A Battery Safety Early Warning Method Based on Multi-Signal Feature Fusion

Abstract

This invention discloses a battery safety early warning method based on multi-signal feature fusion, relating to the field of lithium-ion battery thermal runaway technology. This invention simultaneously focuses on multiple characteristic parameter signals to obtain more accurate early warnings. The established electrothermal model is applicable to different temperature environments and comprehensively considers multiple parameters such as temperature, current, voltage, and internal impedance, resulting in more accurate state estimation and rapid prediction of battery state. Addressing the safety risks posed by battery temperature, temperature is introduced as a constraint in the prediction process. A method using temperature-compensated resistance is employed to estimate the State of Charge (SOC), while a joint SOC and State of Health (SOH) estimation method is used to improve the accuracy of battery health estimation and further enhance the accuracy of safety early warnings. Finally, a multi-head attention mechanism is used to process parallel parameter data in parallel, more effectively handling multiple characteristic parameters and improving the depth and accuracy of data analysis.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery thermal runaway technology, specifically to a battery safety early warning method that integrates multiple signal features. Background Technology

[0002] With the widespread application of lithium-ion batteries in electric vehicles, energy storage systems, and consumer electronics, the safety of batteries as key energy storage components is a prerequisite and foundation for large-scale application. Lithium-ion batteries, with their high energy density, long cycle life, and high charge / discharge efficiency, have found widespread use in modern society. However, as high-energy-density components, the risk of thermal runaway can lead to significant losses.

[0003] Thermal runaway is one of the root causes of battery safety accidents. Especially when batteries are overcharged, over-discharged, subjected to excessively high external temperatures, or suffer internal malfunctions, the internal chemical reactions are accelerated, leading to a rapid increase in temperature and ultimately thermal runaway. The final stage of thermal runaway occurs rapidly and quickly, and can even trigger fires or explosions, threatening public safety. Whether in electric vehicles or energy storage power stations, the catastrophic consequences of battery thermal runaway far exceed those of conventional industrial accidents. Therefore, how to achieve early warning of battery thermal runaway is a key issue for breakthroughs in more fields.

[0004] Currently, traditional battery safety monitoring technologies mostly rely on single signal characteristics (such as battery voltage and temperature) for real-time monitoring. However, a single signal characteristic cannot comprehensively reflect the overall health status of the battery and its potential safety risks, especially under complex conditions such as battery aging, changes in charging and discharging conditions, and fluctuations in ambient temperature. Determining the risk of thermal runaway based on a single signal may be delayed or result in misjudgment. Battery health depends not only on instantaneous voltage and temperature changes but is also influenced by multiple factors, such as the internal chemical aging process, charging speed, and depth of discharge. As battery usage time increases, its performance gradually degrades, thus requiring more precise early warning systems to identify safety risks under different conditions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a battery safety early warning method based on the fusion of multiple signal features, thereby solving the problems of lag and inaccuracy in existing battery safety monitoring technologies that rely on a single signal feature for monitoring.

[0006] A battery safety early warning method based on multi-signal feature fusion includes the following steps:

[0007] A second-order RC equivalent circuit model of a single cell, a thermal model of a single cell, a heat transfer model of a single cell, a heat transfer model between single cells in a battery module, a thermal balance model of the battery module, and a usable capacity model of the battery module are constructed; the battery module includes several single cells.

[0008] Based on the second-order RC equivalent circuit model of a single cell, parameter identification is performed using hybrid pulse power characteristic testing to obtain the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the single cell.

[0009] Establish a joint estimation model for battery health and state of charge;

[0010] Establish safety constraints for battery module operation and state-space expressions for multi-state joint estimation of battery module;

[0011] Under the constraints of the battery module's thermal balance model, battery module's available capacity model, battery module's operational safety constraints, and the state-space expression of the battery module's multi-state joint estimation, based on the established individual cell thermal model, individual cell heat transfer model, and heat transfer model between individual cells in the battery module, the ohmic internal resistance, polarization internal resistance, polarization capacitance of the individual cell, and the established joint estimation model of battery health and state of charge are obtained. The state prediction data of the individual cell is obtained using the AEKF method.

[0012] Based on the state prediction data of individual cells, calculate the changes in state of charge, voltage drop, current change, and temperature rise;

[0013] The state prediction data of a single cell, changes in state of charge, voltage drop, current change, and temperature rise are input into a pre-trained transform model to obtain a battery thermal runaway safety warning result, which is a risk factor.

[0014] Furthermore, the second-order RC equivalent circuit model of the single cell is as follows:

[0015] U0(t)=U OCV (t)-R0I(t)-U1(t)-U2(t) (1)

[0016] In the formula, U0(t) represents the terminal voltage of a single cell, U OCV U(t) represents the open-circuit voltage of a single cell, U1(t) and U2(t) represent the polarization voltages of the two RC circuits of the single cell, R0 represents the ohmic internal resistance of the single cell, I(t) is the charging and discharging current of the single cell, and t represents the time.

[0017] The thermal model of the single cell is as follows:

[0018]

[0019] In the formula, Q(t) is the heat generated by a single cell, T(t) and T are the average temperatures of the single cells in the battery module, and E is the average temperature of the cell. h U is the temperature entropy coefficient. OCV Represents the open-circuit voltage of a single cell;

[0020] The heat transfer model of the single cell is as follows:

[0021]

[0022] In the formula, Q in (t) represents the temperature at which the core of a single cell loses heat to the surface. c (t) represents the core temperature of a single cell, g represents the thermal conductivity of the material, x represents the distance from the core to the surface of the single cell, and Q represents the thermal conductivity of the material. conv (t) represents the heat loss from the surface of a single cell, h1 is the convective heat transfer coefficient, and T s (t) represents the surface temperature of a single cell, T e (t) represents the ambient temperature, and A1 represents the surface area of ​​a single battery cell;

[0023] The heat transfer model between individual cells in the battery module is as follows:

[0024]

[0025] In the formula, Q b (t) represents the heat transferred between individual cells in the battery module, where i represents the cell number, T s,i (t) represents the surface temperature of the i-th individual cell in the battery module, h2 is the heat transfer coefficient, A2 is the cross-sectional area for heat conduction, L is the length or thickness of the heat conduction path, and T s,i+1 (t) and T s,i-1 (t) represents the surface temperature of the two cells adjacent to the i-th cell;

[0026] The thermal balance model of the battery module is as follows:

[0027]

[0028] In the formula, B c For core heat capacity, B s For surface heat capacity, T c,i (t) represents the core temperature of the i-th individual cell;

[0029] The capacity model of the battery module is as follows:

[0030]

[0031] In the formula, V tol V represents the available capacity of the battery module. charV represents the available charging capacity of the battery module. disc V represents the available discharge capacity of the battery module; N represents the total number of individual cells; V m The capacity of the single battery cell with the smallest available rechargeable capacity; V h S represents the capacity of the single cell with the smallest usable discharge capacity. OC,h S represents the state of charge (S) of the single cell h with the smallest usable discharge capacity. OC,m The state of charge (SOC) of the single cell m with the smallest available rechargeable capacity.

[0032] Furthermore, the second-order RC equivalent circuit model based on a single cell is used for parameter identification through hybrid pulse power characteristic testing to obtain the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the single cell, specifically:

[0033] A1: Discretize the state of charge (SOC) of a single cell and obtain the open-circuit voltage of the single cell under different SOCs to obtain the open-circuit voltage U of the single cell under different SOCs. OCV The fitted curve;

[0034] The method for discretizing the state of charge of a single cell is as follows:

[0035]

[0036] In the formula, S OC (t) represents the state of charge of a single cell at time t, S OC (t+1) represents the state of charge of a single cell at time t+1, η is the coulombic efficiency of the cell, Δt is the time step between continuous state changes, and V c This represents the current actual capacity of a single battery cell.

[0037] A2: Measure the terminal voltage and charging / discharging current of a single cell.

[0038] A3: Based on the second-order RC equivalent circuit model of a single cell, the observation equation of the single cell is established. Then, based on the observation equation of the single cell, the ohmic internal resistance, polarization internal resistance and polarization capacitance of the single cell are identified by testing the hybrid pulse power characteristics.

[0039] Specifically, the open-circuit voltage U of the individual cells under different states of charge is used. OCV By fitting the curve, the open-circuit voltage of the individual cell at the current moment is determined. Substituting the measured terminal voltage, charging / discharging current, and open-circuit voltage of the individual cell into the observation equation, the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the individual cell are identified. The observation equation is as follows:

[0040]

[0041] In the formula, R1 and R2 are the polarization internal resistances of a single cell, and their parallel connection represents electrochemical polarization; C1 and C2 are the polarization capacitances of a single cell, and their parallel connection represents concentration polarization; J is the minimum sum of squared errors. U represents the actual terminal voltage of a single cell at time t. OCV (S OC (t) represents the open-circuit voltage of a single cell at time t. max This represents the maximum number of discrete time steps.

[0042] Furthermore, the joint estimation model for battery health and state of charge is as follows:

[0043]

[0044] In the formula, S OH (t) represents the battery health at time t; V N This refers to the rated maximum capacity of a single battery cell.

[0045] Furthermore, the safety constraints for the operation of the battery module include the current actual capacity constraint of the individual battery, the temperature constraint of the individual battery, and the state of charge constraint of the individual battery; the temperature includes the core temperature of the individual battery, the surface temperature of the individual battery, and the average temperature of the individual batteries in the battery module.

[0046] The safety constraints for the operation of the battery module are as follows:

[0047]

[0048] In the formula, T c，max (t) represents the maximum battery core temperature, T cs，max (t) represents the maximum surface and core temperature difference, T max (t) represents the maximum average temperature of the battery. To prevent overcharging and over-discharging, S is set. OC，min S represents the minimum state of charge of a single cell. OC，max This represents the maximum state of charge (SOC) of a single cell.

[0049] The state-space expression for the multi-state joint estimation of the battery module is as follows:

[0050]

[0051] In the formula, z is the mass of a single cell, and a k The weighting coefficient represents the influence of charging and discharging current at different historical moments on the current terminal voltage, where β represents the past time. max This represents the total time duration.

[0052] Furthermore, the state prediction data of the individual battery includes the terminal voltage of the individual battery, the charge and discharge current of the individual battery, the temperature, the state of charge of the individual battery, and the battery health of the individual battery.

[0053] Furthermore, under the constraints of the battery module's thermal balance model, usable capacity model, operational safety constraints, and state-space expression for multi-state joint estimation, based on the established individual cell thermal model, individual cell heat transfer model, and heat transfer model between individual cells in the battery module, the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the individual cells are obtained, along with the established joint estimation model of battery health and state of charge. The state prediction data of the individual cells is then obtained using the AEKF method. Specifically:

[0054] B1: Using the thermal model of a single cell, the heat transfer model of a single cell, and the heat transfer model between cells in a battery module, predict the heat generation of a single cell, the heat loss from the surface of a single cell, the temperature loss from the core to the surface of a single cell, and the heat transfer between cells in a battery module.

[0055] B2: Establish the observational relationship between the core temperature and surface temperature of a single cell;

[0056]

[0057] In the formula, w t For process noise, v t To observe noise;

[0058] B3: Based on the heat generated by the individual cells, the heat loss on the surface of the individual cells, the temperature loss from the core to the surface of the individual cells, and the heat transferred between the individual cells in the battery module, the core temperature and surface temperature of the individual cells are predicted by using the state-space expression of the multi-state joint estimation of the battery module and the observed relationship between the core temperature and the surface temperature of the individual cells.

[0059] B4: Based on the predicted core temperature and surface temperature of a single cell, obtain the predicted average temperature of the single cells in the battery module.

[0060] B5: Using the identified ohmic internal resistance, polarization internal resistance, polarization capacitance of individual cells and the average temperature of individual cells in the battery module, the AEKF method is used to estimate the ohmic internal resistance, polarization internal resistance, and polarization capacitance of individual cells when the temperature changes.

[0061]

[0062] B6: Estimate the current actual capacity of a single battery cell;

[0063] The method for estimating the current actual capacity of a single battery cell is as follows:

[0064]

[0065] In the formula, k is a discrete-time variable;

[0066] B7: Using the second-order RC equivalent circuit model of a single cell, the estimated current actual capacity of the single cell, the ohmic internal resistance, polarization internal resistance and polarization capacitance of the single cell under temperature change, the AEKF method is used to estimate the state of charge and obtain the state of charge considering temperature compensation.

[0067]

[0068] B8: Discretize the established joint estimation model of battery health and state of charge to obtain the joint estimation model of battery health and state of charge considering temperature compensation.

[0069]

[0070] B9: Based on the estimated current actual capacity of a single cell, the battery health is estimated using a joint estimation model that considers battery health and state of charge under temperature compensation.

[0071]

[0072] In the formula, ΔSoc represents the change in state of charge, ΔT represents the temperature rise, ΔU0 represents the voltage drop, and ΔI represents the current change.

[0073] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0074] Compared to traditional early warning models, this invention, by simultaneously focusing on multiple characteristic parameter signals (such as terminal voltage, state of charge, battery health, and pressure), achieves more accurate early warnings. The electrothermal coupling model is applicable to different temperature environments and comprehensively considers multiple parameters such as temperature, current, voltage, and internal impedance, resulting in more accurate state estimation and rapid prediction of battery status. Addressing the potential safety risks posed by battery temperature, temperature is introduced as a constraint in the prediction process. A resistance-based estimation of SOC (State of Charge) with temperature compensation is employed, along with a joint estimation method of SOC and SOH (State of Health) to improve the accuracy of battery health estimation and further enhance the accuracy of safety early warnings. The method of directly inputting the estimated battery status into the Transform model for safety early warning effectively improves the real-time running speed of the neural network. The use of a multi-head attention mechanism for parallel processing of parallel parameter data more effectively handles multiple characteristic parameters, improving the depth and accuracy of data analysis. In summary, this invention provides high-precision battery health monitoring and safety early warning functions, significantly improving the accuracy and real-time performance of battery safety early warnings. By accurately predicting potential battery failures or anomalies, this method can effectively reduce the risk of potential thermal runaway and improve the safety and reliability of battery systems. Attached Figure Description

[0075] Figure 1 This is a flowchart of a battery safety early warning method based on multi-signal feature fusion in an embodiment of the present invention;

[0076] Figure 2 This is a diagram of a battery safety early warning architecture that incorporates battery health and temperature constraints under multiple characteristic parameters in an embodiment of the present invention.

[0077] Figure 3 This is a flowchart of a battery safety early warning control system considering multi-feature fusion in an embodiment of the present invention;

[0078] Figure 4 The figure shows the core temperature simulation results in an embodiment of the present invention. Detailed Implementation

[0079] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0080] A battery safety early warning method based on multi-signal feature fusion, such as Figure 1 As shown, it includes the following steps:

[0081] Step 1: Construct the second-order RC equivalent circuit model of a single cell, the thermal model of a single cell, the heat transfer model of a single cell, the heat transfer model between single cells in the battery module, the thermal balance model of the battery module, and the available capacity model of the battery module; the battery module includes several single cells;

[0082] Step 1-1: Construct a second-order RC equivalent circuit model of a single cell;

[0083] U0(t)=U OCV (t)-R0I(t)-U1(t)-U2(t) (1)

[0084] In the formula, U0(t) represents the terminal voltage of a single cell, U OCV U(t) represents the open-circuit voltage of a single cell, U1(t) and U2(t) represent the polarization voltages of the two RC circuits of the single cell, R0 represents the ohmic internal resistance of the single cell, I(t) is the charging and discharging current of the single cell, and t represents the time.

[0085] Steps 1-2: The temperature change inside a single cell is usually described by the balance between heat generation and heat loss within the cell. Combining the Joule thermal effect and the chemical thermal effect, a thermal model of the single cell is established.

[0086]

[0087] In the formula, Q(t) is the heat generated by a single cell, T(t) and T are the average temperatures of the single cells in the battery module, and E is the average temperature of the cell. h U is the temperature entropy coefficient. OCV Represents the open-circuit voltage of a single cell;

[0088] Steps 1-3: Simplify the heat transfer path of a single cell to a process where heat is transferred from the inside to the surface and then to the external environment, and construct a heat transfer model for the single cell.

[0089]

[0090] In the formula, Q in (t) represents the temperature at which the core of a single cell loses heat to the surface. c (t) represents the core temperature of a single cell, g represents the thermal conductivity of the material, x represents the distance from the core to the surface of the single cell, and Q represents the thermal conductivity of the material. conv (t) represents the heat loss from the surface of a single cell, h1 is the convective heat transfer coefficient, and T s (t) represents the surface temperature of a single cell, T e (t) represents the ambient temperature, and A1 represents the surface area of ​​a single battery cell;

[0091] Steps 1-4: Establish a heat transfer model between individual cells in the battery module;

[0092]

[0093] In the formula, Q b (t) represents the heat transferred between individual cells in the battery module, where i represents the cell number, T s,i (t) represents the surface temperature of the i-th individual cell in the battery module, h2 is the heat transfer coefficient, A2 is the cross-sectional area for heat conduction, L is the length or thickness of the heat conduction path, and T s,i+1 (t) and T s,i-1 (t) represents the surface temperature of the two cells adjacent to the i-th cell;

[0094] Steps 1-5: Heat transfer characteristics include heat exchange within a single cell, between a single cell and the environment, and between adjacent cells; construct a thermal balance model for the battery module.

[0095]

[0096] In the formula, B c For core heat capacity, B s For surface heat capacity, T c,i (t) represents the core temperature of the i-th individual cell;

[0097] Steps 1-6: The battery module needs to operate with the minimum available capacity to avoid over-discharge or over-charge, which could damage the battery. Therefore, the sum of the capacity of the battery m with the minimum available charging capacity and the capacity of the battery h with the minimum available discharging capacity in the battery module is taken as the minimum single-cell available capacity of the battery module, and then the capacity model of the battery module is constructed.

[0098]

[0099] In the formula, V tol V represents the available capacity of the battery module. char V represents the available charging capacity of the battery module. disc V represents the available discharge capacity of the battery module; N represents the total number of individual cells; V m The capacity of the single battery cell with the smallest available rechargeable capacity; V h S represents the capacity of the single cell with the smallest usable discharge capacity. OC,h S represents the state of charge (S) of the single cell h with the smallest usable discharge capacity. OC,m The state of charge of the single cell m with the smallest available rechargeable capacity;

[0100] Step 2: Based on the second-order RC equivalent circuit model of a single cell, parameter identification is performed using hybrid pulse power characteristic testing to obtain the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the single cell. The parameter identification process is as follows: Figure 2 As shown;

[0101] Step 2-1: Discretize the state of charge (SOC) of a single cell and obtain the open-circuit voltage of the single cell under different SOC states, thus obtaining the open-circuit voltage U of the single cell under different SOC states. OCV The fitted curve;

[0102] The method for discretizing the state of charge of a single cell is as follows:

[0103]

[0104] In the formula, S OC (t) represents the state of charge of a single cell at time t, S OC (t+1) represents the state of charge of a single cell at time t+1, η is the coulombic efficiency of the cell, Δt is the time step between continuous state changes, and V c This represents the current actual capacity of a single battery cell.

[0105] Step 2-2: Measure the terminal voltage and charging / discharging current of a single cell.

[0106] Steps 2-3: Establish the observation equation of a single cell based on the second-order RC equivalent circuit model of the single cell, and then use the hybrid pulse power characteristics test to identify the ohmic internal resistance, polarization internal resistance and polarization capacitance of the single cell based on the observation equation of the single cell.

[0107] Specifically, in the hybrid pulse power characteristic test, the open-circuit voltage U of the single cell under different states of charge obtained in step 2-1 was used at 15℃, 25℃, and 35℃. OCV The fitted curve is used to determine the open-circuit voltage of the individual cell at the current moment. The terminal voltage and charging / discharging current of the individual cell measured in step 2-2, along with the open-circuit voltage at the current moment, are substituted into the observation equation of the individual cell to identify its ohmic internal resistance, polarization internal resistance, and polarization capacitance. The observation equation is as follows:

[0108]

[0109] In the formula, R1 and R2 are the polarization internal resistances of a single cell, and their parallel connection represents electrochemical polarization; C1 and C2 are the polarization capacitances of a single cell, and their parallel connection represents concentration polarization; J is the minimum sum of squared errors. U represents the actual terminal voltage of a single cell at time t. OCV (S OC (t) represents the open-circuit voltage of a single cell at time t (based on the state of charge at time t, the open-circuit voltage U of the single cell under different states of charge). OCV The open-circuit voltage of the corresponding single cell is determined from the fitted curve, t max The maximum number of discrete time steps;

[0110] Step 3: Establish a joint estimation model for battery health and state of charge;

[0111]

[0112] In the formula, S OH (t) represents the battery health at time t; V N This refers to the rated maximum capacity of a single battery cell.

[0113] Step 4: Establish the safety constraints for battery module operation and the state-space expression for the joint estimation of multiple states of the battery module;

[0114] Step 4-1: Since the relevant parameters, namely the internal resistance and terminal voltage of a single cell, change with temperature, safety constraints for the operation of the battery module are constructed. The safety constraints for the operation of the battery module include the current actual capacity constraint of the single cell, the temperature constraint of the single cell, and the state of charge constraint of the single cell. The temperature includes the core temperature of the single cell, the surface temperature of the single cell, and the average temperature of the single cells in the battery module.

[0115] The safety constraints for the operation of the battery module are as follows:

[0116]

[0117] In the formula, T c，max (t) represents the maximum battery core temperature, T cs，max (t) represents the maximum surface and core temperature difference, T max (t) represents the maximum average temperature of the battery. To prevent overcharging and over-discharging, S is set. OC，min S represents the minimum state of charge of a single cell. OC，max This represents the maximum state of charge (SOC) of a single cell.

[0118] Step 4-2: To accurately reflect the thermal characteristics of the battery, the average temperature of the individual cells in the battery module is obtained, and a state-space expression for the multi-state joint estimation of the battery module with respect to temperature is established.

[0119] The state-space expression for the multi-state joint estimation of the battery module is as follows:

[0120]

[0121] In the formula, z is the mass of a single cell, and a k The weighting coefficient represents the influence of charging and discharging current at different historical moments on the current terminal voltage, where β represents the past time. max This represents the total time duration.

[0122] Step 5: Under the constraints of the battery module's thermal balance model, usable capacity model, operational safety constraints, and state-space expression for multi-state joint estimation, based on the individual cell thermal model, heat transfer model, and inter-cell heat transfer model established in Step 1, the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the individual cells obtained in Step 2, and the joint estimation model of battery health and state of charge established in Step 3, the AEKF method is used to obtain the state prediction data of the individual cells. The process is as follows: Figure 3 As shown;

[0123] The state prediction data for a single battery cell includes the terminal voltage, charge / discharge current, temperature, state of charge, and battery health of the single battery cell.

[0124] Step 5-1: Using the thermal model of a single cell, the heat transfer model of a single cell, and the heat transfer model between cells in a battery module, predict the heat generated by a single cell, the heat loss from the surface of a single cell, the temperature loss from the core to the surface of a single cell, and the heat transferred between cells in a battery module.

[0125] Step 5-2: Establish the observational relationship between the core temperature and surface temperature of a single cell;

[0126]

[0127] In the formula, w t For process noise, v t To observe noise;

[0128] Step 5-3: Based on the heat generation of a single cell, the heat loss from the surface of a single cell, the temperature loss from the core to the surface of a single cell, and the heat transfer between cells in the battery module, the core temperature and surface temperature of a single cell are predicted using the state-space expression of the multi-state joint estimation of the battery module and the observed relationship between the core temperature and surface temperature of the single cell. The predicted core temperature of the single cell is as follows: Figure 4 As shown;

[0129] Step 5-4: Based on the predicted core temperature and surface temperature of the individual cells, obtain the predicted average temperature of the individual cells in the battery module.

[0130] Step 5-5: Introduce a temperature compensation mechanism. Using the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the individual cells identified in Step 2 and the average temperature of the individual cells in the battery module obtained in Step 5-5, use the AEKF method to estimate the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the individual cells when the temperature changes.

[0131]

[0132] Steps 5-6: Estimate the current actual capacity of a single battery cell;

[0133] The method for estimating the current actual capacity of a single battery cell is as follows:

[0134]

[0135] In the formula, k is a discrete-time variable;

[0136] Step 5-7: Using the second-order RC equivalent circuit model of a single cell, the estimated current actual capacity of the single cell, and the ohmic internal resistance, polarization internal resistance, and polarization capacitance of the single cell under temperature change obtained in step 5-5, the state of charge is estimated using the AEKF method to obtain the state of charge considering temperature compensation.

[0137]

[0138] Steps 5-8: Discretize the joint estimation model of battery health and state of charge established in step 3 to obtain the joint estimation model of battery health and state of charge considering temperature compensation.

[0139]

[0140] Step 5-9: Based on the estimated current actual capacity of the individual cells, estimate the battery health using the joint estimation model of battery health and state of charge obtained in Step 5-8, which considers temperature compensation.

[0141] Step 6: Calculate the state of charge change, voltage drop, current change, and temperature rise based on the state prediction data of individual cells;

[0142]

[0143] In the formula, ΔSoc is the change in state of charge, ΔT is the temperature rise, ΔU0 is the voltage drop, and ΔI is the current change;

[0144] Step 7: Input the state prediction data of individual cells, state of charge change, voltage drop, current change and temperature rise into the pre-trained transform model to obtain the battery thermal runaway safety warning result. The battery thermal runaway safety warning result is a risk factor, with a value between 0 and 1. The closer it is to 1, the higher the risk level. 0-0.3 is safe, 0.3-0.6 requires attention to the battery's operation, as overcharging or over-discharging may occur. When the prediction result is higher than 0.6, it can be regarded as the early stage of thermal runaway.

[0145] The weighted process of the self-attention mechanism in the transform model is carried out under the safety constraints of the battery module operation, specifically as follows:

[0146]

[0147] In the formula, W e To query the weight matrix, λ u,1 , λ u,2 , λ soc,2 , λ soc,1 , λ i,1 , λ i,2 , λ t,1 , λ t,2 , λ soh,1 The weighted indices are voltage, voltage drop, state of charge, state of charge change, current, current change, temperature, temperature rise, and battery health.

[0148]

[0149] In the formula, W g and W h X is a randomly generated matrix; X is the input to the Transformer neural network, representing battery state data such as voltage, current, state of charge, and temperature; E is the product of the current state parameters of the input battery and the weight coefficients; the G matrix is ​​compared with historical features of the E matrix to determine the data from historical states that have a significant impact on the current state; H is the output matrix, which outputs the prediction results of battery health status and fault level; A is the attention score matrix, representing the weight ratio of each parameter; the Softmax function is used to normalize the data; d z Used to scale attention scores and avoid gradient vanishing.

Claims

1. A battery safety early warning method based on multi-signal feature fusion, characterized in that, The method comprises the following steps: constructing a second-order RC equivalent circuit model of a single battery, a thermal model of the single battery, a heat transfer model of the single battery, a heat transfer model between single batteries in a battery module, a thermal balance model of the battery module, and a capacity model of the battery module; the battery module comprises a plurality of single batteries; based on the second-order RC equivalent circuit model of the single battery, parameters are identified by using a hybrid pulse power characteristic test to obtain an ohmic resistance, a polarization resistance, and a polarization capacitance of the single battery; establishing a joint estimation model of battery health and state of charge; establishing a safety constraint of battery module operation and a state space expression of the joint estimation of the multiple states of the battery module; under the constraints of the thermal balance model of the battery module, the capacity model of the battery module, the safety constraint of the battery module operation, and the state space expression of the joint estimation of the multiple states of the battery module, based on the thermal model of the single battery, the heat transfer model of the single battery, and the heat transfer model between single batteries in the battery module, the ohmic resistance, the polarization resistance, and the polarization capacitance of the single battery, and the joint estimation model of battery health and state of charge, state prediction data of the single battery are obtained by using an AEKF method; based on the state prediction data of the single battery, a state of charge change, a voltage drop, a current change, and a temperature rise are calculated; the state prediction data of the single battery, the state of charge change, the voltage drop, the current change, and the temperature rise are input into a pre-trained transform model to obtain a battery thermal runaway safety warning result, which is a risk factor risk; the second-order RC equivalent circuit model of the single battery is as follows: （1） wherein U0(t) represents the terminal voltage of the single battery, U OCV (t) represents the open circuit voltage of the single battery, and represents the polarization voltage of the two RC circuits of the single battery, R0 represents the ohmic internal resistance of the single battery, I(t) is the charge and discharge current of the single battery, denotes the time instant; the thermal model of the single battery is as follows: （2） where Q(t) is the heat generation of the single battery, T(t) and is the average temperature of the single battery in the battery module, E h is the temperature-entropy coefficient, U OCV represents the open-circuit voltage of the single battery; the heat transfer model of the single battery is as follows: （3） where Q in (t) is the temperature of the core to surface losses of the monobloc battery, T c (t) is the temperature of the core of the monobloc battery, g is the thermal conductivity of the material, x is the distance from the core to the surface of the monobloc battery, Q conv (t) is the heat lost by the monobloc battery surface, hi is the convective heat transfer coefficient, T s (t) is the temperature of the surface of the monobloc battery, T e (t) is the ambient temperature, Ai is the surface area of the monobloc battery; the heat transfer model between single batteries in the battery module is as follows: （4） wherein Q b (t) is the heat transferred between the single cells in the battery module, i represents the number of the single cell, T s,i (t) is the surface temperature of the i-th single cell in the battery module, h2 is the heat transfer coefficient, A2 is the cross-sectional area of heat conduction, L is the length or thickness of the heat conduction path, and are the surface temperatures of the two single cells adjacent to the i-th single cell. the thermal balance model of the battery module is as follows: （5） where B c is the core thermal capacity, B s is the surface thermal capacity, T c,i (t) is the core temperature of the ith monobloc cell; the capacity model of the battery module is as follows: （6） wherein V tol is the available capacity of the battery module; V char is the available charge capacity of the battery module; V disc is the available discharge capacity of the battery module; N is the total number of single cells; V m is the capacity of the single cell with the smallest available charge capacity; V h is the capacity of the single cell with the smallest available discharge capacity; S OC,h is the state of charge of the single cell h with the smallest available discharge capacity, is the state of charge of the single cell m with the smallest available charge capacity.

2. The battery safety early warning method of multi-signal feature fusion according to claim 1, characterized in that, based on the second-order RC equivalent circuit model of the single battery, parameters are identified by using a hybrid pulse power characteristic test to obtain an ohmic resistance, a polarization resistance, and a polarization capacitance of the single battery, specifically as follows: A1: discretize the state of charge of the single battery, and obtain the open circuit voltage of the single battery at different states of charge, to obtain the open circuit voltage U of the single battery at different states of charge OCV of the fitting curve; the state of charge discretization method of the single battery is as follows: （7） where S OC (t) is the state of charge of the cell at time t, S OC (t+1) is the state of charge of the cell at time t+1, η is the battery coulombic efficiency, Δt is the time step between successive state changes, V c is the current actual capacity of the cell; A2: measuring the terminal voltage of the single battery and the charge and discharge current of the single battery; A3: based on the second-order RC equivalent circuit model of the single battery, an observation equation of the single battery is established, and then the ohmic resistance, the polarization resistance, and the polarization capacitance of the single battery are identified by using a hybrid pulse power characteristic test based on the observation equation of the single battery; Specifically, the open circuit voltage U of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different state of charge obtained OCV The open circuit voltage of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different state of charge obtained The open circuit voltage of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different state of charge obtained The open circuit voltage of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different state of charge obtained The open circuit voltage of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different state of charge obtained The open circuit voltage of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different state of charge obtained The open circuit voltage of the single battery at the current moment is determined by using the fitting curve of the open circuit voltage U of the single battery at different （8） where R1 and R2 are the polarization resistance of the single cell and in parallel represent the electrochemical polarization, C1 and C2 are the polarization capacitance of the single cell and in parallel represent the battery concentration polarization, is the minimum error square sum, is the real single cell terminal voltage value at time t, is the open circuit voltage of the single cell at time t max is the maximum discrete time step number.

3. The battery safety early warning method of multi-signal feature fusion according to claim 2, characterized in that, the joint estimation model of battery health and state of charge is as follows: （9） In the formula, S OH (t) is the battery health at time t; V N is the rated maximum capacity of the monobloc battery.

4. The battery safety early warning method of multi-signal feature fusion according to claim 3, characterized in that, the safety constraint of the battery module operation comprises a current actual capacity constraint of the single battery, a temperature constraint of the single battery, and a state of charge constraint of the single battery; the temperature comprises a core temperature of the single battery, a surface temperature of the single battery, and an average temperature of the single battery in the battery module; the safety constraint of the battery module operation is as follows: （10） where T c，max (t) is the maximum battery core temperature, T cs，max (t) is the maximum surface-to-core temperature difference, T max (t) is the maximum average battery temperature, S OC，min is the minimum value of the state of charge of the individual battery, S OC，max is the maximum value of the state of charge of the individual battery the state space expression of the joint estimation of the multiple states of the battery module is as follows: （11） In the formula, z is the mass of the single battery, a k is a weighting coefficient, indicating the influence weight of the charge and discharge current at different historical moments on the current terminal voltage, β is the past moment, β max is the total time length.

5. The battery safety early warning method of multi-signal feature fusion according to claim 4, characterized in that, the state prediction data of the single battery comprises a terminal voltage of the single battery, a charge and discharge current of the single battery, a temperature, a state of charge of the single battery, and a battery health of the single battery.

6. The battery safety early warning method of multi-signal feature fusion according to claim 5, characterized in that, Under the constraints of the thermal balance model of the battery module, the available capacity model of the battery module, the safety constraints of the battery module operation, and the state space expression of the joint estimation of the battery module, based on the established thermal model of the single battery, the heat transfer model of the single battery, and the heat transfer model between the single batteries in the battery module, the obtained ohmic resistance, polarization resistance, and polarization capacitance of the single battery, and the established joint estimation model of the battery health and state of charge, the state prediction data of the single battery is obtained by using the AEKF method, specifically: B1: using the thermal model of the single battery, the heat transfer model of the single battery, and the heat transfer model between the single batteries in the battery module, predicting the heat generation of the single battery, the heat loss of the single battery surface, the temperature loss from the core to the surface of the single battery, and the heat transfer between the single batteries in the battery module; B2: establishing the observation relationship between the core temperature of the single battery and the surface temperature of the single battery; （12） where w t is the process noise, v t is the observation noise; B3: according to the obtained heat generation of the single battery, heat loss of the single battery surface, temperature loss from the core to the surface of the single battery, and heat transfer between the single batteries in the battery module, using the state space expression of the joint estimation of the battery module and the observation relationship between the core temperature of the single battery and the surface temperature of the single battery, predicting the core temperature of the single battery and the surface temperature of the single battery; B4: according to the predicted core temperature of the single battery and the surface temperature of the single battery, obtaining the predicted average temperature of the single battery in the battery module; B5: using the identified ohmic resistance, polarization resistance, and polarization capacitance of the single battery and the average temperature of the single battery in the battery module, estimating the ohmic resistance, polarization resistance, and polarization capacitance of the single battery at temperature change by using the AEKF method; （13） B6: estimating the current actual capacity of the single battery; The method for estimating the current actual capacity of the single battery is: （14） where k is a discrete time variable; B7: using the second-order RC equivalent circuit model of the single battery, the estimated current actual capacity of the single battery, the ohmic resistance, polarization resistance, and polarization capacitance of the single battery at temperature change, and using the AEKF method to estimate the state of charge, obtaining the state of charge considering temperature compensation; （15） B8: discretizing the established joint estimation model of the battery health and state of charge, obtaining the joint estimation model of the battery health and state of charge considering temperature compensation; （16） B9: according to the estimated current actual capacity of the single battery, using the joint estimation model of the battery health and state of charge considering temperature compensation, estimating the battery health; （17） where ΔSoc is the state of charge change, ΔT is the temperature rise, ΔU0 is the voltage drop, and ΔI is the current change.

Images