cell in the battery pack to block electromagnetic energy transfer. At the same time, the battery management unit dynamically adjusts the parameters of the self-discharge circuit to compensate for unsteady-state losses based on real-time energy flow data.

[0035] Specifically, during the physical assembly stage of the battery pack, waveguide suppression materials such as ferrite cores are embedded in designated physical gaps between individual cells to form a distributed passive suppression layer. This suppression layer is used to absorb and attenuate high-frequency electromagnetic energy generated by high-frequency current pulsations or fault abrupt changes, thereby physically weakening the path of energy transmission through spatial electromagnetic coupling.

[0036] The battery management unit continuously monitors the transient current of each electrical branch through a distributed Hall sensor network, and calculates the energy transfer rate between adjacent individual cells or between specific circuits based on the measured current and voltage data, thereby quantifying the actual effect of electromagnetic suppression in real time. When the real-time energy transfer rate calculated by the battery management unit exceeds a preset percentage threshold, it is determined that physical suppression alone is insufficient, and a circuit compensation mechanism is triggered.

[0037] The formula for calculating the energy transfer rate is as follows: This refers to the unexpected energy received per unit time by a disturbed cell (or circuit) in a battery pack that is affected by energy crosstalk, which comes from other faulty or abnormal cells. This refers to the abnormal power output per unit time of a source cell (or circuit) in a battery pack that is considered a source of energy crosstalk.

[0038] Subsequently, the battery management unit calculates the required self-discharge parameter adjustment based on the current excessive energy transfer rate value and its changing trend, combined with the main energy flow path obtained from the dynamic battery network topology information from step S2. For high-risk individual cells identified as significantly affected by energy crosstalk, the target adjustment amount of their discharge current is determined based on the current reference discharge current of that individual cell, the extent to which the real-time energy transfer rate exceeds the threshold, and a predefined compensation gain coefficient.

[0039] By dynamically adjusting the resistance value of the equivalent load resistor in the dedicated discharge branch corresponding to the high-risk individual cell using a digital potentiometer or analog switch array, the discharge current flowing through that individual cell is precisely adjusted to a new target value. This parameter adjustment operation can be superimposed on the preventive micro-discharge control command triggered in step S3 to achieve composite regulation of the energy state of the target individual cell.

[0040] After parameter adjustments are completed, the battery management unit continues to monitor the energy transfer rate of the relevant electrical circuits and the core temperature change rate of the target individual cell. By verifying whether the energy transfer rate has fallen back to within the safe threshold and whether the trend of the real-time data and predicted curve of the individual cell temperature has returned to a flattening state, the effectiveness of the physical blocking and active discharge synergistic suppression strategy is confirmed, and the dynamic compensation cycle is completed accordingly.

[0041] S5 continuously monitors the overall status of the battery pack and compares the actual data with the predicted values. Based on the comparison error, it dynamically corrects the prediction model and the control parameters of the self-discharge circuit through optimization algorithms.

[0042] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By employing dynamic battery network modeling and differential flatness theory prediction, global dynamic adaptation and proactive risk perception under unsteady conditions are achieved, fundamentally solving the problem of lag in traditional solutions. Combining feedforward control and preventative micro-discharge, it proactively establishes an energy buffer before thermal runaway occurs, effectively compensating for system delays. Simultaneously, the synergistic suppression of physical blocking and adaptive circuitry precisely cuts off remote energy transfer paths, significantly reducing the risk of cascade failure. A closed-loop optimization mechanism continuously improves decision-making accuracy, giving the system high robustness and long-term reliability, thereby significantly enhancing the intrinsic safety level of large-scale battery arrays.

[0043] Example 2: In Example 1, by constructing a dynamic battery network model and combining it with feedforward control, preventative energy buffering of key nodes was achieved, effectively addressing the risk of cascading failure under unsteady conditions. However, this method mainly relies on macroscopic prediction and intervention based on external electrical parameters and network topology. Its sensing and control accuracy is limited for the highly uneven electro-thermal distribution at the microscale caused by manufacturing differences, uneven connection impedance, and inconsistent heat dissipation conditions within the battery pack. When microscopic anomalies such as localized current concentration and hotspot emergence occur within the battery pack, the changes in macroscopic external parameters (such as total voltage and average temperature) often exhibit lag and smoothness, making it difficult for a single macroscopic network model to accurately identify and locate these microscopic risk sources in the early stages, and even more difficult to dynamically and finely set the resulting differentiated safety thresholds. To achieve early and accurate perception and control of the microscopic electro-thermal state inside the battery and improve the ability to suppress the initiation of local thermal runaway, it is necessary to further deepen and refine the safety protection strategy. On the basis of macroscopic network protection, a microscopic state perception and targeted control mechanism should be introduced and integrated.

[0044] In some embodiments, the energy is actively reduced to form a local energy buffer, and step S3 further includes: S31 deploys a high-density array of miniature temperature sensors at sensitive nodes inside the battery pack to generate a real-time temperature field thermogram and calculate regional gradients, thereby identifying and marking high-risk temperature zones.

[0045] Specifically, during battery pack assembly, a high-precision miniature temperature sensor array is deployed in a grid-like layout at thermally sensitive locations such as the tabs, geometric corners, and central area inside the battery pack. The sensors are either thermocouples or fiber optic sensors, covering key points on the surface of each individual cell and the predicted internal hotspot areas, forming a dense sensing network. This network connects to the data acquisition module of the battery management unit, ensuring that the sampling clocks of all sensors are synchronized, and that the sampling frequency is uniformly set to no less than 10Hz to capture transient temperature changes.

[0046] Once the sensor array begins operation, the battery management unit (BMU) synchronously reads the temperature measurements from all sensors at fixed intervals (e.g., 100 milliseconds). Using these spatially discrete measurement points, the BMU generates a digital heatmap reflecting the continuous temperature distribution within the entire battery pack through a spatial interpolation algorithm. This heatmap visualizes the real-time temperature at various locations within the battery pack in two-dimensional or three-dimensional form.

[0047] Based on the generated temperature field thermogram, the battery management unit calculates the temperature gradient between specific key regions. For example, it calculates the average temperature difference between the tab region of each individual cell and the center region of the same individual cell. This calculation is performed by extracting the temperature values ​​at the corresponding coordinate positions from the thermogram and then performing an arithmetic mean and difference operation.

[0048] The battery management unit (BMU) compares the calculated temperature gradient values ​​for each region with a preset gradient safety threshold (e.g., 5°C) in real time. When the calculated temperature gradient for a region consistently exceeds this threshold, the BMU immediately highlights that region on the digital heatmap and records its location coordinates in the battery pack topology, the cell number it belongs to, and the out-of-limit data as a high-risk event. This marking and recording information will serve as direct input for subsequent refined modeling and risk prediction.

[0049] S32 constructs an electro-thermal coupling model based on multi-source data to predict the current concentration and temperature evolution trend in high-risk temperature zones, and dynamically adjusts the safety parameter thresholds for each zone accordingly.

[0050] Specifically, the battery management unit (BMU) first gathers multi-source data required to construct the electro-thermal coupling model. This data includes a refined temperature field thermogram and marked high-risk area information generated in step S31, the bus current and voltage of the battery pack, the real-time voltage and estimated state of charge of each individual cell, and ambient temperature parameters. Subsequently, the BMU inputs this data into a pre-defined coupled numerical model that integrates battery internal resistance variation characteristics, electrochemical heat generation formulas, and material thermophysical parameters (such as thermal conductivity and specific heat capacity). This model is typically based on the finite element method or finite volume method and is used to describe the interaction between charge transport and heat diffusion.

[0051] After the model runs, the battery management unit uses it for calculations. Starting from the current state, and combining it with the load condition curves for the next short time period, it performs a high-resolution spatiotemporal evolution simulation of the battery pack's internal structure. The simulation process solves the coupled equations of charge conservation and energy conservation, and outputs predicted current density distribution and temperature evolution curves for various spatial locations over a future period (e.g., the next 30 seconds), particularly for the high-risk area marked S31. Through this simulation, it is possible to quantify and predict effects such as current concentration in the tab region due to connection resistance, and the resulting local temperature rise rate.

[0052] Based on the simulation results, the battery management unit (BMU) performs dynamic safety strategy adjustments. For specific areas predicted to experience significant temperature rises or abnormal current concentrations, the BMU no longer applies a globally uniform safety threshold, but instead dynamically sets stricter local safety parameter limits. For example, for a tab area where the predicted temperature will rise by 15°C within 10 seconds, the BMU can dynamically lower its safe upper limit for state of charge (SOC) from the global 100% to 85%, or reduce its maximum allowable instantaneous discharge rate from 3C to 1.5C. These adjusted area-specific thresholds are updated in real-time to the protection strategy library, serving as an immediate basis for subsequent control and judgment.

[0053] S33 identifies specific high-risk individual cells or modules that require energy intervention based on the output of the electro-thermal coupling model, and activates an independently controlled segmented equalization circuit.

[0054] Specifically, the prediction results output by the electro-thermal coupling model in S32 are analyzed to identify specific individual cells or modules with excessively high current concentration or excessively rapid temperature rise rates, determining them as targets requiring energy intervention. For the identified targets, the battery management unit activates the corresponding independent segmented balancing circuit. For example, for individual cells or regions predicted to overheat significantly due to current concentration, the balancing circuit applies a small reverse current (discharge rate, e.g., 0.2C-0.5C) to shunt the current, reducing local charge accumulation and heat generation; for regions predicted to have lower temperatures due to uneven heat dissipation, a small positive current (charge rate, e.g., 0.1C) is added to prevent undercharging and balance the overall state.

[0055] S34 The battery management unit continuously monitors the instantaneous temperature change rate of each area. If the temperature rise rate of any area exceeds the preset limit, the charging and discharging current of the relevant circuit or the entire battery pack will be forcibly limited.

[0056] Specifically, the battery management unit continuously acquires raw temperature data from each monitoring point at a sampling frequency of no less than 10Hz based on the miniature temperature sensor array deployed in step S31. Subsequently, the time-series temperature data of each independent monitoring area (the tab or center point of a single cell) is processed in real time. By calculating the temperature difference between the current moment and the previous sampling moment (5 seconds ago) and dividing it by the time interval, the instantaneous temperature change rate of that area, i.e., the heating rate, is obtained.

[0057] The battery management unit (BMU) continuously compares the calculated real-time temperature rise rate of each region with a preset safety limit (2°C / s). When the real-time temperature rise rate of any region exceeds this preset limit, the BMU immediately determines that the region has entered an emergency overheating state and generates a highest-priority protection command. This protection command is executed immediately. The BMU sends a forced control signal to the main control power circuit of the battery pack. This signal directly acts on the power devices (MOSFETs or IGBTs) responsible for charge / discharge management. Through hardware and software linkage, within milliseconds, the total output / input current of the affected circuit (the module containing the high-risk individual battery cell) or the entire battery pack is forcibly limited below a preset safety value. For example, the maximum allowable charging or discharging current is immediately reduced from the current 3C to 1C or lower.

[0058] After the current limiting protection action is triggered and executed, the battery management unit continues to monitor the temperature and temperature rise rate of the relevant area. Only after confirming that the temperature rise rate of the area has continuously fallen back to within the safe limit and stabilized for a certain period of time will the battery management unit gradually and conditionally relax the current limit according to the preset recovery logic.

[0059] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By deploying a high-density temperature sensor array inside the battery, early and refined detection of precursors to thermal runaway is achieved. The electro-thermal coupling model built upon this model accurately predicts local current concentration and temperature rise trends, and dynamically sets region-specific safety thresholds accordingly, achieving a leap from comprehensive global protection to precise local early warning. Combined with a segmented equalization circuit, targeted energy fine-tuning can be performed on high-risk areas, and in conjunction with global protection measures, effectively suppressing the initiation and spread of thermal runaway. Real-time monitoring and instantaneous current limiting mechanisms provide millisecond-level rapid protection, while closed-loop learning continuously optimizes the prediction model, enabling the entire protection scheme to be adaptive and self-evolving, significantly improving the intrinsic safety and operational reliability of large-scale lithium battery arrays.

[0060] Example 3: In Example 2, by deploying a high-density temperature sensor array inside the battery pack and constructing an electro-thermal coupling model, refined perception and management of microscopic thermal risks and current distribution were achieved, significantly improving the early warning and suppression capabilities for thermal runaway. However, this method is essentially still a response to and regulation of physical field anomalies (temperature, current) that have already appeared or are about to appear. For the root cause of lithium battery performance degradation and runaway risk—namely, the implicit inconsistencies in the internal electrochemical state caused by differences in materials and processes (such as loss of active materials and uneven coating thickness)—Example 2 mainly relies on the resulting thermal and electrical results for indirect inference, which has a recognition lag and makes it difficult to quantify its intrinsic degree. In order to identify, quantify, and proactively manage this implicit inconsistency at the electrochemical source level earlier and more directly before physical signals such as thermal anomalies appear, thereby delaying battery pack composition at the source and preventing the deep safety risks caused by it, it is necessary to introduce and integrate a new dimension of state monitoring and regulation based on the intrinsic electrochemical response of the battery on top of the physical field monitoring layer of Example 2.

[0061] In some embodiments, step S32, predicting the current concentration and temperature evolution trend in the high-risk temperature zone, further includes: S321: During the constant current charging phase of the battery pack, the terminal voltage of each individual cell is collected. By calculating the voltage difference between adjacent SOC intervals, a differential voltage curve for each individual cell is generated. The characteristic parameters of this curve are extracted, and the characteristics are compared with the characteristic benchmark library established by historical normal battery data to identify abnormal shift patterns that characterize the degradation of electrochemical performance.

[0062] The characteristic parameters of the differential voltage curve include peak voltage, the SOC position corresponding to the peak value, and the change in the slope of the curve.

[0063] The establishment of the feature benchmark library based on historical normal battery data involves selecting a batch of brand-new, qualified standard single-cell batteries from the production line as a sample set. Under a controlled standard laboratory environment, these sample batteries undergo a specified number of standard charge-discharge cycle tests, with high-precision synchronous acquisition of voltage data for each single-cell battery across the entire SOC range during each constant-current charging phase. Subsequently, based on the acquired voltage-SOC data, the differential voltage curve for each sample battery in each cycle is calculated, and several predefined key feature parameters are extracted from each curve. Next, statistical analysis is performed on the same type of feature parameters extracted from all sample batteries across all cycles, calculating their group mean and standard deviation to determine the benchmark value and normal fluctuation range of each feature parameter. Then, another independent set of normal battery samples is used to validate the statistical model to ensure that the determined feature range effectively covers the variations in normal batteries. After successful validation, the benchmark values ​​and ranges of all feature parameters are integrated into a structured feature database, which is ultimately solidified into a software model and integrated into the storage system of the battery management unit, forming a benchmark library for online comparison.

[0064] S322 calculates a comprehensive health score for each individual cell based on the extracted differential voltage features, compares the health scores of all individual cells in the battery pack, calculates the standard deviation of each individual cell's score from the group average, defines this value as the health deviation index of that individual cell, and corrects the index by combining real-time temperature data to eliminate environmental temperature interference.

[0065] Specifically, based on the differential voltage feature parameters extracted in step S321, a comprehensive health score is calculated for each individual cell. This score is obtained by standardizing each feature parameter (such as peak voltage and peak-corresponding SOC) and then linearly combining it with a weight vector trained using a large amount of normal battery data. The calculation formula is as follows: ,in The health score of the i-th individual cell. For the weight vector, This is a vector composed of the characteristic parameters of the individual battery cell. The battery management unit then calculates the average health score of all individual cells in the entire battery pack. and standard deviation Based on this, the degree of deviation of each individual battery score from the group average is calculated, i.e., the health deviation index. It is defined as the difference between the individual cell score and the group average score divided by the group standard deviation, and the formula is: This index quantifies the degree of implicit inconsistency of a single cell relative to the population average. The larger the value, the further it deviates from the normal state. Finally, to eliminate the interference of ambient temperature on the above electrical characteristic assessment, a temperature compensation coefficient κ(T) is introduced to correct the health deviation index. This coefficient, obtained through previous experimental calibration, is a function of temperature T. The final health deviation index after temperature correction is shown below. From the formula The calculations yielded an assessment index reflecting the health status of the battery itself.

[0066] S323 maps the health deviation index of each individual battery cell to the current adjustment coefficient of the corresponding charging channel, and instructs the charger with multi-channel independent output capability to perform differentiated current distribution.

[0067] The current adjustment coefficient is a health deviation index for each individual cell, calculated in step S322 and temperature-corrected, obtained by the battery management unit. Based on a preset mapping rule, this index is converted into the current adjustment coefficient for the corresponding charging channel. The core logic of this rule is that when the index is negative, the coefficient is less than one to reduce the charging current; when the index is positive, the coefficient is equal to or slightly greater than one to allow a moderate increase in current, while the adjustments of all channels must collectively meet the total power constraint. A common implementation method is to use a linear mapping relationship, controlling the adjustment sensitivity through a preset gain coefficient. Based on the calculated current adjustment coefficients for each channel, the battery management unit multiplies them by the system's preset reference charging current to determine the target charging current value for each individual cell. Afterward, the battery management unit verifies all target current values ​​one by one to ensure that they do not exceed the maximum safe charging current limit allowed by each individual cell.

[0068] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By analyzing the differential voltage curve of the battery online, characteristics reflecting the microscopic electrochemical states such as phase transitions of electrode materials are directly extracted, enabling early quantitative assessment of latent performance inconsistencies. Based on the health deviation index generated by this assessment, the system can perform differentiated dynamic allocation of charging current with milliampere precision, proactively balancing electrochemical stress at its source and significantly delaying battery pack degradation. This method shifts safety protection from macroscopic thermal management to microscopic electrochemical state management, effectively suppressing the risk of thermal runaway caused by uneven material degradation, and fundamentally improving the consistency, safety, and cycle life of the battery pack.

[0069] Example 4: In Example 3, differential voltage analysis and dynamic current allocation were used to accurately quantify and compensate for implicit inconsistencies within individual cells and modules, effectively delaying performance differentiation at the microscale. However, this approach primarily focuses on the state perception and adjustment of individual cells, failing to fully consider the complex dynamic network interactions formed by electrical connections and thermal coupling between multiple modules in a large-scale battery array. This energy and state transfer between cells may cause implicit inconsistencies in a module to propagate through the network path, triggering system-level current or thermal oscillations (such as periodic power fluctuations between modules), leading to global energy distribution imbalances or even cascading instability risks. To address the global dynamic instability issues that may arise from local optimization, it is necessary to model, identify, and actively suppress such mutual coupling and oscillation propagation at the system network level, thereby achieving a leap from individual cell compensation to system-wide collaborative stability.

[0070] In some embodiments, performing differentiated current allocation in step S323 further includes: 3A, quantifies the electrical and thermal coupling strength between individual cells in the dynamic diagram model of the battery pack.

[0071] 3B: Real-time acquisition of differential voltage characteristic sequences and total current fluctuation sequences for each individual cell, enabling quantitative assessment of potential oscillation risks between individual cells, including the oscillation origin module, main propagation path, oscillation frequency, and estimated amplitude. When an oscillation mode with a frequency exceeding a set threshold is identified, the path is marked as a high-risk oscillation propagation path.

[0072] 3C, deploying distributed damping controllers among individual cells, actively counteracts detected oscillation energy by injecting reverse-phase compensation current and coordinating control.

[0073] Specifically, each individual cell is equipped with a local damping controller. The battery management unit sends the oscillation characteristics (such as frequency and phase) identified by the 3A graph neural network to the local damping controller of the relevant individual cell. Based on the received global oscillation mode information, each controller independently calculates and generates a compensation current pulse signal that is out of phase with its detected local current oscillation waveform, and injects this signal through the power converter at the module port to actively cancel the oscillation energy transmitted through the electrical connection. At the same time, each damping controller exchanges status information through a low-latency communication network and coordinates to adjust the amplitude and phase of the compensation signal to prevent adjacent controllers from generating new secondary oscillations due to overcompensation, thereby optimizing the global damping effect.

[0074] 3D, based on the oscillation source tracing results, globally redistributes the charging current, limits the current of the oscillation source cell battery and compensates the affected modules.

[0075] Specifically, current limits are applied to the identified oscillation source cells (e.g., reducing their charging current rate from fast charging to 0.5C) to reduce disturbance output at the source; compensatory charging current fine-tuning is performed on downstream cells significantly affected by oscillations to correct energy imbalances caused by oscillations; and for regions that have achieved effective damping control and returned to stability, their optimal charging current is gradually restored.

[0076] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By introducing a graph neural network for global coupling analysis of the battery array, this method achieves, for the first time, real-time detection and precise localization of system-level oscillations caused by the transmission of implicit inconsistencies between modules. Based on this, distributed damping control actively injects counter-phase compensation current to precisely cancel out oscillations along their propagation path, fundamentally suppressing the chain reaction of risk. Simultaneously, the system can dynamically reconstruct the charging strategy based on oscillation source tracing results, maximizing charging efficiency while eliminating oscillations. This method effectively solves the problem of global instability caused by local faults in large-scale battery systems, achieving a leap from individual cell protection to system-level stability, and significantly improving the overall robustness and safety of the array.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A safety protection method for lithium batteries, characterized in that, include: S1 uses a high-frequency sensor array to collect multi-dimensional data and environmental data of each individual cell in the battery pack in real time, and uses graph theory to model the battery pack as a dynamic network topology graph, where nodes represent individual cells, edges represent energy transfer paths, and the weights of the edges are updated in real time according to the operating conditions, thereby constructing a battery network model that can reflect the dynamics of energy flow in real time; the target parameters in the multi-dimensional data include the voltage, current, temperature, pressure, and gas concentration of each individual cell in the battery pack; S2, based on a dynamic battery network model, uses an electrothermal model and state observer constructed with differential flatness theory to predict the changing trends of key parameters of each individual battery cell, and combines graph theory algorithms to identify key nodes in the network and potential thermal runaway risk propagation paths. S3, based on the prediction and identification results of step S2, before the predicted value of a certain parameter in the multi-dimensional data exceeds the safety threshold, the battery management unit drives the self-discharge circuit to implement preventive micro-discharge on the key node, actively reducing its energy to form a local energy buffer. S4, passive suppression elements are embedded between each individual cell in the battery pack to block the transfer of electromagnetic energy. At the same time, the battery management unit dynamically adjusts the parameters of the self-discharge circuit to compensate for unsteady losses based on real-time energy flow data. S5 continuously monitors the overall status of the battery pack and compares the actual data with the predicted values. Based on the comparison error, it dynamically corrects the prediction model and the control parameters of the self-discharge circuit through optimization algorithms.

2. The safety protection method for lithium batteries as described in claim 1, characterized in that, The identification of key nodes and potential thermal runaway risk propagation paths in the network specifically includes: constructing an electrothermal state-space model using differential flatness theory, and using a state observer to estimate the internal state of the battery and predict parameters based on real-time data; in the dynamic battery network model, calculating the comprehensive node influence degree for each node by integrating its topological importance, predicted temperature rise trend, and real-time energy state indicators, thereby identifying key nodes; taking the cell that is predicted to reach the thermal runaway temperature earliest as the risk source, and determining the most likely risk propagation path based on the dynamic coupling strength and failure probability in the network.

3. The safety protection method for lithium batteries as described in claim 2, characterized in that, The node influence degree is calculated using the following formula: , It is a node importance score based on the current graph topology. It is the predicted rate of temperature rise. It is the future The amount of temperature change within the time window, It is the length of the predicted future time interval. Related to energy density is the product of the state of charge of the i-th individual cell at the current moment and its terminal voltage. It refers to the state of charge of a single battery cell. It is the real-time terminal voltage of a single cell under the current load. , , These are the weighting coefficients, and their sum equals 1.

4. The safety protection method for lithium batteries as described in claim 1, characterized in that, The local energy buffer specifically includes: based on the parameter prediction curve of key nodes, when the predicted parameter value of any key node reaches or exceeds the preventive action threshold, and the prediction trend indicates that it will reach the final safety threshold within a set critical time, a preventive micro-discharge command is generated for that node; according to the degree and risk level of the key node exceeding the action threshold, the target parameters of the preventive micro-discharge are calculated, and a discharge operation window is set; according to the target parameters, a controllable discharge is performed by adjusting the resistance value of the load element in the corresponding discharge branch until the discharge cutoff condition is reached; during and after the discharge, the actual measured value of the node's parameters is compared with the predicted value to verify the protection effect; when it is confirmed that the predicted parameter value has stabilized below the preventive action threshold and the rate of change meets the requirements, it is determined that the local energy buffer has been established and the discharge is terminated.

5. The safety protection method for lithium batteries as described in claim 1, characterized in that, The active reduction of its energy to form a local energy buffer also includes: S31 deploys a high-density array of miniature temperature sensors at sensitive nodes inside the battery pack to generate a real-time temperature field thermogram and calculate regional gradients, thereby identifying and marking high-risk temperature zones. S32, based on multi-source data, constructs an electro-thermal coupling model to predict the current concentration and temperature evolution trend in high-risk temperature zones, and dynamically adjusts the safety parameter thresholds for each zone accordingly. S33, based on the output of the electro-thermal coupling model, identifies specific high-risk individual cells or modules that require energy intervention and activates an independently controlled segmented equalization circuit; S34 The battery management unit continuously monitors the instantaneous temperature change rate of each area. If the temperature rise rate of any area exceeds the preset limit, the charging and discharging current of the relevant circuit or the entire battery pack will be forcibly limited.

6. The safety protection method for lithium batteries as described in claim 5, characterized in that, The high-risk temperature zone includes: a grid-like array of miniature temperature sensors deployed at thermally sensitive locations inside the battery pack, forming a dense sensor network connected to the battery management unit; the battery management unit synchronously reads the temperature measurements from all sensors at fixed intervals and generates a digital heat map of the continuous temperature distribution inside the battery pack using a spatial interpolation algorithm; based on the digital heat map, the battery management unit calculates the temperature gradient between specific key areas; the calculated temperature gradient values ​​of each area are compared in real time with a preset gradient safety threshold, and when the temperature gradient of a certain area continuously exceeds the threshold, the area is marked as a high-risk zone and its location information is recorded.

7. The safety protection method for lithium batteries as described in claim 5, characterized in that, The forced current limiting of the charging and discharging current of the relevant circuit or the entire battery pack includes: the battery management unit continuously acquiring temperature data from the micro temperature sensor array and calculating the instantaneous temperature change rate of each monitoring area; comparing the real-time temperature change rate of each area with a preset safety limit; when the temperature change rate of any area exceeds the safety limit, immediately generating and executing the highest priority protection command; the protection command sends a forced control signal to the main control power circuit to limit the charging and discharging current of the relevant affected circuit or the entire battery pack to below the preset safety value; after the current limiting protection action is executed, continuously monitoring the temperature change rate of the relevant area until it falls back to within the safety limit and stabilizes, and then gradually relaxing the current limit according to the preset recovery logic.

8. The safety protection method for lithium batteries as described in claim 5, characterized in that, The predicted current concentration and temperature evolution trend in high-risk temperature zones are specifically as follows: S321: During the constant current charging phase of the battery pack, the terminal voltage of each individual cell is collected. By calculating the voltage difference between adjacent SOC intervals, the differential voltage curve of each individual cell is generated. The characteristic parameters of this curve are extracted and compared with the characteristic benchmark library established by historical normal battery data to identify abnormal shift patterns that characterize the degradation of electrochemical performance. S322: Based on the extracted differential voltage features, calculate the comprehensive health score for each individual cell, compare the health scores of all individual cells in the battery pack, calculate the standard deviation of each individual cell's score from the group average, define this value as the health deviation index of that individual cell, and combine real-time temperature data to correct the index to eliminate the interference of ambient temperature. S323 maps the health deviation index of each individual battery cell to the current adjustment coefficient of the corresponding charging channel, and instructs the charger with multi-channel independent output capability to perform differentiated current distribution.

9. The safety protection method for lithium batteries as described in claim 7, characterized in that, The current adjustment coefficient includes: determining the corresponding current adjustment coefficient according to a preset rule based on the temperature-corrected single-cell health deviation index; multiplying the current adjustment coefficient by a preset reference charging current to obtain the target charging current value of each single-cell; and verifying that each target charging current value does not exceed the maximum safe charging current limit allowed for the corresponding single-cell.

10. The safety protection method for lithium batteries as described in claim 7, characterized in that, The differentiated current allocation specifically includes: quantifying the electrical and thermal coupling strength between individual cells in the battery pack dynamic graph model; acquiring the differential voltage characteristic sequence and total current fluctuation sequence of each individual cell in real time, and quantitatively assessing the potential oscillation risk between individual cells, including the oscillation origin module, main propagation path, oscillation frequency, and estimated amplitude. When an oscillation mode with a frequency exceeding a set threshold is identified, the path is marked as a high-risk oscillation propagation path; deploying a distributed damping controller between individual cells, actively offsetting the detected oscillation energy by injecting reverse-phase compensation current and coordinating control; and globally redistributing the charging current based on the oscillation source tracing results, limiting the current of the oscillation source cell and compensating the affected modules.

Images