Lithium ion battery multi-state collaborative online estimation and reliability evaluation method and system

Abstract

The invention discloses a lithium ion battery multi-state collaborative online estimation and reliability evaluation method and system, belongs to the technical field of battery management and state estimation, and aims to solve the problems that battery state collaborative online estimation is difficult and the confidence coefficient of an estimation result is unknown. The method comprises the following steps: establishing an improved equivalent circuit model fusing solid-phase and liquid-phase diffusion mechanisms, a multi-heat-source thermal network model and a degradation evolution model; model parameters are cooperatively identified through an off-line experiment and an on-line double-filter frame, the state of charge SOC, the internal temperature SOT and the state of health SOH of the battery are estimated in real time, a first filter updates external variable thermal resistance, and a second filter fuses voltage and single surface temperature measuring point data to cooperatively estimate the SOC and the SOT; after the model is calibrated through experimental data, the core temperature of the battery is estimated online through an observation filtering method based on a single surface temperature measuring point; a self-adaptive uncertainty quantization module based on filter covariance and measurement residual error is introduced, and a confidence interval of a state estimation value is calculated.

Description

Technical Field

[0001] This invention relates to the field of battery management and state estimation technology, specifically to a method and system for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries. Background Technology

[0002] As the core power unit of new energy vehicles and energy storage systems, the performance, safety, and lifespan of lithium-ion batteries directly affect the reliable operation of the entire system. During battery use, the state of charge (SOC), internal temperature (SOT), and state of health (SOH) are three key parameters characterizing its operating status. However, these core state parameters have complex, strongly coupled, and nonlinear relationships, and most cannot be directly measured by sensors, requiring real-time estimation using models and algorithms.

[0003] While existing battery state estimation technologies have made some progress, the following prominent bottlenecks still exist:

[0004] First, at the state estimation level, there are widespread problems of "model decoupling" and "estimation isolation." Current mainstream Battery Management Systems (BMS) typically use simplified equivalent circuit models for State of Charge (SOC) estimation, while thermal management relies on a limited number of surface temperature sensors. This approach of estimating electrical and thermal states separately ignores crucial intrinsic coupling mechanisms: battery electrical parameters such as internal resistance and polarization are extremely sensitive to temperature, and the battery's heat generation rate is directly related to electrical operating conditions such as SOC and current. Simultaneously, battery aging (SOH decay) causes a slow drift in electrical model parameters (such as capacity and internal resistance). If this is not considered, the estimation errors of SOC and SOH will accumulate over time. Therefore, the lack of a proper electrical model model is detrimental to the estimation process. hot The aging collaborative estimation framework is the root cause of the decline in estimation accuracy of existing methods under complex dynamic conditions and throughout the entire life cycle.

[0005] Secondly, at the temperature sensing level, there are dilemmas of "internal temperature blind spots" and "hardware costs." Safety hazards such as thermal runaway in batteries often stem from abnormal increases in the internal core temperature, rather than the surface temperature. Due to technological cost and reliability limitations, embedding temperature sensors inside the battery cells of mass-produced vehicles is highly impractical. Existing technologies address this issue primarily along two paths: one relies on deploying multiple temperature sensors on the battery surface and inferring the internal temperature through spatial interpolation or simple regression. This method not only increases system complexity and cost but also fails to accurately capture the temperature gradient and local hot spots inside the battery cell due to overly simplified models; the other employs complex high-dimensional computational fluid dynamics (CFD) models, which, while highly accurate, involve enormous computational demands and cannot meet the real-time requirements of automotive BMS. Therefore, how to achieve accurate, online estimation of the battery's internal core temperature without significantly increasing hardware costs and computational burden is a pressing industry challenge.

[0006] Furthermore, at the engineering application level, there is a lack of quantitative assessment of the "reliability of estimation results." Existing estimation algorithms typically only provide an "optimal estimate" without specifying its uncertainty or confidence level. In cases of model mismatch, sudden changes in operating conditions, or sensor malfunctions, the algorithm may have already provided an estimate that deviates significantly from the true value, but the BMS remains unaware and continues to make decisions based on erroneous information, thus creating serious safety hazards. Therefore, a method capable of online assessment of its own estimation reliability is crucial for improving the decision-making intelligence and safety redundancy of the BMS.

[0007] In conclusion, current technology has failed to effectively solve the problem of electricity. hot The strongly coupled estimation problem of aging conditions makes it impossible to accurately sense the internal temperature of the battery in a low-cost manner, and it also lacks an inherent evaluation mechanism for the reliability of the estimation results. These three shortcomings severely limit the performance ceiling and safety boundary of the battery management system. To address this, this invention aims to propose a comprehensive solution that integrates collaborative estimation and uncertainty quantification to overcome the limitations of existing technologies. Summary of the Invention

[0008] To address the challenges of collaborative online estimation of battery states, the dependence of internal temperature on expensive sensors, and the unknown confidence level of estimation results, this invention provides a method and system for collaborative online estimation and reliability assessment of multi-state lithium-ion batteries.

[0009] In a first aspect, this invention provides a method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries, comprising the following steps:

[0010] S1. Model building steps: Establish an improved equivalent circuit model to describe the dynamic voltage response of the battery, a multi-heat source thermal network model to describe the multi-dimensional heat transfer characteristics of the battery, and a degradation evolution model to estimate the state of health (SOH) of the battery.

[0011] S2. Parameter identification and state estimation steps: The parameters of the improved equivalent circuit model and the multi-heat source thermal network model are identified offline and online. Based on the dual filter framework composed of the first filter and the second filter, the battery's state of charge (SOC), internal temperature (SOT), and state of health (SOH) are estimated online in a coordinated manner. At the same time, the external variable thermal resistance parameters are updated online through the first filter.

[0012] S3. Experimental Calibration and Data Fusion Steps: The multi-heat source thermal network model is calibrated and verified using surface and embedded temperature sensor data, and based on a single surface temperature measurement point, the data is fused through observation. The filtering method estimates the battery core temperature and other surface temperatures online.

[0013] Preferably, in step S1, the improved equivalent circuit model is an extension based on the traditional first-order RC equivalent circuit model by introducing solid-phase diffusion mechanism and liquid-phase diffusion mechanism;

[0014] The solid-state diffusion mechanism is dynamically characterized by replacing the volume average SOC with the particle surface SOC as the independent variable of the open-circuit voltage (OCV) function. The particle surface SOC and the volume average SOC are related by the following formula:

[0015]

[0016] In the formula, For particle surface SOC; For the average SOC of the body, , This is the current battery capacity. This represents the battery's current maximum usable capacity. The difference between the volume average SOC and the particle surface SOC. Approximated by a first-order inertial element, its discrete form is:

[0017]

[0018] In the formula, For discrete time points, It is the sampling time frequency. The solid-phase diffusion time constant is for Battery current at any given moment It is the solid-phase diffusion ratio coefficient. for The difference between the volume average SOC and the particle surface SOC at time t;

[0019] The liquid-phase diffusion mechanism is derived based on the limiting current density theory and Ohm's law for liquid phase, and is used to characterize the polarization behavior under liquid-phase mass transfer confinement. The liquid-phase diffusion voltage introduced by this mechanism... Represented as:

[0020]

[0021] In the formula, The time constant for liquid-phase diffusion is... It is the liquid phase diffusion ratio. for Liquid phase diffusion voltage at time t;

[0022] Based on this, the battery's terminal voltage output equation is:

[0023]

[0024] In the formula, Terminal voltage; The internal resistance is ohmic;

[0025] The open-circuit voltage is obtained using the surface state of charge (SOC) of the particles. ;

[0026] Polarization voltage,

[0027] In the formula, for Polarization voltage at time , Equivalent resistance This is the equivalent capacitance.

[0028] Preferably, the multi-heat-source thermal network model establishes a thermal node network structure for square lithium-ion batteries, while considering three types of heat sources: volumetric heat generation of the battery body, ohmic heat generation in the positive electrode area, and ohmic heat generation in the negative electrode area.

[0029] Total heat generation power of the battery body Irreversible heat generation power and reversible heat generation power composition:

[0030]

[0031] in,

[0032]

[0033]

[0034] In the formula, The overall temperature of the battery. It is the entropy heat coefficient;

[0035] The multi-heat-source heat network model will include the above total heat production power. The internal heat source is distributed to the corresponding heat network nodes, and the estimated temperature is finally output by solving the heat balance equations of each node. The estimated temperature includes the estimated temperature of the core and various locations on the surface of the battery.

[0036] Preferably, in step S1, the degradation evolution model obtains the battery capacity degradation characteristic curve based on the ampere-hour integral method and the least squares fitting technique, which is used to estimate the battery health state (SOH).

[0037] Battery State of Health (SOH) is defined as the current maximum available capacity. With rated capacity The ratio:

[0038]

[0039] Current maximum available capacity Update using the following relation:

[0040]

[0041] In the formula, These represent the start and end times of the charge / discharge cycle, respectively. These represent the state of charge at the beginning and end of the charge-discharge cycle, respectively.

[0042] By fitting the above relationship using the least squares method, the current maximum available capacity can be directly identified. value.

[0043] Preferably, the parameter identification in step S2 includes offline identification and online identification;

[0044] The offline identification includes performing HPPC tests on individual battery cells under different temperature conditions, identifying the electrical parameters of the improved equivalent circuit model using a genetic algorithm combined with physical constraints, and identifying the thermal parameters of the multi-heat source thermal network model through thermal characteristic experiments.

[0045] The online identification specifically involves updating the external variable thermal resistance parameter in real time through the first filter to reflect the dynamic changes in external operating conditions.

[0046] Preferably, in step S2, the collaborative online estimation based on the dual-filter framework specifically includes:

[0047] The SOH estimation model outputs the current maximum available capacity. This information is fed back to the improved equivalent circuit model to update the denominator capacity parameter in the volume average SOC calculation.

[0048] The improved equivalent circuit model receives battery current. Initial state of charge and temperature feedback from the multi-heat-source heat network model It outputs the model-estimated state of charge. With the estimated terminal voltage and open circuit voltage ; ;

[0049] The multi-heat-source thermal network model receives battery current. Terminal voltage from the improved equivalent circuit model Open circuit voltage and the external thermal resistance parameters dynamically updated from the first filter. and output the temperature estimated by the model. ;

[0050] The second filter, using the improved equivalent circuit model and the multi-heat-source thermal network model as its state equations, receives its output. and The data is then fused with measurements from the measured terminal voltage and individual surface temperature points to collaboratively estimate the battery's actual state of charge in real time. and actual internal temperature ;

[0051] Meanwhile, at a timescale slower than the second filter, the least squares method is used to estimate the SOH of the battery.

[0052] Preferably, the external variable thermal resistance parameter This corresponds to the convective thermal resistance of multiple surface nodes in a multi-heat-source thermal network model that are in direct contact with the external environment.

[0053] In the online identification and estimation process of step S2, the observations used to update these external variable thermal resistance parameters are only the measurement data of a single surface temperature measuring point.

[0054] Preferably, step S3 specifically includes:

[0055] Offline calibration and verification stage: Using experimental data obtained from temperature sensors placed on the battery core and multiple surfaces, the parameters of the multi-heat source thermal network model are identified and their accuracy is verified.

[0056] Online estimation phase: After completing model calibration and verification, the battery is discretized into a multi-heat source thermal network model containing a core temperature node and multiple surface temperature nodes. The temperature of all nodes constitutes the system state.

[0057] During the online phase, an extended Kalman filter (EKF) or an unscented Kalman filter (UKF) is used as an estimator to estimate the system state based on the measurement data of the individual surface temperature measurement points, thereby obtaining the battery core temperature and other surface temperatures.

[0058] Preferably, the method further includes:

[0059] S4: Uncertainty quantification step: Based on the state covariance matrix output by the second filter and the measurement residual, adaptively calculate the confidence intervals of the SOC and SOT estimates;

[0060] The adaptive calculation includes:

[0061] Calculate the baseline value of the confidence interval based on the state covariance matrix;

[0062] Meanwhile, the measurement noise covariance is adaptively adjusted using the exponential moving average of the measurement residuals so that the confidence interval can reflect model errors and measurement anomalies.

[0063] In a second aspect, the present invention provides a multi-state collaborative online estimation and reliability assessment system for lithium-ion batteries, the system comprising:

[0064] The model building module is used to build improved equivalent circuit models, multi-heat source heat network models, and degradation evolution models.

[0065] The parameter identification and state estimation module is used to perform offline and online identification of model parameters and run a dual-filter framework to collaboratively estimate SOC, SOT, SOH and update external variable thermal resistance.

[0066] The data fusion module is used to process temperature sensor data and estimate the core temperature and other surface temperatures based on single surface temperature measurement points.

[0067] The uncertainty quantification module is used to calculate the confidence interval of the state estimate and evaluate the model's reliability.

[0068] The beneficial effects of this invention are:

[0069] 1. High-precision collaborative estimation across multiple states was achieved, improving the model's adaptability under complex operating conditions. This was achieved by establishing an electric... hot A strongly coupled aging model is employed, and a solid-liquid diffusion mechanism is introduced into the equivalent circuit model. This effectively characterizes the dynamic characteristics under low SOC and high current conditions, significantly reducing the estimation errors of SOC and internal temperature. Simultaneously, a dual-filter framework is used to collaboratively estimate state and time-varying parameters (such as external thermal resistance), enabling the model to adapt to environmental changes and ensuring the estimation accuracy and long-term stability of the algorithm throughout its entire lifespan.

[0070] 2. It overcomes the hardware bottleneck of internal temperature sensing, significantly reducing system cost and complexity. By combining a sophisticated thermal network model with observability analysis, it achieves high-precision estimation of the battery core temperature and other surface temperatures based on a single or very few surface temperature measurement points. This overcomes the shortcomings of existing technologies that rely on multiple sensors or computationally intensive models. While ensuring estimation accuracy, it significantly reduces the dependence on hardware sensors, providing a feasible path for low-cost internal temperature monitoring in mass-produced automotive BMS.

[0071] 3. A lightweight online uncertainty quantification mechanism was pioneered, providing a decision-making basis for battery safety management. By adaptively adjusting the fusion filter covariance and measurement residuals, the confidence interval of the state estimate can be calculated online, and the reliability of the model itself can be evaluated. This upgrades the BMS from the traditional "blindly trusting the estimate" to "informed use of the estimate." When the system identifies that the estimation uncertainty exceeds the limit, it can trigger a conservative management strategy in advance, greatly improving the system's proactive safety warning capability and overall reliability under complex operating conditions such as model mismatch and sensor anomalies.

[0072] 4. A complete technical closed loop from state perception to safety decision-making has been formed, demonstrating strong engineering practicality. This invention integrates high-precision estimation, low-cost implementation, and reliability assessment into a complete solution. The solution has controllable computational complexity and is easily embedded into resource-constrained vehicle BMS or energy storage system monitoring units. It not only provides more accurate state information but also offers crucial reliability data support for battery thermal safety, energy state management, and lifespan prediction, demonstrating broad prospects for engineering applications. Attached Figure Description

[0073] Figure 1 This is a general flowchart of the method described in the embodiments of the present invention.

[0074] Figure 2 This is a schematic diagram of the improved equivalent circuit model used in the embodiments of the present invention.

[0075] Figure 3 This is a flowchart illustrating the multi-state estimation process according to an embodiment of the present invention.

[0076] Figure 4This is a schematic diagram of the battery temperature measurement point arrangement for model building, parameter identification, and filter verification in one embodiment of the present invention. Detailed Implementation

[0077] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0079] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0080] Specific Implementation Method 1: The following is combined with... Figures 1 to 4 This embodiment describes a multi-state collaborative online estimation and reliability assessment method for lithium-ion batteries, which includes the following steps:

[0081] S1. Model building steps: Establish an improved equivalent circuit model to describe the dynamic voltage response of the battery, a multi-heat source thermal network model to describe the multi-dimensional heat transfer characteristics of the battery, and a degradation evolution model to estimate the state of health (SOH) of the battery.

[0082] S2. Parameter identification and state estimation steps: The parameters of the improved equivalent circuit model and the multi-heat source thermal network model are identified offline and online. Based on the dual filter framework composed of the first filter and the second filter, the battery's state of charge (SOC), internal temperature (SOT), and state of health (SOH) are estimated online in a coordinated manner. At the same time, the external variable thermal resistance parameters are updated online through the first filter.

[0083] S3. Experimental Calibration and Data Fusion Steps: The multi-heat source thermal network model is calibrated and verified using surface and embedded temperature sensor data, and based on a single surface temperature measurement point, the data is fused through observation. The filtering method estimates the battery core temperature and other surface temperatures online.

[0084] S4: Uncertainty quantification step: Based on the state covariance matrix output by the second filter and the measurement residual, adaptively calculate the confidence intervals of the SOC and SOT estimates;

[0085] The adaptive calculation includes:

[0086] Calculate the baseline value of the confidence interval based on the state covariance matrix;

[0087] Meanwhile, the measurement noise covariance is adaptively adjusted using the exponential moving average of the measurement residuals so that the confidence interval can reflect model errors and measurement anomalies.

[0088] The following description uses an example of a square lithium iron phosphate power battery. See [link to example]. Figure 1 This enables the coordinated estimation and reliability assessment of electrical, thermal, and aging conditions within the vehicle's battery management system (BMS). Figure 1 The diagram shows the overall flowchart of this method, which mainly includes four core steps: modeling, parameter identification and state estimation, experimental calibration and data fusion, and uncertainty quantification.

[0089] 1. Model establishment (corresponding to step S1)

[0090] First, three core sub-models are established: an improved equivalent circuit model, a multi-heat source heat network model, and a degradation evolution model.

[0091] (1) Improved equivalent circuit model

[0092] like Figure 2 As shown, the improved equivalent circuit model described in this model is based on the traditional first-order RC equivalent circuit model, and is extended by introducing solid-phase diffusion mechanism and liquid-phase diffusion mechanism;

[0093] The solid-state diffusion mechanism is dynamically characterized by replacing the volume average SOC with the particle surface SOC as the independent variable of the open-circuit voltage (OCV) function. The particle surface SOC and the volume average SOC are related by the following formula:

[0094]

[0095] In the formula, For particle surface SOC; For the average SOC of the body, , This is the current battery capacity. This represents the battery's current maximum usable capacity. The difference between the volume average SOC and the particle surface SOC. Approximated by a first-order inertial element, its discrete form is:

[0096]

[0097] In the formula, For discrete time points, It is the sampling time frequency. The solid-phase diffusion time constant is for Battery current at any given moment It is the solid-phase diffusion ratio coefficient. for The difference between the volume average SOC and the particle surface SOC at time t;

[0098] The liquid-phase diffusion mechanism is derived based on the limiting current density theory and Ohm's law for liquid phase, and is used to characterize the polarization behavior under liquid-phase mass transfer confinement. The liquid-phase diffusion voltage introduced by this mechanism... Represented as

[0099]

[0100] In the formula, The time constant for liquid-phase diffusion is... It is the liquid phase diffusion ratio. for Liquid phase diffusion voltage at time t;

[0101] Based on this, the battery's terminal voltage output equation is:

[0102]

[0103] In the formula, Terminal voltage; The internal resistance is ohmic;

[0104] The open-circuit voltage is obtained using the surface state of charge (SOC) of the particles. ;

[0105] Polarization voltage,

[0106] In the formula, for Polarization voltage at time , Equivalent resistance This is the equivalent capacitance.

[0107] In this step, the solid-phase diffusion part is used to compensate for the dynamic voltage response error in the low SOC region, while the liquid-phase diffusion part is used to correct the voltage hysteresis effect under high-current charge and discharge conditions. In the solid-phase diffusion modeling, the independent variable of the open-circuit voltage (OCV) function is replaced by the particle surface SOC, thereby more accurately reflecting the lithium-ion concentration change on the electrode particle surface and achieving dynamic characterization of the solid-phase diffusion process. The liquid-phase diffusion process is derived based on the limiting current density theory and Ohm's law for liquid phase, considering the influence of electrolyte concentration gradient and ion conduction impedance on the terminal voltage, thus characterizing the polarization behavior under liquid-phase mass transfer constraints.

[0108] (2) Multi-heat source heat network model

[0109] For this square battery, a thermal network model with 11 nodes is established, arranged as follows: Figure 4 As shown. Node T9 is located at the core node of the battery cell, while the rest are surface nodes. Node T9 needs to collect the internal temperature of the battery, which cannot be measured by thermocouples and needs to be estimated using surface temperature. This invention establishes a multi-heat source thermal network model offline according to... Figure 4 Temperature sensors are deployed at each node. Node T9 has an embedded thermocouple inserted through the bottom of the battery to the center of the cell to acquire the core temperature. The remaining nodes have surface thermocouples on the battery surface to monitor the surface temperature distribution at different locations. After the multi-heat source network model is trained, experimental calibration and data fusion techniques are used to calibrate and validate the model using data from the surface and embedded temperature sensors. A node position closest to the core temperature (e.g., T5) is retained. In practical applications, the surface temperature of node T5 (temperature data from a single surface temperature measurement point) is collected to estimate the battery core temperature and the surface temperatures of other nodes, achieving the engineering goal of acquiring the battery core temperature and other surface temperatures with the fewest possible sensors.

[0110] The multi-heat-source thermal network model establishes a thermal node network structure for square lithium-ion batteries to accurately describe the internal and external thermal conduction and dissipation characteristics of the cell. This model simultaneously considers three types of heat sources: volumetric heat generation of the battery body, ohmic heat generation in the positive tab region, and ohmic heat generation in the negative tab region; thus, it can more realistically reflect the non-uniform temperature rise characteristics of the cell. In terms of heat transfer path modeling, the heat conduction process along the x, y, and z directions of the battery body is considered. External heat exchange is mainly achieved through convection cooling at the bottom and narrow sides. Since the cell studied is assumed to be located in the central region of the battery module and surrounded by other cells, the top and wide sides are considered as adiabatic boundary conditions. This structure can represent the thermal operating characteristics of a typical cell in the module.

[0111] Total heat generation power of the battery body Irreversible heat generation power and reversible heat generation power composition:

[0112]

[0113] in,

[0114]

[0115]

[0116] In the formula, The overall temperature of the battery. It is the entropy heat coefficient;

[0117] The multi-heat-source heat network model will include the above total heat production power. The internal heat source is distributed to the corresponding heat network nodes, and the estimated temperature is finally output by solving the heat balance equations of each node. The estimated temperature includes the estimated temperature of the core and various locations on the surface of the battery.

[0118] (3) Degenerative evolution model

[0119] The degradation evolution model obtains the battery capacity degradation characteristic curve based on the ampere-hour integral method and the least squares fitting technique, which is used to estimate the battery health state (SOH).

[0120] Battery State of Health (SOH) is defined as the current maximum available capacity. With rated capacity The ratio:

[0121]

[0122] Current maximum available capacity Update using the following relation:

[0123]

[0124] In the formula, These represent the start and end times of the charge / discharge cycle, respectively. These represent the state of charge at the beginning and end of the charge-discharge cycle, respectively.

[0125] make:

[0126] The maximum available capacity can then be directly identified by fitting the above relationship using the least squares method. Value. Maximum available capacity. The value is fed back to the improved equivalent circuit model to update the denominator capacity parameter in the volume average SOC calculation, i.e. The denominator.

[0127] In summary, the calculation of heat relies on parameters such as instantaneous terminal voltage, open-circuit voltage, and battery current provided by the improved equivalent circuit model. Conversely, key parameters of the equivalent circuit model (such as equivalent resistance R and equivalent capacitance C) change dynamically with temperature and need to be updated in real time by the multi-heat-source thermal network model. Therefore, the improved equivalent circuit model (electrical) and the multi-heat-source thermal network model (thermal) are strongly coupled and must be solved collaboratively within a unified framework to realize the battery's electrical... The accurate description of thermal behavior and the estimation of its state cannot be analyzed independently.

[0128] The degradation evolution model used in step S1 to estimate the battery's state of health (SOH) is related to the electrical... The thermal coupling model belongs to sub-models at different time scales. Electricity Thermal models describe transient voltage and temperature changes during charge and discharge, with a relatively fast time scale; while SOH models characterize the long-term evolution of battery performance degradation, with a slower rate of change. Typically, it can be assumed that SOH remains approximately constant over several adjacent charge-discharge cycles, thus being treated as a quasi-static variable in state estimation. Due to the complexity of lithium-ion battery aging mechanisms, involving multiple factors such as solid electrolyte interfacial film growth, active lithium loss, electrode peeling, and electrolyte decomposition, it is difficult to accurately describe its evolution using a single physical model. Therefore, the degradation evolution model employs an identification modeling method based on experimental data, using the ampere-hour integral method and least squares fitting technique to obtain the capacity degradation characteristic curve.

[0129] 2. Parameter identification and state estimation (corresponding to step S2)

[0130] This process is divided into two stages: offline and online.

[0131] (1) Offline identification

[0132] The offline identification includes performing HPPC tests on individual battery cells under different temperature conditions, identifying the electrical parameters of the improved equivalent circuit model using a genetic algorithm combined with physical constraints, and identifying the thermal parameters of the multi-heat source thermal network model through thermal characteristic experiments.

[0133] Electrical parameter identification: First, HPPC (Hybrid PulsePower Characterization) tests are performed on individual battery cells under different temperature conditions. The obtained experimental data are divided into multiple subsets according to temperature and SOC level for subsequent parameter fitting. The electrical parameters of the improved equivalent circuit model are then identified using these test data in conjunction with a genetic algorithm (GA), with physical constraints added during the identification process to ensure that the obtained parameters satisfy physical laws such as monotonicity and non-negativity.

[0134] Thermal parameter identification: For thermal model parameters (such as specific heat capacity, thermal resistance, etc.), identification is required through specially designed thermal characteristic experiments. For example, steady-state heating experiments can be used to obtain convective thermal resistance through constant power heating. Natural cooling experiments were used to estimate heat capacity, combined with... Figure 4 The 11 thermocouple data shown were used to identify the heat capacity, internal thermal resistance, and basic convective thermal resistance through nonlinear least squares inversion. This ensures that the thermal model parameters are both experimentally repeatable and accurately reflect the true thermal conductivity characteristics of the square battery in three dimensions (x, y, z).

[0135] (2) Online identification

[0136] The online identification specifically involves updating the external variable thermal resistance parameter in real time through the first filter to reflect the dynamic changes in external operating conditions.

[0137] The external variable thermal resistance parameter refers to Figure 4 Middle node The parameter set consists of the convective thermal resistance values ​​corresponding to several surface nodes, T10 and T11. In this step, a single Kalman filter is used to update the convective thermal resistance values ​​corresponding to multiple surface nodes online, and the observation is only the measured value of a single surface temperature measuring point T5.

[0138] (3) Collaborative online estimation based on a dual-filter framework

[0139] See Figure 3 The dual-filter framework consists of a first filter and a second filter, which jointly estimate the battery state. The second filter is used to estimate the State of Charge (SOC) and State of Response (SOT) in real time, while the first filter dynamically updates the external variable thermal resistance parameter to reflect changes in the system's thermal boundary. Simultaneously, on a slower timescale, the State of Response (SOH) is estimated using the least squares method, and the result is fed back to correct the capacity denominator (current maximum available capacity) in the SOC calculation. This ensures the accuracy and consistency of the estimation results throughout the battery's entire lifespan.

[0140] The online estimation is as follows:

[0141] The SOH estimation model outputs the current maximum available capacity. This information is fed back to the improved equivalent circuit model to update the denominator capacity parameter in the volume average SOC calculation.

[0142] The improved equivalent circuit model receives battery current. Initial state of charge and temperature feedback from the multi-heat-source heat network model It outputs the model-estimated state of charge. With the estimated terminal voltage and open circuit voltage ; ;

[0143] The multi-heat-source thermal network model receives battery current. Terminal voltage from the improved equivalent circuit model Open circuit voltage and the external thermal resistance parameters dynamically updated from the first filter. and output the temperature estimated by the model. ;

[0144] The second filter, using the improved equivalent circuit model and the multi-heat-source thermal network model as its state equations, receives its output. and It also fuses the data with the measured terminal voltage and individual surface temperature measurement data to output the state of charge. and internal temperature After uncertainty analysis, the actual state of charge of the battery is finally estimated in real time through collaborative calculation. and actual internal temperature ;

[0145] Meanwhile, at a timescale slower than the second filter, the least squares method is used to estimate the SOH of the battery.

[0146] 3. Experimental calibration and data fusion (corresponding to step S3)

[0147] This step is crucial for achieving single-point estimation.

[0148] Offline calibration and validation: In the laboratory, according to Figure 4 Eleven thermocouples were deployed, with node T9 being an embedded thermocouple inserted through the bottom of the battery to the center of the cell to obtain the core temperature; the remaining nodes were surface thermocouples deployed on the battery surface to monitor the surface temperature distribution at different locations. Full temperature field data of the battery under various operating conditions were collected. This data was used to accurately calibrate the aforementioned thermal network model and verify the accuracy of the EKF observer based on single-point T5 in estimating the core temperature T9 (RMSE required). ).

[0149] This invention ensures sufficient observability between the selected measurement point and the core node, as well as other surface nodes, during the model design phase. In engineering implementation, it is preferable to place the measurement point on a surface near the center of the cell or near the tab to increase sensitivity to core temperature dynamics. For the aforementioned state-space model, this invention employs an extended Kalman filter (EKF) as an online estimator to achieve core temperature and other surface temperatures estimation based on single-measurement-point data. At each time step, the electrical and thermal models are calculated in parallel, outputting... and EKF compared this predicted value with the measured value. The data is then fused with T5 to output the optimal estimate. and Each time a complete charge-discharge cycle is completed, the SOH estimation model is triggered and updated. This information is then fed back into the SOC calculation of the electrical model.

[0150] During implementation, to ensure estimation convergence and engineering feasibility, observability analysis and sensitivity simulation of the single measurement point location are required in the offline phase. Observability verification is performed by constructing the observability matrix of the linearized observation matrix and calculating its rank (or by evaluating through numerical singular value decomposition). Sensitivity analysis selects the measurement point location by calculating the amplitude of the sensitivity coefficient of each state component under typical operating conditions. If a single measurement point is unobservable for a certain node under specific operating conditions, this can be addressed by utilizing temporal changes in operating conditions (such as charge-discharge curves) or by selecting a location that improves sensitivity spatially. In the experimental calibration phase, only a small number of embedded points are needed for offline calibration to verify the filter's estimation accuracy. Conventional deployments still primarily use single-surface measurement points to save costs and system complexity. Ultimately, the goal is to estimate the temperature of the battery's interior and other surface points using the surface temperature of a single measurement point. The obtained temperature is used to detect whether the battery's operating state is within a reasonable range. When the battery temperature is too high, timely measures need to be taken to prevent a chain reaction of thermal runaway, ensuring the driving safety of the vehicle and the driver. On the other hand, as a key environmental parameter in the electrical model, it is used to estimate the battery's state of charge (SOC) and ensure an accurate estimate of the battery's capacity.

[0151] Online Deployment: In the actual BMS, only one surface NTC thermistor corresponding to measurement point T5 is installed. During system operation, EKF uses this single measurement point data to drive the calibrated thermal model, which can estimate the core temperature T9 and the temperatures of all other surface nodes in real time.

[0152] 4. Uncertainty quantification (corresponding to step S4)

[0153] To improve system reliability, an uncertainty quantification module is introduced.

[0154] Confidence interval calculation: Online output variable of the second filter EKF State estimate and its covariance matrix For SOC and core temperature T9, the 95% confidence interval is... The calculation is as follows:

[0155]

[0156] In the formula, For the first Temperature measuring points Temperature value at time, For variables standard deviation , covariance matrix The diagonal elements.

[0157] Adaptive noise adjustment: Simultaneously, calculate voltage and temperature measurement residuals. Exponential Moving Average (EMA):

[0158]

[0159] In the formula, As the attenuation factor, , and They are respectively and The exponential moving average of the squared residuals at time t;

[0160] Temperature measurement residual Calculate using the following formula:

[0161]

[0162] This represents the measured value, i.e., the terminal voltage. and the temperature of T5 , express and The functional relationship between them This represents the prior calculated value of the estimated state, i.e. and .

[0163] Subsequently, the measurement noise covariance is adaptively adjusted. :

[0164]

[0165] In the formula, The lower limit of measurement noise set empirically, a constant. It serves to conservatively amplify the information to prevent exaggerated claims.

[0166] This operation automatically widens the confidence interval when there is model mismatch or measurement anomalies, thus truly reflecting the increase in estimation uncertainty.

[0167] Confidence flag: If the root mean square value of the residuals continuously exceeds the threshold, the "low confidence" flag is set, and the BMS can trigger an alarm or switch to a conservative strategy accordingly.

[0168] The above adaptive The update will be directly used for filter gain calculation, thereby increasing the covariance. The uncertainty automatically expands when there is model error or measurement anomaly, reflecting greater uncertainty; conversely, when the residuals are stable, the uncertainty converges to a smaller value. In addition to confidence intervals based on covariance, this invention also employs a simple confidence determination strategy based on residual thresholds: the short-term root mean square value of the residuals... Compared with empirical thresholds In comparison, if If the threshold is not met, a model confidence reduction flag is triggered, and one or a combination of the following actions are performed: indicating the need for recalibration, increasing the filter process noise R to increase robustness, or initiating a conservative thermal management strategy in the BMS (e.g., reducing the maximum charging current). This threshold strategy, together with the confidence interval, constitutes a concise and effective online uncertainty management mechanism, which is easy to implement in resource-constrained embedded environments.

[0169] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries, characterized in that, Includes the following steps: S1. Model building steps: Establish an improved equivalent circuit model to describe the dynamic voltage response of the battery, a multi-heat source thermal network model to describe the multi-dimensional heat transfer characteristics of the battery, and a degradation evolution model to estimate the state of health (SOH) of the battery. S2. Parameter identification and state estimation steps: The parameters of the improved equivalent circuit model and the multi-heat source thermal network model are identified offline and online. Based on the dual filter framework composed of the first filter and the second filter, the battery's state of charge (SOC), internal temperature (SOT), and state of health (SOH) are estimated online in a coordinated manner. At the same time, the external variable thermal resistance parameters are updated online through the first filter. S3. Experimental Calibration and Data Fusion Steps: The multi-heat source thermal network model is calibrated and verified using surface and embedded temperature sensor data, and based on a single surface temperature measurement point, the data is fused through observation. The filtering method estimates the battery core temperature and other surface temperatures online.

2. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 1, characterized in that, In step S1, the improved equivalent circuit model is based on the traditional first-order RC equivalent circuit model, and is extended by introducing solid-phase diffusion mechanism and liquid-phase diffusion mechanism. The solid-state diffusion mechanism is dynamically characterized by replacing the volume average SOC with the particle surface SOC as the independent variable of the open-circuit voltage (OCV) function. The particle surface SOC and the volume average SOC are related by the following formula: In the formula, For particle surface SOC; For the average SOC of the body, , This is the current battery capacity. This represents the battery's current maximum usable capacity. The difference between the volume average SOC and the particle surface SOC. Approximated by a first-order inertial element, its discrete form is: In the formula, For discrete time points, It is the sampling time frequency. The solid-phase diffusion time constant is for Battery current at any given moment It is the solid-phase diffusion ratio coefficient. for The difference between the volume average SOC and the particle surface SOC at time t; The liquid-phase diffusion mechanism is derived based on the limiting current density theory and Ohm's law for liquid phase, and is used to characterize the polarization behavior under liquid-phase mass transfer confinement. The liquid-phase diffusion voltage introduced by this mechanism... Represented as: In the formula, The time constant for liquid-phase diffusion is... It is the liquid phase diffusion ratio. for Liquid phase diffusion voltage at time t; Based on this, the battery's terminal voltage output equation is: In the formula, Terminal voltage; The internal resistance is ohmic; The open-circuit voltage is obtained using the surface state of charge (SOC) of the particles. ; Polarization voltage, In the formula, for Polarization voltage at time , Equivalent resistance This is the equivalent capacitance.

3. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 2, characterized in that, The multi-heat-source thermal network model establishes a thermal node network structure for square lithium-ion batteries, while considering three types of heat sources: volumetric heat generation of the battery body, ohmic heat generation in the positive electrode area, and ohmic heat generation in the negative electrode area. Total heat generation power of the battery body Irreversible heat generation power and reversible heat generation power composition: in, In the formula, The overall temperature of the battery. It is the entropy heat coefficient; The multi-heat-source heat network model will include the above total heat production power. The internal heat source is distributed to the corresponding heat network nodes, and the estimated temperature is finally output by solving the heat balance equations of each node. The estimated temperature includes the estimated temperature of the core and various locations on the surface of the battery.

4. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 3, characterized in that, In step S1, the degradation evolution model obtains the battery capacity degradation characteristic curve based on the ampere-hour integral method and the least squares fitting technique, which is used to estimate the battery health state SOH. Battery State of Health (SOH) is defined as the current maximum available capacity. With rated capacity The ratio: Current maximum available capacity Update using the following relation: In the formula, These represent the start and end times of the charge / discharge cycle, respectively. These represent the state of charge at the beginning and end of the charge-discharge cycle, respectively. By fitting the above relationship using the least squares method, the current maximum available capacity can be directly identified. value.

5. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 1, characterized in that, The parameter identification in step S2 includes offline identification and online identification; The offline identification includes performing HPPC tests on individual battery cells under different temperature conditions, identifying the electrical parameters of the improved equivalent circuit model using a genetic algorithm combined with physical constraints, and identifying the thermal parameters of the multi-heat source thermal network model through thermal characteristic experiments. The online identification specifically involves updating the external variable thermal resistance parameter in real time through the first filter to reflect the dynamic changes in external operating conditions.

6. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 5, characterized in that, In step S2, the collaborative online estimation based on the dual-filter framework specifically includes: The SOH estimation model outputs the current maximum available capacity. This information is fed back to the improved equivalent circuit model to update the denominator capacity parameter in the volume average SOC calculation. The improved equivalent circuit model receives battery current. Initial state of charge and temperature feedback from the multi-heat-source heat network model It outputs the model-estimated state of charge. With the estimated terminal voltage and open circuit voltage ; ; The multi-heat-source thermal network model receives battery current. Terminal voltage from the improved equivalent circuit model Open circuit voltage and the external thermal resistance parameters dynamically updated from the first filter. and output the temperature estimated by the model. ; The second filter, using the improved equivalent circuit model and the multi-heat-source thermal network model as its state equations, receives its output. and The data is then fused with measurements from the measured terminal voltage and individual surface temperature points to collaboratively estimate the battery's actual state of charge in real time. and actual internal temperature ; Meanwhile, at a timescale slower than the second filter, the least squares method is used to estimate the SOH of the battery.

7. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 1 or 5, characterized in that, The external variable thermal resistance parameter This corresponds to the convective thermal resistance of multiple surface nodes in a multi-heat-source thermal network model that are in direct contact with the external environment. In the online identification and estimation process of step S2, the observations used to update these external variable thermal resistance parameters are only the measurement data of a single surface temperature measuring point.

8. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 1, characterized in that, Step S3 specifically includes: Offline calibration and verification stage: Using experimental data obtained from temperature sensors placed on the battery core and multiple surfaces, the parameters of the multi-heat source thermal network model are identified and their accuracy is verified. Online estimation phase: After completing model calibration and verification, the battery is discretized into a multi-heat source thermal network model containing a core temperature node and multiple surface temperature nodes. The temperature of all nodes constitutes the system state. During the online phase, an extended Kalman filter (EKF) or an unscented Kalman filter (UKF) is used as an estimator to estimate the system state based on the measurement data of the individual surface temperature measurement points, thereby obtaining the battery core temperature and other surface temperatures.

9. The method for multi-state collaborative online estimation and reliability assessment of lithium-ion batteries according to claim 1, characterized in that, The method further includes: S4: Uncertainty quantification step: Based on the state covariance matrix output by the second filter and the measurement residual, adaptively calculate the confidence intervals of the SOC and SOT estimates; The adaptive calculation includes: Calculate the baseline value of the confidence interval based on the state covariance matrix; Meanwhile, the measurement noise covariance is adaptively adjusted using the exponential moving average of the measurement residuals so that the confidence interval can reflect model errors and measurement anomalies.

10. A multi-state collaborative online estimation and reliability assessment system for lithium-ion batteries, used to implement the multi-state collaborative online estimation and reliability assessment method for lithium-ion batteries as described in any one of claims 1-9, characterized in that, The system includes: The model building module is used to build improved equivalent circuit models, multi-heat source heat network models, and degradation evolution models. The parameter identification and state estimation module is used to perform offline and online identification of model parameters and run a dual-filter framework to collaboratively estimate SOC, SOT, SOH and update external variable thermal resistance. The data fusion module is used to process temperature sensor data and estimate the core temperature and other surface temperatures based on single surface temperature measurement points. The uncertainty quantification module is used to calculate the confidence interval of the state estimate and evaluate the model's reliability.

Images