A lithium battery state of charge and core temperature joint estimation method and device

Abstract

The application discloses a lithium battery state of charge and core temperature combined estimation method and device, first, a lithium battery is obtained for multi-working condition experiment data set; then, an electrochemical thermal model is established, and reaction rate constant and diffusion coefficient are corrected; the relationship between electrode solid phase average lithium ion concentration and positive and negative electrode solid phase surface lithium ion concentration and liquid phase lithium concentration distribution are calculated; then, SOC and lithium battery voltage are calculated; the electrochemical thermal model is coupled with a neural network model and trained, time series data containing current and voltage are input, and finally, positive and negative electrode solid phase surface lithium ion concentration is output; based on the model output after training, the output is transmitted to the electrochemical thermal model with corrected parameters, and voltage, SOC and core temperature calculation results are output. The lithium battery SOC calculated by the method has better multi-working condition adaptability and performs more excellent in a wide temperature range while improving the calculation precision.

Description

Technical Field

[0001] This invention relates to the field of lithium battery SOC and temperature estimation technology, and in particular to a method and apparatus for jointly estimating the state of charge and core temperature of a lithium battery. Background Technology

[0002] In new energy vehicles and large-scale energy storage systems, lithium batteries serve as the core energy carrier, and their accurate monitoring directly impacts the system's safety, durability, and energy management efficiency. The Battery Management System (BMS) needs to accurately estimate the battery's State of Charge (SOC) and temperature state in real time. SOC is a key indicator of the battery's remaining capacity, providing a basis for range prediction and charge / discharge strategies. Core temperature directly reflects the heat accumulation of electrochemical reactions inside the battery. Compared to surface temperature, it more accurately represents the true temperature state inside the cell, providing crucial temperature feedback for electrochemical models. This facilitates online correction of model parameters and high-fidelity electro-thermal coupling modeling, playing a vital role in early warning of thermal risks and prevention of thermal runaway. Since lithium-ion batteries are highly nonlinear and time-varying electrochemical-thermal coupled systems, their electrical and thermal behaviors are closely related and mutually influential. Therefore, achieving high-precision joint estimation of SOC and core temperature under complex real-world conditions has significant engineering implications.

[0003] Currently, battery state estimation methods are mainly divided into direct measurement methods, model-driven methods, and data-driven methods, each with its own limitations. Direct measurement methods, represented by the ampere-hour integration method, are simple and easy to implement, but they rely on high-precision current sensors and are sensitive to initial errors, making it difficult to eliminate accumulated errors over long-term operation. Model-driven methods mainly include equivalent circuit models and electrochemical models: equivalent circuit models have lower computational load, but model parameters are significantly affected by factors such as temperature and aging, requiring frequent calibration and exhibiting poor consistency throughout the battery's lifespan; electrochemical models describe the internal processes of the battery at the mechanistic level, exhibiting good physical consistency, but due to the large number of partial differential equations involved, they have a heavy computational burden and are difficult to solve in real-time within a battery management system (BMS).

[0004] Neural network-based data-driven methods do not rely on precise mechanistic models and can learn the nonlinear relationships between voltage, current, temperature, and state from historical data. However, these methods are "black box" models, lacking physical constraints, and their performance is heavily dependent on the coverage of the training data. When actual operating conditions (such as high-rate operation, extreme temperatures, or aging conditions) deviate from the distribution of the training data, the model's generalization ability decreases, and the estimation results may be severely distorted. Furthermore, in terms of thermal state estimation, most existing methods rely only on surface temperature and fail to fully consider the internal and external temperature differences and thermal hysteresis effects of the battery under high-current conditions, thus failing to accurately capture the true internal thermal state of the battery.

[0005] Existing lithium battery SOC estimation methods (a method for estimating the SOC of all-electric aircraft lithium batteries based on a multi-field coupled physical information neural network, CN 120559474 A) introduce temperature and mechanical deformation fields into a single-particle model of the all-electric aircraft lithium battery to obtain a lithium battery SPM model; construct a neural network, using the calibrated multi-field coupled all-electric aircraft lithium battery SPM model as the physical constraint of the neural network, to obtain a multi-field coupled physical information neural network; construct a joint loss function by summing the SOC mean square error of the neural network and the voltage mean square error of the calibrated SPM model, train the multi-field coupled physical information neural network based on the joint loss function, to obtain a trained physical information neural network; estimate the SOC of the all-electric aircraft lithium battery based on the trained physical information neural network, to obtain the predicted SOC. This application can achieve accurate estimation and prediction of the SOC of all-electric aircraft lithium batteries under complex operating conditions. This method has the following three shortcomings:

[0006] 1) The electrochemical model used in this scheme has significant errors under high-rate conditions. The LSTM neural network model is good at solving time series prediction, but not good at data feature extraction. The electrochemical model used in this scheme is the SPM model, which ignores the kinetic behavior of liquid lithium diffusion in the electrolyte. However, SPM is prone to introducing large errors under high-rate conditions (above C / 2) due to ignoring electrolyte kinetics. Therefore, this scheme uses the single-particle model SPMe, which takes into account electrolyte kinetics. The neural network model used in this scheme is LSTM. This scheme uses CNN as the feature extraction layer, LSTM as the time series modeling layer, and finally a fully connected layer as the output layer. By using a hybrid neural network model, time series features can be extracted better.

[0007] 2) The coupling degree of the thermal model in this scheme is limited: the coupled temperature field in this scheme only uses the battery temperature to correct the electrochemical model parameters (reaction rate constant k and lithium-ion diffusion coefficient D) in the Arrhenius equation; this scheme, on the other hand, calculates the battery core temperature first based on the second-order thermal model and the battery voltage, current, and surface temperature. Then, the Arrhenius equation is used to correct the rate constants of the positive and negative electrode reactions. Diffusion coefficient of lithium ions in positive and negative electrode solid phases The liquid-phase lithium-ion diffusion coefficient was corrected using the Vogel-Tammann-Fulcher (VTF) equation. ;

[0008] 3) The neural network training logic that couples physical information in this scheme is somewhat lacking: This scheme couples physical information by calculating the average lithium concentration on the negative electrode using the predicted SOC value from the neural network (the document states it's the negative electrode surface, but the formula actually calculates the average concentration; the difference between the average concentration and the negative electrode surface concentration needs to be calculated using the solid-phase lithium concentration diffusion equation or its reduced-order formula), which serves as input to the SPM model to further simulate the battery's voltage response, constructing a joint loss function using the SOC mean square error and voltage mean square error. This scheme, however, predicts the lithium concentration on the solid-phase surface of both the positive and negative electrodes using a neural network. , The voltage can be calculated by inputting the data into the SPMe model, and then the average lithium concentration in the solid phase of the positive and negative electrodes can be obtained by using the reduced-order polynomial equation of the solid-phase lithium diffusion partial differential equation. That is, the SOC is calculated, and the neural network is trained using the voltage and the mean square error of SOC as the physical information loss function.

[0009] Another existing SOC estimation method (a joint state estimation method for power battery SOC and SOT based on an electro-thermal-neural network coupled model, CN 110703114 B) includes: S1 Selecting the power battery to be tested, collecting and organizing relevant technical parameters, establishing a continuous electro-thermal-electro-model (ETSM) for the power battery in the time domain, and determining the model input and output parameters; S2 Simulating two urban driving cycles and two highway driving cycles at different temperatures to collect battery state data; S3 Identifying parameters of the ETSM model to obtain the characteristic parameters of the electro-thermal-electro-model; S4 Generating a training database using the parameterized ETSM model, training the neural network, establishing an ETNN model, and estimating the terminal voltage Vsp and core temperature Tc; S5 Using UKF to jointly estimate the battery SOC and SOT. This invention can effectively estimate the state of a power battery under extreme temperature and high current conditions. This method has the following two shortcomings:

[0010] 1) The electrochemical model used in this scheme will have a large error under high-rate operating conditions. The neural network model used in this scheme is not clearly described: Similar to point 1 of patent (CN 120559474 A), the electrochemical model used in that scheme is SPM, while this scheme uses the SPMe model; the neural network model used in that scheme is not described in detail. The hybrid neural network model used in this scheme uses CNN as the feature extraction layer, LSTM as the temporal modeling layer, and finally outputs through a fully connected layer.

[0011] 2) The coupling between the model and the neural network in this scheme is not deep. The main function of the electro-thermal-neural network is in voltage correction: The joint SOC and SOT estimation method based on the electro-thermal-neural network in this scheme uses surface temperature and current as electrical inputs to the electro-thermal sub-model (ETSM) to calculate the core temperature Tc and voltage Vsp. Then, the calculated core temperature Tc and voltage Vsp are used as inputs to the neural network to obtain the voltage V corrected by the neural network. Finally, using the electro-thermal model formula as the state function and the neural network output as the measurement function, an unscented Kalman filter (UKF) is used for SOC estimation. Therefore, the main role of the electro-thermal-neural network in this scheme is to use the neural network to correct the accuracy of the simplified SPM model under high current and extreme ambient temperatures, outputting a more accurate voltage V before using the UKF for SOC estimation. This scheme first adopts the SPMe electrochemical model, which is more adaptable to extreme working conditions, and then deeply couples the electrochemical thermal model with the neural network. The output of the neural network is calculated by the electrochemical thermal model to obtain the SOC and voltage V. Then, the weighted mean square error of SOC and V is used as the physical information loss function. Through backpropagation training, the output of the neural network model has both data fitting ability and meets the laws of the physical model. Summary of the Invention

[0012] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and apparatus for jointly estimating the state of charge and core temperature of lithium batteries, in order to solve the following problems:

[0013] (1) Solving the problem of poor generalization ability and distorted estimation results under unknown or extreme conditions due to the lack of physical constraints in pure data-driven methods: Existing data-driven methods based on neural networks are "black box" models, and their performance is heavily dependent on the coverage of training data. When the actual operating conditions (such as high magnification, extreme temperature, different aging degrees) deviate from the distribution of training data, the generalization ability of the model decreases significantly, and the estimation results may seriously deviate from the physical reality. This invention needs to construct a neural network model that integrates clear physical mechanisms to enhance its robustness and physical consistency across the entire operating range.

[0014] (2) Solving the problem of decreased estimation accuracy of traditional electrochemical models under high-rate conditions due to neglect of electrolyte kinetics: The simplified electrochemical models (such as SPM) used in existing methods have large errors in voltage and SOC prediction when charging and discharging at high rates (above C / 2) because they do not consider the mass transfer process of lithium ions in the electrolyte. This invention aims to provide a battery state estimation method that can maintain high accuracy under high-rate and complex dynamic conditions.

[0015] (3) Solving the problem of neglecting the internal temperature difference and thermal hysteresis effect under high current conditions in existing methods: Most existing methods only rely on the battery surface temperature for thermal management or parameter correction, failing to accurately estimate the battery core temperature and reflect the true internal thermal state of the battery, thus affecting the accurate correction of temperature-sensitive parameters (reaction rate, diffusion coefficient). This invention needs to establish a thermal model that can accurately estimate the internal core temperature of the battery and effectively couple it with the electrochemical process to achieve accurate online correction of key temperature-sensitive parameters.

[0016] (4) To address the problem that existing neural network model structures are insufficient in extracting features from battery time-series data and capturing long-term dependencies: Although recurrent neural networks (such as LSTM) used in existing solutions are good at processing time-series data, they have limitations in automatically extracting local and global deep features from raw current and voltage data. This invention needs to design a more effective hybrid neural network structure to simultaneously capture local features (such as current impulse response) and long-term time-series evolution patterns (such as lithium concentration diffusion dynamics) of battery operating data.

[0017] The objective of this invention is achieved through the following technical solution: a method for jointly estimating the state of charge and core temperature of a lithium battery, the method comprising the following steps:

[0018] (1) Perform dynamic testing on lithium batteries under multiple operating conditions to obtain experimental data sets;

[0019] (2) Establish an electrochemical thermal model, calculate the core temperature of the lithium battery, and correct the reaction rate constant and diffusion coefficient; based on the conservation of lithium ion mass in the solid phase and the conservation of lithium ion mass in the liquid phase, obtain the relationship between the average lithium ion concentration in the solid phase of the electrode and the lithium ion concentration on the surface of the solid phase of the positive and negative electrodes, as well as the lithium concentration distribution in the liquid phase; then calculate the SOC and obtain the liquid phase potential difference based on the liquid phase lithium concentration distribution, and then calculate the lithium battery voltage.

[0020] (3) An adaptive particle swarm optimization algorithm combined with an adaptive adjustment strategy for inertial weights is used to optimize the electrochemical thermal model parameters, battery geometric parameters, and electrochemical model parameters;

[0021] (4) Couple the electrochemical thermal model with the neural network model, input the time series data containing current and voltage, and finally output the lithium ion concentration on the solid surface of the positive and negative electrodes; train the model based on the mean square error physical information loss function containing voltage and SOC, and transmit the model output after training to the electrochemical thermal model with corrected parameters to output the calculated results of voltage, SOC and core temperature.

[0022] Further, step (1) specifically involves: screening lithium batteries through standard battery charge and discharge experiments, conducting thermocouple embedding experiments on the screened lithium batteries, and conducting multi-condition dynamic testing experiments on the lithium batteries with embedded thermocouples, thereby obtaining the voltage, current, surface temperature, core temperature and SOC of the lithium batteries in the multi-condition dynamic test.

[0023] Furthermore, in step (2), the thermal model adopts a second-order equivalent thermal resistance model to describe the heat transfer inside the battery, establishes the core thermal balance equation of the battery and discretizes it, and combines the internal heat generation power formula to establish the relationship between the lithium battery voltage, current, average lithium concentration of the positive and negative electrode solid phases and temperature.

[0024] Furthermore, in step (2), the electrochemical model adopts a single-particle model that considers electrolyte kinetics, including solid-phase lithium ion mass conservation, liquid-phase lithium ion mass conservation, solid-phase charge conservation, liquid-phase charge conservation, and electrode reaction kinetics.

[0025] Furthermore, in step (2), the positive and negative electrode reaction rate constants and the solid phase diffusion coefficient are corrected using the Arrhenius law, and the liquid phase diffusion coefficient is corrected using the VTF equation.

[0026] Furthermore, for the optimization of thermal model parameters, the core heat capacity and core-to-surface thermal resistance are corrected by minimizing the root mean square error of the core temperature prediction; for the optimization of geometric model parameters, the core heat capacity and core-to-surface thermal resistance are corrected by minimizing the root mean square error between the predicted solid-phase lithium concentration and the theoretical solid-phase lithium concentration; for the optimization of electrochemical model parameters, the initial solid-phase diffusion coefficient, positive and negative electrode SEI film resistance, and initial positive and negative electrode reaction rate constants are corrected by minimizing the root mean square error of the voltage prediction.

[0027] Furthermore, in step (4), the neural network model is a CNN-LSTM hybrid neural network model, which extracts local and short-term pattern features based on CNN and captures long-term dynamic evolution of lithium concentration by LSTM.

[0028] Secondly, the present invention also provides a lithium battery state of charge and core temperature joint estimation device, including a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, it implements the lithium battery state of charge and core temperature joint estimation method.

[0029] Thirdly, the present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the aforementioned method for jointly estimating the state of charge and core temperature of a lithium battery.

[0030] Fourthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method for jointly estimating the state of charge and core temperature of a lithium battery.

[0031] The beneficial effects of this invention are:

[0032] (1) Model: The electrochemical model used in the prior art is the SPM model, which ignores the kinetic behavior of liquid lithium diffusion in the electrolyte; however, SPM is prone to introducing large errors under high-rate (above C / 2) conditions due to the neglect of electrolyte kinetics. Therefore, this scheme uses the single-particle model SPMe, which takes into account electrolyte kinetics.

[0033] (2) Algorithm: The existing technology uses LSTM as the neural network model. This scheme uses CNN as the feature extraction layer, LSTM as the temporal modeling layer, and finally a fully connected layer as the output layer. By using a hybrid neural network model, temporal features can be extracted better.

[0034] (3) Regarding the accuracy of SOC estimation:

[0035] The lithium battery SOC calculated by the method of this invention not only improves the calculation accuracy, but also has better adaptability to multiple operating conditions and performs better in a wide temperature range. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 The overall flowchart of a method for jointly estimating the state of charge and core temperature of a lithium battery provided by the present invention;

[0038] Figure 2 A flowchart for obtaining multi-condition dynamic testing experiments for training and testing sets;

[0039] Figure 3 Flowchart for optimizing parameters of an electrochemical thermal model;

[0040] Figure 4The calculation of the core temperature of the thermal model (taking test condition 1 at 25℃ as an example) and the error statistics chart are provided.

[0041] Figure 5 For the voltage calculation of the electrochemical model (taking test condition 1 at 25℃ as an example) and error statistics;

[0042] Figure 6 This is a system architecture diagram of an electrochemical thermal model coupled with a hybrid neural network.

[0043] Figure 7 The figure shows the SOC estimation results (taking test condition 3 at 10℃ / 25℃ / 40℃ as an example);

[0044] Figure 8 This is a structural diagram of a lithium battery state of charge and core temperature joint estimation device provided by the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are merely illustrative and not intended to limit the invention.

[0046] like Figure 1 As shown, the present invention provides a method for jointly estimating the state of charge and core temperature of a lithium battery, the specific process of which is as follows:

[0047] (1) Obtain the experimental dataset of lithium battery under multi-condition dynamic testing as the training set and test set.

[0048] Experimental flowchart as follows Figure 2 As shown, it includes standard battery charge and discharge experiments, thermocouple embedding experiments, and multi-condition dynamic testing experiments.

[0049] Taking a 21700 cylindrical battery with a rated capacity of 4.8Ah NCM811 positive electrode and a graphite / silicon composite negative electrode as an example.

[0050] First, standard charge-discharge tests were conducted on the batteries. The lithium batteries were placed in a constant temperature chamber at 25°C for 2 hours to ensure uniform temperature distribution within each cell. The batteries were then charged at a constant current of 0.3C to the charging cutoff voltage of 4.20V, followed by constant voltage charging until the current dropped below C / 30. After resting for 30 minutes, the batteries were discharged at a rate of 0.2C until the voltage reached the discharge cutoff voltage of 2.50V. Based on the standard charge-discharge test data, cells with high consistency were selected.

[0051] Thermocouple embedding experiments were conducted on the screened lithium batteries. A cross-sectional computed tomography scan was performed on a selected lithium battery sample to confirm that the battery core had a certain gap (approximately 3 mm). The lithium battery was then placed in a vacuum glove box, and a hole (0.45 mm in diameter) was drilled in the center of the negative electrode side of the cylindrical battery. An extremely fine thermocouple (0.1 mm in wire diameter) was inserted. After sealing with glue in the vacuum glove box, the mixture was allowed to stand until the glue was completely dry to directly measure the core temperature of the battery. This method provides crucial experimental data for the calibration and validation of thermal models;

[0052] Multi-condition dynamic testing was conducted on lithium batteries with embedded thermocouples. The lithium batteries were charged at a 1C rate using CC / CV charging, and then left to stand at the test temperatures (10℃ / 25℃ / 40℃) for 2 hours to ensure uniform temperature distribution among individual cells. To prevent overvoltage during the random charging portion of the dynamic tests, the batteries were discharged to 90% SOC at a constant current of 1C, and then subjected to four dynamic tests to discharge to 10% SOC. Test 1 was the DST dynamic stress test, Test 2 was the BJDST urban dynamic stress test, Test 3 was the FUDS urban driving test, and Test 4 was the US06 highway driving test. The battery voltage during the multi-condition dynamic tests was obtained. Current Surface temperature Core temperature and SOC.

[0053] (2) Establish an electrochemical thermal model

[0054] The thermal model uses a second-order equivalent thermal resistance model to describe the heat transfer inside the battery (from the core to the surface). Assuming that the temperature inside the battery cells is uniform, the thermal balance equation of the battery core is:

[0055]

[0056] in, For core heat capacity, For time, For the core temperature, For internal heat generation power, Surface temperature, The thermal resistance from the core to the surface.

[0057] Discretizing the core thermal balance equation of the battery, we get:

[0058]

[0059] in, For sampling points, The sampling interval is denoted as .

[0060] By using the discrete equation of the core thermal balance of a lithium battery and the formula for internal heat generation power, the relationship between the battery voltage, current, average lithium concentration in the solid phase of the positive and negative electrodes, and temperature can be established. The formula for internal heat generation power is as follows:

[0061]

[0062] in, For current, For voltage, The positive open-circuit potential, The negative open-circuit potential, This represents the average lithium-ion concentration in the positive electrode solid phase. The average lithium-ion concentration in the negative electrode solid phase.

[0063] Based on the core temperature calculated using the thermal model, parameters in the electrochemical model that are significantly affected by temperature changes can be corrected, including the reaction rate constant and diffusion coefficient. Electrochemical model solid-state parameters (positive and negative electrode reaction rate constants) and solid-phase diffusion coefficient The correction is made using Arrhenius's law:

[0064]

[0065] in, It refers to the core temperature Lowered parameter values ​​(positive and negative electrode reaction rate constants) and solid-phase diffusion coefficient ), For these parameters, pre-exponential factors, For activation energy, The gas constant is... This is a reference temperature.

[0066] Because the viscosity and ion migration behavior of the electrolyte undergo drastic nonlinear changes at low temperatures, the liquid phase parameters (liquid phase diffusion coefficient) in the electrochemical model... The Vogel-Tammann-Fulcher (VTF) equation, which better describes the phenomenon, is used for correction to improve the accuracy of low-temperature estimation.

[0067]

[0068] in, To achieve the core temperature The corrected electrolyte phase diffusion coefficient, Pre-exponential factor, Activation energy of liquid-phase lithium diffusion coefficient This is the pseudo-glass transition temperature of the electrolyte system.

[0069] The electrochemical model adopts a single-particle model with electrolyte dynamics (SPMe), which includes solid-phase lithium-ion mass conservation, liquid-phase lithium-ion mass conservation, solid-phase charge conservation, liquid-phase charge conservation, and electrode reaction kinetics.

[0070] The mass conservation of lithium ions in solid phases is expressed by the diffusion equations of lithium ions in solid phases at both positive and negative electrodes, as follows:

[0071]

[0072] in, The concentration of lithium ions in the solid phase of the positive and negative electrodes. The coordinates are the radius coordinates of the spherical particles.

[0073] The relationship between the average lithium-ion concentration in the solid phase and the surface lithium-ion concentration of the electrode can be obtained using the partial differential equation of solid-phase diffusion and the polynomial order reduction method. After obtaining the surface lithium-ion concentration of the electrode solid phase, the average lithium-ion concentration can be calculated using this relationship, and thus the state of charge (SOC) can be further calculated. The relationship is as follows:

[0074]

[0075]

[0076] in, The concentration of lithium ions on the solid surface of the positive and negative electrodes. This represents the average lithium-ion concentration in the solid phases of both the positive and negative electrodes. The radius of the spherical particle is The cross-sectional area of ​​the electrode. The surface area of ​​the spherical particles is denoted as ... denoted as denoted denoted as denoted denoted as denoted denoted as denoted The thickness of the positive and negative electrodes. This represents the volume fraction of the solid-phase active material. is Faraday's constant.

[0077] The formula for calculating SOC is:

[0078]

[0079] in, The average lithium concentration in the solid phase of the cathode at 100% SOC. The average lithium concentration in the solid phase of the cathode at 0% SOC. The average lithium concentration in the solid phase of the negative electrode at 100% SOC. The average lithium concentration in the solid phase of the negative electrode at 0% SOC.

[0080] The mass conservation of lithium ions in the liquid phase is expressed by the diffusion equations of lithium ions in the positive and negative electrodes and the separator, respectively:

[0081]

[0082]

[0083] in, The concentrations of lithium ions in the liquid phase at the positive and negative electrodes are denoted as . For position coordinates, The integral of the positive and negative electrode liquids in electrolysis. For migration number, This refers to the lithium-ion concentration in the liquid phase of the diaphragm. This represents the integral number of the diaphragm electrolyzer liquid.

[0084] Using the parabolic profile approximation method, an approximate solution for the liquid phase lithium concentration distribution can be obtained. This approximate solution can then be used to determine the liquid phase lithium concentrations at the positive and negative electrodes and the separator, as shown in the following formulas:

[0085]

[0086]

[0087] in, , , and As an intermediate variable, The negative electrode time constant, The positive electrode time constant, The integral number of the negative electrode electrolyzing liquid. This represents the integral number of the positive electrode electrolytic liquid. This represents the integral number of the diaphragm electrolyzer liquid. The thickness of the negative electrode. The thickness of the positive electrode. For the thickness of the diaphragm electrode, This represents the initial concentration of lithium ions in the liquid phase.

[0088] The conservation of charge in the liquid phase is reflected in the calculation of the liquid phase potential difference, as shown by the formula:

[0089]

[0090] in, This is the positive liquid phase potential. This represents the negative electrode liquid phase potential. The ionic conductivity of the negative electrode material. The ionic conductivity of the cathode material. The ionic conductivity of the membrane material. is the ideal gas constant.

[0091] The overpotentials at the positive and negative electrodes can be calculated using the Butler-Volmer electrode reaction kinetic equations. The calculation equations are as follows:

[0092]

[0093]

[0094]

[0095] in, The current density at the solid-liquid interface between the positive and negative electrodes. The positive and negative electrode exchange current density, This represents the overpotential between the positive and negative poles. These are the rate constants for the positive and negative electrode reactions. This represents the maximum lithium-ion concentration in the solid phase of both the positive and negative electrodes.

[0096] The solid-state potentials of the positive and negative electrodes can be obtained from the conservation of solid-state charge. Here, the calculation of the solid-state potentials of the positive and negative electrodes is decomposed into the sum of all polarizations:

[0097]

[0098] in, The positive and negative electrode solid-phase potentials These represent the positive and negative electrode liquid phase potentials. The open-circuit potentials are the positive and negative terminals.

[0099] The equation for calculating the terminal voltage is:

[0100]

[0101] in, This is a positive overpotential. This is the negative overpotential. The solid-electrolyte interphase (SEI) film resistance is the positive electrode resistance. The negative electrode SEI film resistor, This represents the current density at the solid-liquid interface of the positive electrode. This represents the current density at the negative electrode solid-liquid interface.

[0102] (3) Model parameter identification and optimization

[0103] The Adaptive Particle Swarm Optimization (APSO) algorithm is used for group identification and optimization, based on the thermal model parameters ( ), battery geometric parameters ( ) and electrochemical model parameters ( The algorithm groups components and sequentially identifies and optimizes them with the objectives of minimizing core temperature error, solid-phase lithium concentration error, and voltage error. This method systematically solves the problem of numerous parameters and difficult identification in coupled models, providing a high-precision basic model for the fusion framework. APSO, based on the classic particle swarm optimization (PSO) algorithm, introduces an adaptive mechanism to dynamically adjust algorithm parameters, improving convergence speed and optimization accuracy.

[0104] During the initialization phase, 30 particles containing position and velocity vectors are generated. The position of the first particle is initialized using an initial guess, while the positions of the remaining particles and the velocities of all particles are randomly generated within their respective domains.

[0105] During the iteration process, the formula for the particle update rate is:

[0106]

[0107] in, For particle velocity, This represents the current iteration number. For inertial weights, , As a learning factor, , A random number in the range [0, 1]. This is the optimal position in the particle's history. To be the globally optimal position The position of the particle.

[0108] After the velocity is updated, the particle's position is updated accordingly:

[0109]

[0110] In this process, the inertia weight adopts an adaptive adjustment strategy, decreasing linearly from its maximum value to its minimum value as the iteration progresses. Its calculation formula is as follows:

[0111]

[0112] in, For maximum inertia weight, For minimum inertia weight, This represents the maximum number of iterations.

[0113] When a particle's new position exceeds the preset parameter boundaries, the algorithm performs boundary processing, restricting its position value within the boundaries and resetting its velocity in the corresponding dimension to zero to prevent it from continuously moving into infeasible regions. Finally, the algorithm uses the objective function to calculate the fitness value of each particle's new position. Based on the calculation results, it updates the individual historical best position of each particle and the global best position of the entire population, thereby driving the entire particle swarm to continuously move towards better solution regions.

[0114] The model parameters that need optimization mainly include thermal model parameters, battery geometric parameters, and electrochemical model parameters, as shown in the flowchart below. Figure 3 As shown.

[0115] For thermal model parameter optimization, the root mean square error (RMSE) of the core temperature prediction is minimized. and .

[0116]

[0117] in, To predict voltage, This is the measured voltage.

[0118] The core temperature estimation results after parameter optimization (taking the No. 1 test condition at 25℃ as an example) and error analysis are as follows: Figure 4 As shown, the maximum error of the core temperature of the lithium battery is within 0.5℃, and the average absolute error of the core temperature calculation under all operating conditions is within 0.0670℃.

[0119] For geometric model parameter optimization, the root mean square error between the predicted solid-phase lithium concentration and the theoretical solid-phase lithium concentration is minimized to correct the error. and .

[0120]

[0121]

[0122] in, To predict the average lithium concentration in the solid phase of both the positive and negative electrodes, This represents the theoretical average concentration of lithium in the solid phase of both the positive and negative electrodes.

[0123] For the optimization of electrochemical model parameters, correction is achieved by minimizing the root mean square error of voltage prediction. , and .

[0124]

[0125] in, To predict voltage, This is the measured voltage.

[0126] The voltage calculation results after parameter optimization (taking the first test condition at 25℃ as an example) and error statistics are as follows: Figure 5 As shown, the maximum error of the lithium battery voltage is within 0.2V, and the average absolute error of the voltage calculation under all operating conditions is within 0.0590V.

[0127] (4) Construct a hybrid neural network framework (ETM-CNN-LSTM) that couples the electrochemical thermal model and train it using the physical information loss function;

[0128] The system structure of the electrochemical thermal model coupled with the hybrid neural network is as follows: Figure 6 As shown, a physical principle-based electrochemical thermal model (mechanism model) is coupled with a data-driven hybrid neural network (CNN-LSTM) for joint high-precision estimation of multiple states (voltage, core temperature, SOC) of lithium batteries. The hybrid neural network is responsible for extracting features from the input sequence (current, voltage) and outputting key intermediate physical quantities (lithium ion concentration on the solid surface of the positive and negative electrodes). The calibrated electrochemical thermal model receives these intermediate physical quantities and calculates the final state variables (voltage, SOC, core temperature) to be estimated based on the physical equations. The core of this approach is to use the physical model to ensure the physical consistency and generalization ability of the estimation results, while using the neural network to enhance the model's adaptability to complex dynamic conditions and temperature changes, especially improving the estimation accuracy under extreme conditions such as high and low temperatures.

[0129] A CNN-LSTM hybrid neural network model was chosen, which combines the advantages of Convolutional Neural Networks (CNNs) in local feature extraction with the advantages of Long Short-Term Memory Networks (LSTMs) in handling long-sequence temporal dependencies. The CNN-LSTM hybrid neural network model specifically includes a data input layer, a CNN feature extraction layer, an LSTM temporal modeling layer, and a fully connected output layer.

[0130] In the data input layer, time-series data containing current and voltage are input, and independent Z-score normalization is performed on the input data to improve training stability. The CNN feature extraction layer contains two one-dimensional convolutional layers, each followed by a ReLU activation function and a max-pooling layer with a pooling window of 2. By reshaping the data dimension, convolution, and data transposition, local, short-term pattern features (such as the influence of dynamic test condition current pulses) are extracted from multivariate time series. The LSTM time-series modeling layer contains 128 hidden units. Based on the local features extracted by the CNN, long-term dynamic evolution of lithium concentration is captured through forget gate, input gate, and output gate mechanisms, and a dropout rate of 0.2 is applied to improve generalization ability. In the fully connected output layer, the final state of the LSTM is mapped to the output of two neurons, namely the lithium ion concentration on the positive and negative electrode solid surface. .

[0131] The output of the hybrid neural network model The data is transmitted to the corrected electrochemical thermal model, and the calculated voltage, state of charge (SOC), and core temperature are output. During the time-series calculation, the electrochemical parameters are corrected using the core temperature from the previous moment, and the neural network outputs... The input is then fed into the electrochemical model with corrected parameters, and the average lithium concentration of the positive and negative electrode solid phases is obtained through the relationship between the solid phase surface area and the average lithium concentration. This allows for the calculation of the state of charge (SOC) and open-circuit potential; on the other hand... Compared with the solution obtained based on the liquid phase lithium concentration distribution The overpotential is obtained through the Butler-Volmer electrode reaction kinetic equation, and the voltage is further calculated by combining the liquid phase potential, open circuit potential, and SEI film resistance voltage drop. The current, voltage, and open circuit potential are input into the thermal model to calculate the core temperature, and are prepared to be transferred to the electrochemical model parameter correction stage at the next moment to realize the calculation of SOC, voltage, and core temperature.

[0132] Establish a physical information loss function that includes the mean square error of voltage and SOC:

[0133]

[0134] in, To predict the SOC value, This represents the measured SOC value. This loss function forces the neural network output (lithium ion concentration on the solid surface of the positive and negative electrodes) not only to fit the data but also to satisfy the physical laws of the electrochemical thermal model.

[0135] Using the No. 1 dynamic test dataset at 10℃ / 25℃ / 40℃ as the training set, and minimizing the physical information loss function as the training objective, the neural network learns data features and strictly follows the physical laws of batteries during training, ultimately forming a joint estimation model for lithium battery state that combines high accuracy and strong physical consistency.

[0136] (5) Validate the trained hybrid neural network model using dynamic test data under multiple working conditions.

[0137] Using dynamic test datasets 2 to 4 at 10℃ / 25℃ / 40℃ as the test set, the electrochemical thermal model coupled with a hybrid neural network model was tested. Simultaneously, an LSTM neural network model with the same hyperparameters (Model 1) and an LSTM model coupled with the electrochemical thermal model (Model 2) were established to compare the computational performance of the three models.

[0138] SOC estimation results (taking dynamic test No. 3 at 10℃ / 25℃ / 40℃ as an example) are as follows Figure 7 As shown in the table, the error statistics indicate that the maximum error of the lithium battery SOC calculated by the method of this invention is within 0.05, and the average absolute error of the SOC calculation under all operating conditions is 0.0083, which is 15.31% lower than the common LSTM method (Model 1) and 5.56% lower than the LSTM method coupled with an electrochemical-thermal model (Model 2). At a low temperature of 10℃, the average absolute error of the SOC calculation by the method of this invention is 21.12% lower than that of LSTM (Model 1); at a high temperature of 40℃, the average absolute error of the SOC calculation by the method of this invention is 19.80% lower than that of LSTM (Model 1) and 18.15% lower than that of the LSTM method coupled with an electrochemical-thermal model (Model 2). These results demonstrate that the method proposed in this invention improves calculation accuracy while exhibiting better adaptability to multiple operating conditions and a wide temperature range.

[0139]

[0140] Corresponding to the aforementioned embodiment of a method for jointly estimating the state of charge and core temperature of a lithium battery, the present invention also provides an embodiment of a device for jointly estimating the state of charge and core temperature of a lithium battery.

[0141] See Figure 8 The present invention provides a lithium battery state of charge and core temperature joint estimation device, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement a lithium battery state of charge and core temperature joint estimation method in the above embodiment.

[0142] The embodiment of the lithium battery state of charge and core temperature joint estimation device provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 8 The diagram shown is a hardware structure diagram of any device with data processing capabilities, where the lithium battery state of charge and core temperature joint estimation device provided by this invention is located. (Except for...) Figure 8 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0143] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0144] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0145] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a method for jointly estimating the state of charge and core temperature of a lithium battery as described in the above embodiments.

[0146] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0147] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method for jointly estimating the state of charge and core temperature of a lithium battery.

[0148] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A method for jointly estimating the state of charge and core temperature of a lithium battery, characterized in that, The method includes the following steps: (1) Perform dynamic testing on lithium batteries under multiple operating conditions to obtain experimental data sets; (2) Establish an electrochemical thermal model, calculate the core temperature of the lithium battery, and correct the reaction rate constant and diffusion coefficient; based on the conservation of lithium ion mass in the solid phase and the conservation of lithium ion mass in the liquid phase, obtain the relationship between the average lithium ion concentration in the solid phase of the electrode and the lithium ion concentration on the surface of the solid phase of the positive and negative electrodes, as well as the lithium concentration distribution in the liquid phase; then calculate the SOC and obtain the liquid phase potential difference based on the liquid phase lithium concentration distribution, and then calculate the lithium battery voltage. (3) An adaptive particle swarm optimization algorithm combined with an adaptive adjustment strategy for inertial weights is used to optimize the electrochemical thermal model parameters, battery geometric parameters, and electrochemical model parameters; (4) Couple the electrochemical thermal model with the neural network model, input the time series data containing current and voltage, and finally output the lithium ion concentration on the solid surface of the positive and negative electrodes; train the model based on the mean square error physical information loss function containing voltage and SOC, and transmit the model output after training to the electrochemical thermal model with corrected parameters to output the calculated results of voltage, SOC and core temperature.

2. The method for jointly estimating the state of charge and core temperature of a lithium battery according to claim 1, characterized in that, Step (1) specifically involves screening lithium batteries through standard battery charge and discharge experiments, conducting thermocouple embedding experiments on the screened lithium batteries, and performing multi-condition dynamic testing experiments on the lithium batteries with embedded thermocouples to obtain the voltage, current, surface temperature, core temperature and SOC of the lithium batteries in the multi-condition dynamic test.

3. The method for jointly estimating the state of charge and core temperature of a lithium battery according to claim 1, characterized in that, In step (2), the thermal model uses a second-order equivalent thermal resistance model to describe the heat transfer inside the battery, establishes the core thermal balance equation of the battery and discretizes it, and combines the internal heat generation power formula to establish the relationship between the lithium battery voltage, current, average lithium concentration of the positive and negative electrode solid phases and temperature.

4. The method for jointly estimating the state of charge and core temperature of a lithium battery according to claim 1, characterized in that, In step (2), the electrochemical model adopts a single-particle model that considers electrolyte kinetics, including solid-phase lithium ion mass conservation, liquid-phase lithium ion mass conservation, solid-phase charge conservation, liquid-phase charge conservation, and electrode reaction kinetics.

5. The method for jointly estimating the state of charge and core temperature of a lithium battery according to claim 1, characterized in that, In step (2), the positive and negative electrode reaction rate constants and the solid phase diffusion coefficient are corrected using the Arrhenius law, and the liquid phase diffusion coefficient is corrected using the VTF equation.

6. The method for jointly estimating the state of charge and core temperature of a lithium battery according to claim 1, characterized in that, For thermal model parameter optimization, the core heat capacity and core-to-surface thermal resistance are corrected by minimizing the root mean square error of the core temperature prediction; for geometric model parameter optimization, the core heat capacity and core-to-surface thermal resistance are corrected by minimizing the root mean square error between the predicted solid-phase lithium concentration and the theoretical solid-phase lithium concentration; for electrochemical model parameter optimization, the initial solid-phase diffusion coefficient, positive and negative electrode SEI film resistance, and initial positive and negative electrode reaction rate constants are corrected by minimizing the root mean square error of the voltage prediction.

7. The method for jointly estimating the state of charge and core temperature of a lithium battery according to claim 1, characterized in that, In step (4), the neural network model is a CNN-LSTM hybrid neural network model, which extracts local and short-term pattern features based on CNN and captures long-term dynamic evolution of lithium concentration by LSTM.

8. A joint estimation device for the state of charge and core temperature of a lithium battery, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements a method for jointly estimating the state of charge and core temperature of a lithium battery as described in any one of claims 1-7.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for jointly estimating the state of charge and core temperature of a lithium battery as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a method for jointly estimating the state of charge and core temperature of a lithium battery as described in any one of claims 1-7.

Images