A method and device for estimating the state of charge of a vehicle battery throughout its life cycle

Abstract

The application provides a vehicle battery full life cycle state of charge estimation method and device, through collecting the current charging and discharging working condition and actual operation data of the vehicle battery, using the particle swarm method to determine and dynamically update the discretization second-order equivalent circuit model parameters, combining the discrete state equation and the linearized observation equation, through the intelligent control adaptive extended Kalman filter algorithm start-stop and combining the ampere-hour integral method bottom-up method to determine the state of charge estimation value of the vehicle battery, solves the technical defects of poor adaptability, narrow working condition adaptation range, low estimation accuracy and error accumulation in the existing vehicle battery state of charge estimation method, realizes the high-precision real-time estimation of the state of charge of the battery in the full life cycle and full working condition, reduces the calculation load of the vehicle-mounted battery management system, and improves the algorithm practicality and landing nature.

Description

Technical Field

[0001] This application relates to the field of vehicle battery management technology, and in particular to a method and apparatus for estimating the state of charge of a vehicle battery throughout its entire life cycle. Background Technology

[0002] Currently, most existing SOC estimation technologies for lithium iron phosphate batteries use fixed model parameters and do not dynamically adjust them based on battery aging state (SOH, which is the ratio of the battery's current performance to its brand-new state) and actual operating conditions. As the battery ages (SOH decreases), the deviation between the model parameters and the battery's actual electrochemical characteristics increases, resulting in a significant reduction in the accuracy of SOC estimation and failing to cover the needs of the entire battery life cycle.

[0003] Meanwhile, most existing technologies are based on HPPC (Hybrid Pulse Power Characterization) operating condition identification to obtain model parameters. These parameters are only applicable to discharge conditions and not to AC or DC charging conditions. Therefore, existing algorithms cannot be used to estimate SOC during battery charging. In addition, the SOC-OCV (Open Circuit Voltage, which refers to the stable voltage between the positive and negative electrodes after the battery is disconnected from the load) curve of lithium iron phosphate batteries has a clear plateau region. Currently, the SOC estimation algorithm is usually activated when the SOC is less than 20%. In actual user scenarios, the SOC estimation algorithm has a low trigger probability. When the algorithm cannot be triggered, the SOC can usually only be calculated by ampere-hour integration. At this time, the measurement error of the current sensor will cause the SOC estimation error to accumulate continuously, further affecting the estimation accuracy. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method and device for estimating the state of charge (SOC) of a vehicle battery throughout its entire life cycle. By collecting the current charging and discharging conditions and actual operating data of the vehicle battery, the particle swarm optimization method is used to determine and dynamically update the parameters of the discretized second-order equivalent circuit model. Combining the discrete state equation and the linearized observation equation, the estimated SOC of the vehicle battery is determined by intelligent control of the adaptive extended Kalman filter algorithm and a fallback method of ampere-hour integration. This solves the technical defects of existing vehicle battery SOC estimation methods, such as poor adaptability throughout the entire life cycle, narrow operating condition adaptability, low estimation accuracy, and easy accumulation of errors. It achieves high-precision real-time estimation of the SOC of the battery throughout its entire life cycle and under all operating conditions, while reducing the computational load of the on-board battery management system and improving the practicality and applicability of the algorithm.

[0005] This application provides a method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle. The method includes: In response to obtaining the charging and discharging conditions of the target vehicle battery in the current sampling period, the actual operating data of the target vehicle battery under the charging and discharging conditions is collected, and the actual terminal voltage data of the target vehicle battery is determined based on the actual operating data. Based on the actual terminal voltage data, the test state data of the target vehicle battery under the charge and discharge conditions, and the preset discretized second-order equivalent circuit model, the initial model parameters of the target vehicle battery under the charge and discharge conditions are determined using the preset particle swarm method. Based on the comparison between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and the preset voltage error threshold, the initial model parameters are updated using the particle swarm method to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. Based on the target model parameters, the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation are determined respectively. In response to the open-circuit voltage change corresponding to the historical state of charge estimation value determined in the previous sampling period satisfying the preset estimation start condition, based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, the target state of charge estimation value of the target vehicle battery in the current sampling period is determined using a preset adaptive extended Kalman filter method.

[0006] Furthermore, the charging and discharging conditions include at least AC charging conditions, DC charging conditions, and discharging conditions; the actual operating data includes at least actual current values, actual voltage values, and actual temperature values; and the test status data includes at least test current values, test voltage values, and test temperature values.

[0007] Furthermore, the step of determining the initial model parameters of the target vehicle battery under the charge-discharge condition based on the actual terminal voltage data, the test state data of the target vehicle battery under the charge-discharge condition, and the preset discretized second-order equivalent circuit model, using a preset particle swarm optimization method, includes: Based on the test state data of the target vehicle battery under the charging and discharging conditions, the preset particle swarm method is initialized to set particle parameter information; wherein, the particle parameter information includes particle weight, learning factor, iteration number, number of particles, particle position information and velocity information, and the particle position information is the model parameter to be identified corresponding to the preset discretized second-order equivalent circuit model; Based on the particle parameter information, the model terminal voltage data is determined using the discretized second-order equivalent circuit model, and the error between the model terminal voltage data and the actual terminal voltage data is used as the optimization target to construct an offline identification fitness function for battery parameters. The particle position information is substituted into the offline identification fitness function of the battery parameters to calculate the fitness value of each particle, and based on the fitness value of each particle, the individual extreme value and the group extreme value are determined respectively; wherein, the individual extreme value includes the position information and fitness value corresponding to the current particle, and the group extreme value includes the velocity information and position information corresponding to the particle with the target fitness value; Based on the individual extreme value and the group extreme value, the velocity information of the particles is iteratively updated using a preset particle velocity update formula, and the particle position information is iteratively updated using a preset position update formula. The fitness value is recalculated based on the updated velocity and position information of the particles, and the individual extreme value and the group extreme value are updated to determine whether the preset number of iterations has been reached. If the preset number of iterations is reached, the iteration stops and the initial model parameters of the target vehicle battery under the charging and discharging conditions are output; wherein, the initial model parameters include at least the initial ohmic internal resistance value, the initial polarization internal resistance value, and the initial polarization capacitance value.

[0008] Furthermore, the step of updating the initial model parameters using the particle swarm optimization method based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold, to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging condition, includes: Based on the discretized second-order equivalent circuit model and the test current value in the test state data, determine the model terminal voltage data; The mean absolute error between the actual terminal voltage data and the model terminal voltage data is determined using a preset voltage loss function. The average absolute error value is compared with a preset voltage error threshold to obtain the error comparison result; If the mean absolute error value is less than the preset voltage error threshold, then the initial model parameters are determined as the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. If the mean absolute error value is greater than or equal to the preset voltage error threshold, the particle swarm optimization method is triggered to identify and update the initial model parameters online to obtain the target model parameters.

[0009] Furthermore, the discrete state equations are pre-defined through the following steps: The preset battery state of charge calculation formula is discretized using the ampere-hour integral method to obtain the initial discrete state equation, and the system state variables are defined as battery state of charge, low-order polarization voltage and high-order polarization voltage. Based on the discretized second-order equivalent circuit model, the first coefficient matrix and the second coefficient matrix are constructed respectively. The discrete state equation is determined based on the system state variables, the first coefficient matrix, the second coefficient matrix, and the experimental state data, with the process noise covariance matrix introduced.

[0010] Furthermore, the linearized observation equation is pre-defined through the following steps: Based on the discretized second-order equivalent circuit model, a nonlinear observation equation for the terminal voltage is established. A first-order Taylor expansion of the open-circuit voltage in the terminal voltage nonlinear observation equation is performed with respect to the state of charge to linearize the terminal voltage nonlinear observation equation, resulting in an initial linearized observation equation. Taking the partial derivative of the initial linearized observation equation, we obtain the Jacobian matrix and the observation matrix; Based on the Jacobian matrix, the observation matrix, and the system state variables, and by configuring the corresponding measurement noise covariance matrix, the linearized observation equation is constructed.

[0011] Furthermore, the following steps are used to determine whether the change in open-circuit voltage corresponding to the historical state of charge estimation value determined in the previous sampling period meets the preset estimation start condition: Based on the actual terminal voltage data and the target model parameters, determine the open-circuit voltage corresponding to the current sampling period; Based on the historical open-circuit voltage value corresponding to the historical state of charge estimation value determined in the previous sampling period and the open-circuit voltage, the difference between adjacent open-circuit voltages is calculated to obtain the open-circuit voltage change, and the open-circuit voltage change corresponding to a preset number of time windows is continuously acquired. When the open-circuit voltage change corresponding to the preset number of time windows is greater than or equal to the preset open-circuit voltage threshold, it is determined that the preset estimated opening condition is met. If the change in open-circuit voltage is less than the preset open-circuit voltage threshold for any time window, it is determined that the estimated opening condition is not met.

[0012] Furthermore, the method also includes: In response to the open-circuit voltage change not meeting the estimated start-up conditions, the target state of charge estimate is determined based on the actual operating data using a preset ampere-hour integration method.

[0013] Furthermore, the step of determining the target state of charge estimate of the target vehicle battery in the current sampling period based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, using a preset adaptive extended Kalman filter method, includes: Based on the actual current value in the actual operating data, the target model parameters and the historical state vector corresponding to the previous iteration cycle, the prior state estimation is performed using the discrete state equation to obtain the prior state vector corresponding to the current iteration cycle. Based on the state equation parameters and the historical error covariance matrix corresponding to the previous iteration cycle, determine the prior error covariance matrix corresponding to the current iteration cycle. Based on the observation matrix in the observation equation parameters, the ohmic internal resistance value in the target model parameters, the actual current value, and the prior state vector, the model prediction terminal voltage is determined using the linearized observation equation, and the difference between the actual voltage value in the actual operating data and the model prediction terminal voltage is determined as the innovation error. Based on the prior error covariance matrix, the measurement noise covariance matrix in the observation equation parameters, and the observation matrix, the Kalman gain matrix is ​​determined, and based on the innovation error, the measurement noise covariance matrix and the process noise covariance matrix in the state equation parameters are adaptively updated. The prior state vector is corrected a posteriori by using the Kalman gain matrix and the innovation error to obtain the target state vector corresponding to the current iteration period, and the target state of charge estimate of the target vehicle battery in the current sampling period is determined in the target state vector. Based on the Kalman gain matrix and the observation matrix, the prior error covariance matrix is ​​updated to obtain the posterior error covariance matrix, which is then used to calculate the next iteration cycle.

[0014] This application embodiment also provides a state of charge estimation device for the entire life cycle of a vehicle battery, the state of charge estimation device comprising: The data acquisition module is used to respond to the acquisition of the charging and discharging conditions of the target vehicle battery in the current sampling period, collect the actual operating data of the target vehicle battery under the charging and discharging conditions, and determine the actual terminal voltage data of the target vehicle battery based on the actual operating data. The model parameter module is used to determine the initial model parameters of the target vehicle battery under the charging and discharging conditions based on the actual terminal voltage data, the test state data of the target vehicle battery under the charging and discharging conditions, and the preset discretized second-order equivalent circuit model, using a preset particle swarm optimization method. The parameter update module is used to update the initial model parameters using the particle swarm optimization method based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold, so as to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. The equation parameter module is used to determine the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation based on the target model parameters. The data estimation module is used to determine the target state of charge (SBC) estimate of the target vehicle battery in the current sampling period based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, using a preset adaptive extended Kalman filter method.

[0015] This application embodiment also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the above-described method for estimating the state of charge of a vehicle battery throughout its entire life cycle are performed.

[0016] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described method for estimating the state of charge of a vehicle battery throughout its entire life cycle.

[0017] The present application provides a method and apparatus for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle. The method includes: in response to acquiring the charge / discharge condition of a target vehicle battery in the current sampling period, collecting actual operating data of the target vehicle battery under the charge / discharge condition, and determining the actual terminal voltage data of the target vehicle battery based on the actual operating data; determining the initial model parameters corresponding to the target vehicle battery under the charge / discharge condition using a preset particle swarm optimization method based on the actual terminal voltage data, the test state data of the target vehicle battery under the charge / discharge condition, and a preset discretized second-order equivalent circuit model; and determining the initial model parameters corresponding to the target vehicle battery under the charge / discharge condition based on the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold. The comparison results are used to update the initial model parameters using the particle swarm optimization method to obtain the target model parameters of the target vehicle battery under the charging and discharging conditions. Based on the target model parameters, the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation are determined respectively. In response to the open-circuit voltage change corresponding to the historical state of charge estimation value determined in the previous sampling period satisfying the preset estimation start condition, the target state of charge estimation value of the target vehicle battery in the current sampling period is determined using the preset adaptive extended Kalman filter method based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation.

[0018] Compared with existing technologies that use fixed model parameters and model parameters obtained by identifying operating conditions based on hybrid pulse power characteristics, this method collects the current charging and discharging conditions and actual operating data of the vehicle battery, uses the particle swarm optimization method to determine and dynamically update the parameters of the discretized second-order equivalent circuit model, and combines discrete state equations and linearized observation equations. It uses intelligent control to start and stop the adaptive extended Kalman filter algorithm and combines the ampere-hour integration method as a fallback to determine the estimated value of the vehicle battery's state of charge. This method solves the technical defects of existing vehicle battery state of charge estimation methods, such as poor adaptability throughout the entire life cycle, narrow operating condition adaptation range, low estimation accuracy, and easy accumulation of errors. It achieves high-precision real-time estimation of the state of charge of the battery throughout its entire life cycle and under all operating conditions, while reducing the computational load of the vehicle battery management system and improving the practicality and feasibility of the algorithm.

[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is one of the flowcharts for a method to estimate the state of charge of a vehicle battery throughout its entire life cycle, provided in an embodiment of this application. Figure 2 A second flowchart illustrating a method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle, as provided in this application embodiment; Figure 3 This is one of the structural schematic diagrams of a vehicle battery state-of-charge estimation device for the entire life cycle provided in an embodiment of this application; Figure 4 This is a second schematic diagram of a device for estimating the state of charge of a vehicle battery throughout its entire life cycle, provided in an embodiment of this application. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0023] Research has shown that with the rapid development of the new energy vehicle industry, lithium iron phosphate batteries have become the mainstream choice for new energy vehicle power systems due to their advantages such as high safety, long cycle life, low cost, and good thermal stability. State of Charge (SOC) refers to the ratio of the battery's current remaining charge to its rated charge. It is a core parameter of the Battery Management System (BMS), and its estimation accuracy directly determines the accuracy of the predicted driving range, power performance, and battery life of new energy vehicles. It is a fundamental guarantee for the safe and reliable operation of new energy vehicles.

[0024] Currently, most existing SOC estimation technologies for lithium iron phosphate batteries use fixed model parameters and do not dynamically adjust them based on battery aging state (SOH, which is the ratio of the battery's current performance to its brand-new state) and actual operating conditions. As the battery ages (SOH decreases), the deviation between the model parameters and the battery's actual electrochemical characteristics increases, resulting in a significant reduction in the accuracy of SOC estimation and failing to cover the needs of the entire battery life cycle.

[0025] Furthermore, most existing technologies are based on HPPC (Hybrid Pulse Power Characterization) operating condition identification to obtain model parameters. These parameters are only applicable to discharge conditions and not to AC or DC charging conditions. Therefore, existing algorithms cannot be used for SOC estimation during battery charging. In addition, the SOC-OCV (Open Circuit Voltage, which refers to the stable voltage between the positive and negative electrodes after the battery is disconnected from the load) curve of lithium iron phosphate batteries has a clear plateau region. Currently, the SOC estimation algorithm is usually activated when the SOC is less than 20%. In actual user scenarios, the SOC estimation algorithm has a low trigger probability. When the algorithm cannot be triggered, the SOC can usually only be calculated by ampere-hour integration. At this time, the measurement error of the current sensor will cause the SOC estimation error to accumulate continuously, further affecting the estimation accuracy.

[0026] Based on this, this application provides a method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle. By collecting the current charging and discharging conditions and actual operating data of the vehicle battery, the method uses particle swarm optimization to determine and dynamically update the parameters of the discretized second-order equivalent circuit model. Combining the discrete state equation and the linearized observation equation, the method uses intelligent control to start and stop the adaptive extended Kalman filter algorithm and combines the ampere-hour integral method as a fallback to determine the estimated SOC of the vehicle battery. This method solves the technical defects of existing vehicle battery SOC estimation methods, such as poor adaptability throughout the entire lifecycle, narrow operating condition adaptability, low estimation accuracy, and easy accumulation of errors. It achieves high-precision real-time estimation of the SOC of the battery throughout its entire lifecycle and under all operating conditions, while reducing the computational load of the vehicle battery management system and improving the practicality and feasibility of the algorithm.

[0027] Please see Figure 1 , Figure 1 This is one of the flowcharts for a method to estimate the state of charge (SOC) of a vehicle battery throughout its entire lifecycle, as provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle includes: S101. In response to obtaining the charging and discharging conditions of the target vehicle battery in the current sampling period, collect the actual operating data of the target vehicle battery under the charging and discharging conditions, and determine the actual terminal voltage data of the target vehicle battery based on the actual operating data.

[0028] In this embodiment of the application, the charging and discharging conditions include at least AC charging conditions, DC charging conditions, and discharging conditions; the actual operating data includes at least actual current values, actual voltage values, and actual temperature values.

[0029] Here, charging and discharging conditions refer to the current operating state of the target vehicle battery, including three core operating conditions: AC charging, DC charging, and discharging; sampling period refers to the fixed time interval for the vehicle battery management system to collect battery data (usually 10ms to 100ms, 50ms is used in this embodiment); actual operating data refers to the battery operating data collected in real time by the current, voltage, and temperature sensors of the vehicle BMS, including at least the actual current value, actual voltage value, and actual temperature value.

[0030] In this embodiment, AC charging mode refers to the operating state of the target vehicle being charged through an AC charging pile, with a charging current typically ranging from 10A to 30A; DC charging mode refers to the operating state of the target vehicle being fast-charged through a DC charging pile, with a charging current typically ranging from 50A to 150A; and discharging mode refers to the operating state of the target vehicle when it is driving normally, with the battery outputting electrical energy, and the discharging current typically ranging from -20A to -100A (in this embodiment, the discharging current is negative and the charging current is positive).

[0031] For example, the actual current value is collected by an on-board current sensor with an accuracy of ±0.1A and a range of -200A to 200A; the actual voltage value is collected by an on-board voltage sensor with an accuracy of ±5mV and a range of 2.5V to 4.2V; the actual temperature value is collected by a temperature sensor built into the battery pack with an accuracy of ±0.5℃ and a range of -20℃ to 60℃. In this embodiment, the actual operating data collection period is 50ms, and filtering and noise reduction are performed every 10 data points to remove abnormal values ​​(such as voltage jumps exceeding 0.5V).

[0032] Among them, the actual terminal voltage data is the actual voltage value in the actual operating data. The effective voltage data obtained after filtering and noise reduction is used for subsequent model calculation and error comparison.

[0033] In this way, acquiring the battery's current operating condition and real-time operating data provides a real and effective data foundation for subsequent model parameter identification and SOC estimation, ensuring the relevance and accuracy of subsequent steps; it enables real-time identification of operating conditions and accurate data collection, eliminates abnormal data interference, ensures data validity, and provides reliable support for subsequent steps.

[0034] S102. Based on the actual terminal voltage data, the test state data of the target vehicle battery under the charging and discharging conditions, and the preset discretized second-order equivalent circuit model, the initial model parameters corresponding to the target vehicle battery under the charging and discharging conditions are determined using the preset particle swarm optimization method.

[0035] It should be noted that the discretized second-order equivalent circuit model is a mathematical model built based on the electrochemical characteristics of the battery. It is used to describe the relationship between the battery terminal voltage, current and SOC, polarization voltage. It is adapted to the digital calculation of the vehicle embedded system through discretization and includes parameters such as ohmic internal resistance, low-order polarization internal resistance / capacitance, and high-order polarization internal resistance / capacitance. Particle Swarm Optimization (PSO) is an intelligent optimization algorithm that finds the optimal solution by simulating the flight and cooperation of particles in space. It is used for the identification and updating of model parameters in this invention.

[0036] Here, experimental state data refers to the battery current, voltage, and temperature data collected by experimental equipment under three operating conditions—AC charging, DC charging, and discharging—for the target vehicle battery in a laboratory environment. This data is used for the initialization of the particle swarm optimization method and the construction of the fitness function.

[0037] For example, the test state data was collected through a laboratory power battery testing system. The test environment temperature was 25℃ (standard environment). Tests were conducted for three operating conditions: AC charging condition test current was 20A, charging time was 2h; DC charging condition test current was 100A, charging time was 30min; and discharging condition test current was -50A, discharging to 0% SOC. The acquisition frequency was 10ms, and at least 10,000 sets of test current values, test voltage values, and test temperature values ​​were collected for each operating condition as the basic data for offline identification using the particle swarm optimization method.

[0038] In this embodiment, the preset discretized second-order equivalent circuit model is a discretized mathematical model constructed based on the second-order RC equivalent circuit of the battery (including ohmic internal resistance R0, low-order polarization internal resistance R1, low-order polarization capacitor C1, high-order polarization internal resistance R2, and high-order polarization capacitor C2), and its discretization equation is shown below.

[0039] ; ; .

[0040] in, The sampling period is , These represent the low-order and high-order polarization voltages at time k, respectively. Let k be the actual current value at time k. Calculate the terminal voltage of the model at time k. Let be the open-circuit voltage of SOC at time k.

[0041] Here, the discretized second-order equivalent circuit model is obtained by discretizing the second-order RC equivalent circuit model of the battery (the second-order Thevenin equivalent circuit equation of the power battery).

[0042] In this embodiment, the particle swarm optimization method finds the model parameters that minimize the error between the calculated terminal voltage and the actual terminal voltage through iterative optimization, and uses these as the initial model parameters (parameters in the battery's brand-new state).

[0043] In this way, the initial model parameters corresponding to the battery in its brand-new (BOL) state under the current operating conditions are determined, which solves the problem of fixed model parameters and lack of adaptation to operating conditions in the prior art, and provides an initial benchmark for subsequent parameter updates and SOC estimation. The initial model parameters adapted to the current operating conditions are obtained to ensure the matching between the model and the battery in its brand-new state and the current operating conditions, laying the foundation for subsequent full life cycle parameter adaptation and high-precision SOC estimation.

[0044] In one possible implementation of this application, step S102 may include: S1021. Based on the test state data of the target vehicle battery under the charging and discharging conditions, initialize the preset particle swarm method to set particle parameter information.

[0045] The particle parameter information includes particle weight, learning factor, iteration number, number of particles, particle position information, and velocity information. The particle position information is the model parameter to be identified corresponding to the preset discretized second-order equivalent circuit model.

[0046] In this embodiment, the initialization of the particle swarm optimization method is to set the basic parameters of the algorithm to ensure the efficiency and accuracy of iterative optimization.

[0047] Among them, particle weight is used to adjust the global and local search capabilities of particles; learning factor is used to adjust the speed at which particles learn towards individual and group extrema; iteration count is the criterion for the algorithm to stop optimization; number of particles is the total number of particles participating in optimization; particle position information corresponds to the model parameters to be identified (R0, R1, C1, R2, C2), and particle velocity information corresponds to the update step size of particle position; the model parameters to be identified are the core parameters in the discretized second-order equivalent circuit model, which are the key to affecting the accuracy of the model.

[0048] For example, based on the experimental state data of AC charging, the particle swarm optimization method is initialized with the following specific parameters: particle weight ω = 0.7, learning factor c1 = 1.49, learning factor c2 = 1.49, number of iterations = 100, number of particles = 50; the particle position information (parameters to be identified) ranges as follows: R0 ∈ [0.005Ω, 0.02Ω], R1 ∈ [0.01Ω, 0.03Ω], C1 ∈ [500F, 2000F], R2 ∈ [0.008Ω, 0.025Ω], C2 ∈ [1000F, 3000F]; the initial particle velocity ranges as [-0.001, 0.001] (corresponding to resistance parameters) and [-100, 100] (corresponding to capacitance parameters).

[0049] S1022. Based on the particle parameter information, the model terminal voltage data is determined using the discretized second-order equivalent circuit model, and the error between the model terminal voltage data and the actual terminal voltage data is used as the optimization target to construct an offline identification fitness function for battery parameters.

[0050] In this step, the model terminal voltage data is calculated by substituting the model parameters corresponding to the particle position into the discretized second-order equivalent circuit model and combining the experimental current value in the experimental state data; the fitness function is the optimization objective of the particle swarm method, used to measure the quality of the particle position (model parameters). The smaller the error, the better the fitness value, thereby guiding the particles to seek optimization in the direction of the optimal parameters.

[0051] In this embodiment of the application, the expression for the offline identification fitness function of battery parameters is as follows.

[0052] .

[0053] in, Indicates fitness. This represents the voltage data at the model terminals. This represents the actual terminal voltage data. Indicates the number of iterations.

[0054] Here, the smaller the error term in the denominator of the fitness function for offline identification of battery parameters, the larger the value of the fitness function, and the better the particle.

[0055] S1023. Substitute the particle position information into the offline identification fitness function of the battery parameters, calculate the fitness value of each particle, and determine the individual extreme value and the group extreme value based on the fitness value of each particle.

[0056] The individual extreme value includes the position information and fitness value of the current particle, and the group extreme value includes the velocity information and position information of the particle corresponding to the target fitness value.

[0057] In this embodiment, the individual extreme value is the position and corresponding fitness value of each particle during its own iteration process; the group extreme value is the position, velocity and fitness value of the particle with the best fitness value among all particles; by determining the individual extreme value and the group extreme value, the particles are guided to continuously optimize their own positions and move closer to the global optimal solution.

[0058] For example, suppose the position information of 50 particles is substituted into the fitness function to calculate the fitness value of each particle; suppose the fitness value of particle 3 is 0.98 (maximum), then the position information of particle 3 (R0=0.012Ω, R1=0.02Ω, C1=1200F, R2=0.015Ω, C2=2000F) and fitness value of 0.98 are the population extreme values, and at the same time, the position and fitness value of this particle are also its own individual extreme values; the individual extreme values ​​of other particles are the position and fitness value with the maximum current fitness value of each particle.

[0059] S1024. Based on the individual extreme value and the group extreme value, the velocity information of the particles is iteratively updated using a preset particle velocity update formula, and the particle position information is iteratively updated using a preset position update formula.

[0060] In this embodiment, the particle velocity update formula is as follows.

[0061] .

[0062] In this embodiment of the application, the position update formula is as follows.

[0063] .

[0064] in, This represents the number of iterations. and For learning factors; and A random number between 0 and 1; For particle weights; For speed information; For particle position information; For individual extreme value locations; This represents the location of the group's extreme value.

[0065] Here, the particle velocity update formula is used to adjust the particle's flight velocity based on the individual extreme value and the group extreme value, so that the particle moves to a better position; the particle position update formula is used to adjust the particle's position based on the updated velocity, so as to achieve iterative optimization; the two formulas work together to ensure that the particle can quickly converge to the optimal solution.

[0066] S1025. Recalculate the fitness value based on the updated velocity information and position information of the particles, update the individual extreme value and the group extreme value, and determine whether the preset number of iterations has been reached.

[0067] In this step, after the particle position is updated, its corresponding model parameters change, and the fitness value needs to be recalculated to determine whether it is better than the current individual extreme value and the group extreme value. If it is better, it is updated. At the same time, it is determined whether the current iteration number has reached the preset value. If it has not reached the preset value, it continues to iterate. If it has reached the preset value, it stops the optimization search to ensure that the algorithm converges to the optimal solution.

[0068] For example, the updated particle positions are substituted into the fitness function to recalculate the fitness value of each particle; assuming that the updated fitness value of particle 1 is 0.96, which is better than its initial fitness value of 0.92, the individual extreme value of particle 1 is updated to the current position and fitness value of 0.96; the population extreme value is still the fitness value of particle 3 of 0.98, which has not changed; the current iteration number is 1, which has not reached the preset 100 times, so the iterative update continues.

[0069] S1026. If the preset number of iterations is reached, stop the iteration and output the initial model parameters of the target vehicle battery under the charging and discharging conditions.

[0070] The initial model parameters include at least the initial ohmic internal resistance, the initial polarization internal resistance, and the initial polarization capacitance.

[0071] In this step, when the number of iterations reaches a preset value, the particle swarm optimization algorithm converges. At this time, the particle position corresponding to the swarm extremum is the optimal initial model parameter. The initial model parameter is the model parameter adapted to the current operating condition in the battery's brand-new state, including ohmic internal resistance, low-order / high-order polarization internal resistance, and low-order / high-order polarization capacitance, which is the basis for subsequent parameter updates.

[0072] S103. Based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and the preset voltage error threshold, the initial model parameters are updated using the particle swarm optimization method to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions.

[0073] It should be noted that the voltage loss function is used to quantify the error between the calculated terminal voltage and the measured terminal voltage, and to characterize the degree of matching between the model parameters and the actual characteristics of the battery.

[0074] Here, the model terminal voltage data refers to the battery terminal voltage calculated by discretizing the second-order equivalent circuit model and substituting the current model parameters and the actual current value; the error value refers to the deviation between the actual terminal voltage data and the model terminal voltage data, which is calculated by a preset voltage loss function.

[0075] In this embodiment, the preset voltage error threshold is a pre-calibrated model accuracy judgment threshold (set to 50mV in this embodiment) used to determine whether the current model parameters meet the accuracy requirements; if the error value exceeds the threshold, the parameters are re-identified by the particle swarm optimization method and updated to target model parameters that are adapted to the current battery aging state.

[0076] This approach dynamically monitors the accuracy of model parameters and updates them based on battery aging (SOH decrease), addressing the issue in existing technologies where fixed parameters lead to decreased estimation accuracy with battery aging. This ensures that model parameters consistently match the actual characteristics of the battery. It achieves adaptive updates of model parameters throughout the entire lifecycle, ensuring model accuracy remains within a preset range and guaranteeing high accuracy for subsequent SOC estimations. Simultaneously, it avoids meaningless and frequent parameter updates, reducing computational power consumption.

[0077] In one possible implementation of this application, step S103 may include: S1031. Based on the discretized second-order equivalent circuit model and the test current value in the test state data, determine the model terminal voltage data.

[0078] Here, the model terminal voltage data is calculated by discretizing the second-order equivalent circuit model, substituting the currently used model parameters (initial model parameters or parameters after the last update) and the test current value in the test state data, and is used to compare with the actual terminal voltage data to quantify the model error.

[0079] S1032. Determine the average absolute error between the actual terminal voltage data and the model terminal voltage data using a preset voltage loss function.

[0080] In this embodiment of the application, the expression for the voltage loss function is as follows.

[0081] .

[0082] in, This is the actual terminal voltage data. For model terminal voltage data, The mean absolute error value. The time period of a certain power-on driving cycle.

[0083] Here, the voltage loss function is used to calculate the average absolute error between the actual terminal voltage and the model terminal voltage, which characterizes the degree of matching between the model parameters and the actual characteristics of the battery. The smaller the error value, the higher the model accuracy. The average absolute error value is the average of the errors of all sampling points in a single power-on driving cycle, avoiding misjudgment caused by errors of a single sampling point.

[0084] S1033. The average absolute error value is compared with the preset voltage error threshold to obtain the error comparison result.

[0085] Here, the preset voltage error threshold is a pre-calibrated model accuracy qualification threshold, which is set to 50mV in this embodiment. It is used to determine whether the current model parameters meet the SOC estimation accuracy requirements. The error comparison results are divided into two types: the average absolute error value is less than the preset voltage error threshold (accuracy qualified) and the average absolute error value is greater than or equal to the preset voltage error threshold (accuracy failed).

[0086] For example, the calculated average absolute error value J=45mV is compared with the preset voltage error threshold of 50mV, and the error comparison result is "J<50mV"; if the battery ages and J=60mV, then the error comparison result is "J≥50mV".

[0087] S1034. If the average absolute error value is less than the preset voltage error threshold, then the initial model parameters are determined as the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions.

[0088] In this step, when the mean absolute error value is less than the preset voltage error threshold, it indicates that the current model parameters match the actual characteristics of the battery and the accuracy meets the requirements. No update is needed, and the initial model parameters are directly used as the target model parameters for subsequent SOC estimation.

[0089] S1035. If the mean absolute error value is greater than or equal to the preset voltage error threshold, the particle swarm method is triggered to identify and update the initial model parameters online to obtain the target model parameters.

[0090] In this step, when the mean absolute error value is greater than or equal to the preset voltage error threshold, it indicates that the current model parameters have become mismatched with the actual characteristics of the battery due to battery aging, and the accuracy cannot meet the requirements. It is necessary to trigger the particle swarm method for online identification and re-optimize to obtain target model parameters that are suitable for the current battery aging state. The online identification process is the same as the offline identification process, but the input data is the actual operating data of the current sampling period, rather than laboratory test data, to ensure that the parameters match the current operating conditions and aging state.

[0091] S104. Based on the target model parameters, determine the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation.

[0092] It should be noted that the discrete state equation describes the discrete mathematical equations of state variables such as battery SOC and polarization voltage as they change over time, and is the basis for the AEKF algorithm to predict the state. The linearized observation equation is constructed based on the second-order equivalent circuit model. The nonlinear equation is linearized through the first-order Taylor expansion, and the mapping relationship between the state variables and the measured terminal voltage is established. This is the basis for the AEKF algorithm to correct the state.

[0093] Here, the state equation parameters include a first coefficient matrix, a second coefficient matrix, and a process noise covariance matrix; the observation equation parameters include an observation matrix, a Jacobian matrix, and a measurement noise covariance matrix.

[0094] In this embodiment of the application, the expression for the discrete state equation is as follows.

[0095] .

[0096] in, For the quantity to be estimated, For process noise, The process noise covariance matrix (in the formula) express), This is the actual current value. This is the first coefficient matrix. This is the second coefficient matrix. This is the target state vector.

[0097] In this embodiment, the expression for the linearized observation equation is shown below.

[0098] .

[0099] in, This represents actual operational data; Represents the observation matrix; Indicates measurement noise; Represents the measurement noise covariance matrix (in the formula) express); This is a quantity to be estimated.

[0100] Here, the discrete state equation is used to describe the battery state variables (SOC, low-order polarization voltage U). p1 Higher-order polarization voltage U p2 The time-varying discretized equations have parameters including the first coefficient matrix A, the second coefficient matrix B, and the process noise covariance matrix Q. The linearized observation equations are linearized equations constructed based on the second-order equivalent circuit model and are used to establish the mapping relationship between the state variables and the measured terminal voltage. The parameters of the observation equations include the observation matrix H and the measurement noise covariance matrix R.

[0101] The target model parameters are used to determine the specific values ​​of coefficient matrices A and B and observation matrix H, ensuring that the equations match the current battery state and operating conditions.

[0102] This provides discrete state equations and linearized observation equations adapted to the current battery state and operating conditions for the subsequent execution of the AEKF algorithm, solving the problem that nonlinear models cannot be directly used for Kalman filtering, and ensuring the adaptability and computational accuracy of the AEKF algorithm. Obtaining state equations and observation equations adapted to the current target model parameters provides a precise mathematical basis for the state prediction and correction of the AEKF algorithm, improving the estimation accuracy and stability of the algorithm.

[0103] In one possible implementation of this application, the step of pre-setting the discrete state equation in step S104 may include: S1041. The preset battery state of charge calculation formula is discretized using the ampere-hour integral method to obtain the initial discrete state equation, and the system state variables are defined as battery state of charge, low-order polarization voltage and high-order polarization voltage.

[0104] In this embodiment of the application, the formula for calculating the state of charge of the battery under the ampere-hour integration method is as follows.

[0105] .

[0106] in, and They represent State of charge at time 1 and any other time; Rated energy of the battery; Energy charging and discharging efficiency, typically 1; for The test current value at time t, where the charging current is defined as negative and the discharging current as positive.

[0107] In this embodiment of the application, the expression for the initial discrete state equation is as follows.

[0108] .

[0109] in, and They represent State of charge at time 1 and any other time; Rated energy of the battery; Energy charging and discharging efficiency, typically 1; for The test current value at time t.

[0110] In this embodiment of the application, the expression for the system state variable is as follows.

[0111] .

[0112] in, These are system state variables (quantities to be estimated). The battery is in its state of charge. For low-order polarization voltage, It is a higher-order polarization voltage.

[0113] S1042. Based on the discretized second-order equivalent circuit model, construct the first coefficient matrix and the second coefficient matrix respectively.

[0114] In this embodiment of the application, the expression of the first coefficient matrix is ​​as follows.

[0115] .

[0116] In this embodiment of the application, the expression for the second coefficient matrix is ​​as follows.

[0117] .

[0118] in, and These represent the first coefficient matrix and the second coefficient matrix, respectively.

[0119] Here, the first coefficient matrix is ​​used to describe the change law of the system state variables themselves, and the second coefficient matrix is ​​used to describe the influence of the input variable (current) on the system state variables; the elements of the two matrices are determined by the model parameters (R1, C1, R2, C2) of the discretized second-order equivalent circuit model to ensure that the matrices match the battery model.

[0120] S1043. Based on the system state variables, the first coefficient matrix, the second coefficient matrix, and the test state data, and by introducing the process noise covariance matrix, the discrete state equation is determined.

[0121] Here, the process noise covariance matrix Q is used to quantify the uncertainty of the system model (such as model error). It is an important parameter of the AEKF algorithm and is usually set as a diagonal matrix, with the diagonal elements being the process noise variance of each state variable. The discrete state equation is a mathematical description of the changes of the system state variables over time. Combining the system state variables, the A matrix, the B matrix, the input current, and the process noise, a complete state prediction equation is formed.

[0122] In this embodiment, the test current value is used as the input, and the polarization voltage value and state of charge value are used as the output, resulting in the following expression for the discrete state equation.

[0123] .

[0124] In one possible implementation of this application, in specific implementation, the step of pre-setting the linearized observation equation in step S104 may include: S1044. Based on the discretized second-order equivalent circuit model, establish the nonlinear observation equation for the terminal voltage.

[0125] Here, the nonlinear observation equation for the terminal voltage is established based on the voltage balance relationship of the discretized second-order equivalent circuit model. It describes the relationship between the battery terminal voltage and the system state variables, ohmic internal resistance, and current. Since the relationship between OCV and SOC is nonlinear, this equation is a nonlinear equation.

[0126] In this embodiment, the expression for the nonlinear observation equation of the terminal voltage is shown below.

[0127] U(k) = OCV(SOC(k)) - U p1 (k) - U p2 (k) - R0·I(k).

[0128] Where U(k) is the actual terminal voltage value at time k, and OCV(SOC(k)) is the open-circuit voltage corresponding to SOC at time k (obtained from the table by looking up the SOC-OCV curve).

[0129] S1045. Perform a first-order Taylor expansion of the open-circuit voltage in the terminal voltage nonlinear observation equation with respect to the state of charge to linearize the terminal voltage nonlinear observation equation and obtain the initial linearized observation equation.

[0130] Here, since the AEKF algorithm is only applicable to linear systems, it is necessary to linearize the nonlinear observation equations. The first-order Taylor expansion is a commonly used nonlinear linearization method. It expands the system state variables at the operating point (the optimal state estimate of the previous time step), ignores higher-order small quantities, and transforms the nonlinear equations into linear equations. The open-circuit voltage OCV(SOC(k)) is the only nonlinear term in the equation, so only a first-order Taylor expansion is performed on it.

[0131] Specifically, the optimal state estimate at the selected operating point is time k-1. (k-1|k-1) = [SOC0,U p10 U p20 ] Perform a first-order Taylor expansion of OCV(SOC(k)) at SOC0: OCV(SOC(k)) ≈ OCV(SOC0) + (dOCV / dSOC)·(SOC(k)-SOC0); where dOCV / dSOC is the derivative of OCV with respect to SOC, obtained by fitting the SOC-OCV curve (e.g., when SOC=50%, dOCV / dSOC=0.02V / %). Substituting the expansion into the nonlinear observation equation, we obtain the expression for the initial linearized observation equation as shown below.

[0132] U(k) ≈ OCV(SOC0) + (dOCV / dSOC)·(SOC(k) - SOC0) - U p1 (k) - U p2 (k) -R0·I(k).

[0133] S1046. Take the partial derivative of the initial linearized observation equation to obtain the Jacobian matrix and the observation matrix.

[0134] Here, the Jacobian matrix is ​​the partial derivative matrix of the nonlinear function with respect to the state variables during the linearization process, used to describe the impact of changes in state variables on the observed value (terminal voltage); the observation matrix H is the core matrix used for state correction in the AEKF algorithm, which is simplified from the Jacobian matrix and has a dimension of 1×3 (corresponding to 1 observed value and 3 state variables).

[0135] In this embodiment, since the output equation of the second-order equivalent circuit is a nonlinear equation, it needs to be linearized by performing a first-order Taylor expansion. Then, according to the system... Optimal estimate of time predict The state variable at time t, and simultaneously observe that state, to obtain Observed variables at time .use By modifying the state variables, we obtain Time-optimal state estimation This is then used as the initial value for the next time step, thus enabling the estimation of the nonlinear system.

[0136] Here, the expression for the observation matrix is ​​shown below.

[0137] ; ; ; .

[0138] in, / This is the ratio of the change in terminal voltage to the change in SOC when the SOC changes by 0.01%. The relationship between terminal voltage and SOC can be obtained by referring to the parameter tables of different cells. Furthermore, SOC is estimated using an adaptive extended Kalman filter algorithm based on the SOC state equation and observation equation.

[0139] S1047. Based on the Jacobian matrix, the observation matrix, and the system state variables, and configuring the corresponding measurement noise covariance matrix, the linearized observation equation is constructed.

[0140] Here, the measurement noise covariance matrix R is used to quantify the uncertainty in the observation process (such as the measurement error of the voltage sensor). It is usually set as a 1×1 diagonal matrix, and its value is determined by the sensor accuracy calibration. The linearized observation equation is a linear mapping relationship between the observed value (terminal voltage) and the system state variables. Combining the observation matrix, system state variables, and measurement noise, a complete state correction equation is formed.

[0141] S105. In response to the open-circuit voltage change corresponding to the historical state of charge estimation value determined in the previous sampling period satisfying the preset estimation start condition, based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, the target state of charge estimation value of the target vehicle battery in the current sampling period is determined using a preset adaptive extended Kalman filter method.

[0142] It should be noted that the Adaptive Extended Kalman Filter (AEKF) method adds an adaptive mechanism to the Extended Kalman Filter (EKF), which can dynamically adjust the noise covariance matrix and is suitable for state estimation of nonlinear systems. It is used for real-time estimation of SOC in this invention. The open-circuit voltage change (ΔOCV) is the difference in open-circuit voltage between adjacent sampling periods, which is used to determine the plateau region of the SOC-OCV curve of the lithium iron phosphate battery, and thus control the start and stop of the AEKF algorithm. The state of charge (SOC) is the ratio of the battery's current remaining capacity to its rated capacity. It is the core parameter characterizing the battery's remaining capacity and its value ranges from 0 to 100%.

[0143] In this embodiment, the Extended Kalman Filter (EKF) algorithm is suitable for handling the state estimation problem of nonlinear dynamic systems. The EKF algorithm adds an adaptive mechanism to the Extended Kalman Filter algorithm, which adaptively adjusts the filter parameters to cope with changes in model parameters or uncertainties in noise statistical characteristics, thereby improving the accuracy of state estimation. Therefore, the method described in this embodiment estimates the SOC based on the EKF algorithm to solve the impact of noise interference on the estimation accuracy.

[0144] Here, the state equations of the AEKF algorithm for nonlinear discrete systems are expressed as follows.

[0145] .

[0146] For nonlinear discrete systems, the observation equations of the AEKF algorithm are expressed as follows.

[0147] .

[0148] in, It is a state vector; The system input vector; This is the system output vector; The mean is zero and the covariance is White noise; The mean is zero and the covariance is White noise, and They are independent of each other.

[0149] Furthermore, by expanding the state equation and observation equation using a first-order Taylor series, the linearized state equation is shown below.

[0150] .

[0151] The linearized observation equation is shown below.

[0152] .

[0153] in, , , , These are the coefficient matrices after linearization.

[0154] In this embodiment of the application, the change in open circuit voltage refers to the difference between the historical open circuit voltage corresponding to the historical SOC in the previous sampling period and the open circuit voltage in the current sampling period, which is used to determine whether the SOC-OCV curve of the lithium iron phosphate battery is in the plateau region. Here, the preset estimation start condition means that the open-circuit voltage change is greater than the preset open-circuit voltage threshold (0.01V in this embodiment) for a consecutive preset number of time windows. This indicates that the battery is in the non-plateau region of the SOC-OCV curve, which is suitable for enabling the adaptive extended Kalman filter method.

[0155] In this embodiment, the Adaptive Extended Kalman Filter (AEKF algorithm) outputs the target SOC estimate for the current sampling period by combining discrete state equations and linearized observation equations through steps such as prior state estimation, innovation error calculation, Kalman gain calculation, noise covariance adaptive update, and posterior state correction.

[0156] In this way, this step intelligently controls the start and stop of the AEKF algorithm, avoiding the SOC-OCV curve plateau region and preventing estimation errors caused by algorithm erroneous activation. Simultaneously, it utilizes the adaptive characteristics of the AEKF algorithm to resist sensor noise and model errors, achieving high-precision real-time SOC estimation. It balances SOC estimation accuracy with the computational load of the onboard BMS, avoids error accumulation in the plateau region, and adaptively suppresses noise interference, achieving high-precision real-time SOC estimation under all operating conditions and throughout the entire lifecycle.

[0157] In one possible implementation of this application, step S105 may include: S1051. Based on the actual current value in the actual operating data, the target model parameters, and the historical state vector corresponding to the previous iteration cycle, the prior state is estimated using the discrete state equation to obtain the prior state vector corresponding to the current iteration cycle.

[0158] Here, prior state estimation is the core step of the AEKF algorithm for state prediction. Its core purpose is to predict the battery state in the current iteration period based on the optimal state estimation result (historical state vector) of the previous iteration period, combined with the actual current and target model parameters of the current sampling period, through discrete state equations, so as to provide an initial benchmark for subsequent state correction.

[0159] Among them, the historical state vector is the posterior state vector obtained after correction by the AEKF algorithm in the previous iteration cycle. It includes the estimated value of the historical target state of charge, the low-order polarization voltage, and the high-order polarization voltage at the previous moment. The actual current value is of the current sampling cycle (after filtering and denoising). The target model parameters are used to ensure that the parameters of the discrete state equation match the current battery aging state and operating conditions, thereby ensuring the accuracy of the prior state estimation.

[0160] In this embodiment of the application, the prior state vector corresponding to the current iteration cycle is obtained by the following equation.

[0161] (k|k-1) = A· (k-1|k-1) + B·I(k).

[0162] in, (k|k-1) represents the prior state vector corresponding to the current iteration cycle; (k-1|k-1) represents the historical state vector corresponding to the previous iteration cycle; I(k) represents the actual current value; A and B represent the first coefficient matrix and the second coefficient matrix, respectively.

[0163] Here, the prior state vector (k|k-1) represents the predicted state in the current iteration period before observation correction, including SOC and U. p1 U p2 The predicted value.

[0164] S1052. Based on the state equation parameters and the historical error covariance matrix corresponding to the previous iteration cycle, determine the prior error covariance matrix corresponding to the current iteration cycle.

[0165] In this embodiment, the prior error covariance matrix is ​​the core matrix characterizing the uncertainty of the prior state estimation and is used to quantify the prior state vector. The degree of error between (k|k-1) and the actual state vector x(k) of the battery provides a quantitative basis for subsequent Kalman gain calculation and state correction.

[0166] The historical error covariance matrix is ​​the posterior error covariance matrix obtained after correction using the AEKF algorithm in the previous iteration period (time k-1). (k-1|k-1) includes the previous state of charge (SOC) and the lower-order polarization voltage U. p1 Higher-order polarization voltage U p2The estimation error variance and covariance; the state equation parameters mainly refer to the first coefficient matrix A and the process noise covariance matrix Q, where the first coefficient matrix is ​​used to transmit the error covariance of the previous time step, and the process noise covariance matrix is ​​used to quantify the error caused by the uncertainty of the system model.

[0167] In this embodiment of the application, the prior error covariance matrix corresponding to the current iteration cycle is determined by the following formula.

[0168] (k|k-1) = A· (k-1|k-1)·A + Q.

[0169] in, (k|k-1) represents the prior error covariance matrix corresponding to the current iteration period; A represents the first coefficient matrix. (k-1|k-1) represents the historical error covariance matrix corresponding to the previous iteration cycle; Q represents the process noise covariance matrix.

[0170] S1053. Based on the observation matrix in the observation equation parameters, the ohmic internal resistance value in the target model parameters, the actual current value, and the prior state vector, the model prediction terminal voltage is determined using the linearized observation equation, and the difference between the actual voltage value in the actual operating data and the model prediction terminal voltage is determined as the innovation error.

[0171] Here, the innovation error is the core bridge connecting the prior state estimation and the actual observation in the AEKF algorithm. It is used to characterize the degree of deviation between the prior state prediction result and the actual observation data, providing a direct basis for subsequent Kalman gain calculation and state correction. Specifically, the model-predicted terminal voltage is the predicted terminal voltage value obtained by ignoring measurement noise based on the prior state vector and the linearized observation equation. Its calculation requires combining the observation matrix H in the observation equation parameters, the ohmic internal resistance R0 in the target model parameters, the actual current value I(k) in the current sampling period, and the prior state vector. (k|k-1).

[0172] In this embodiment, since the linearized observation equation is based on the voltage balance relationship of the discretized second-order equivalent circuit model, the voltage loss caused by the ohmic internal resistance needs to be supplemented. Therefore, the calculation formula for the model prediction terminal voltage is as follows.

[0173] (k|k-1) = H· (k|k-1) - R0·I(k).

[0174] in, (k|k-1) represents the model-predicted terminal voltage, and H is the observation matrix. (k|k-1) is the prior state vector, and R0·I(k) is the voltage drop caused by the ohmic internal resistance, which is consistent with the core logic of the linearized observation equation.

[0175] In this embodiment, the innovation error is obtained by the difference between the actual terminal voltage data and the model-predicted terminal voltage in the current sampling period. The absolute value directly reflects the accuracy of the prior state estimation. The larger the error, the greater the deviation between the prior prediction and the actual battery state, and the greater the magnitude of subsequent state correction needs to be. The innovation error is calculated using the following formula.

[0176] ε(k) = U_real(k) - (k|k-1), Where ε(k) represents the innovation error; U_real(k) represents the actual voltage value; (k|k-1) represents the model's predicted terminal voltage.

[0177] S1054. Based on the prior error covariance matrix, the measurement noise covariance matrix in the observation equation parameters, and the observation matrix, determine the Kalman gain matrix, and based on the innovation error, adaptively update the measurement noise covariance matrix and the process noise covariance matrix in the state equation parameters.

[0178] Here, the Kalman gain matrix is ​​the core weight matrix for state correction in the AEKF algorithm. Its core function is to balance the uncertainty of the prior state estimation (characterized by the prior error covariance matrix) with the reliability of the observation data (characterized by the measurement noise covariance matrix), determine the magnitude of the innovation error correction on the prior state vector, and ensure that the state correction can both offset the prior prediction bias and not be excessively affected by observation noise.

[0179] In this embodiment of the application, the Kalman gain matrix is ​​determined by the following formula.

[0180] K(k) = (k|k-1)·H ·(H· (k|k-1)·H + R) - ¹.

[0181] Where K(k) represents the Kalman gain matrix; (k|k-1) represents the prior error covariance matrix corresponding to the current iteration period; H Let H be the transpose of the observation matrix H, and R be the measurement noise covariance matrix.

[0182] Here, since the model error and sensor noise will change dynamically with the operating conditions and battery aging state during battery operation, fixing the noise covariance matrix will lead to a decrease in the robustness of the algorithm. Therefore, it is necessary to achieve adaptive updating of the noise covariance matrix based on the innovation error.

[0183] In this embodiment of the application, the update formula for the measurement noise covariance matrix is ​​as follows.

[0184] R(k) = R(k-1) + α·(ε(k)·ε (k) - H· (k|k-1)·H - R(k-1)).

[0185] The update formula for the process noise covariance matrix Q is shown below.

[0186] Q(k) = Q(k-1) + β·(K(k)·ε(k)·ε (k)·K (k) - (k|k-1) + A· (k-1|k-1)·A ).

[0187] Wherein, α and β are preset adaptive coefficients (in this embodiment, α=0.01 and β=0.001), used to control the update amplitude and avoid the impact of sudden changes in the noise covariance matrix on the stability of the algorithm.

[0188] In this way, through the above adaptive update, the noise covariance matrix is ​​always matched with the current battery operating state and observation bias, thereby improving the anti-interference ability and estimation robustness of the AEKF algorithm.

[0189] S1055. The prior state vector is corrected a posteriori using the Kalman gain matrix and the innovation error to obtain the target state vector corresponding to the current iteration period, and the target state of charge estimate of the target vehicle battery in the current sampling period is determined in the target state vector.

[0190] Here, the posterior correction is the core correction step of the AEKF algorithm. Its core purpose is to use the Kalman gain matrix and innovation error to accurately correct the prior state vector, offset the prior prediction bias, obtain the optimal state estimation result (target state vector) for the current iteration period, and finally extract the SOC estimate as the target value for the current sampling period.

[0191] In this embodiment of the application, the target state vector corresponding to the current iteration cycle is determined by the following formula.

[0192] (k|k) = (k|k-1) + K(k)·ε(k).

[0193] in, (k|k) is the target state vector (i.e., the posterior state vector) of the current iteration period; (k|k-1) is the prior state vector; K(k) is the Kalman gain matrix; ε(k) is the innovation error.

[0194] In this embodiment of the application, the target state vector and the prior state vector have the same dimension, which is still [SO]. (k|k), p1 (k|k), p2 (k|k)] , of which SO (k|k) is the corrected target state of charge estimate (normalized form) for the current sampling period. Multiplying it by 100% converts it to a regular percentage form, which is used for vehicle BMS display and subsequent battery control logic.

[0195] Furthermore, the a posteriori correction process will simultaneously correct the lower-order polarization voltage. p1 (k|k) and higher-order polarization voltage p2 The predicted value of (k|k) ensures the accuracy of prior state estimation in subsequent iterations, forming a closed-loop optimization, and further improving the long-term estimation stability and anti-interference ability of the AEKF algorithm.

[0196] S1056. Based on the Kalman gain matrix and the observation matrix, update the prior error covariance matrix to obtain the posterior error covariance matrix, and use the posterior error covariance matrix to calculate the next iteration cycle.

[0197] In this embodiment of the application, the prior error covariance matrix is ​​updated using the following formula.

[0198] (k|k) = [I K(k)·H]· (k|k-1).

[0199] in, (k|k) represents the posterior error covariance matrix; K(k) represents the Kalman gain matrix; H represents the observation matrix; I represents the identity matrix; (k|k-1) represents the prior error covariance matrix.

[0200] Here, the posterior error covariance matrix will be used as the historical error covariance matrix for the next iteration cycle, and used to calculate the prior error covariance matrix for the next cycle, forming a closed-loop update of the error covariance. This ensures that the error quantization of each iteration of the AEKF algorithm is based on the correction results of the previous cycle, further improving the estimation stability and accuracy of the algorithm.

[0201] Optional, please refer to Figure 2 , Figure 2 This is a second flowchart illustrating a method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle, as provided in this application embodiment. Figure 2 As shown in the embodiment of this application, the method for estimating the state of charge (SOC) of a vehicle battery throughout its entire life cycle includes step S106 in addition to steps S101 to S105. Specifically, step S106 is used to calculate the target SOC estimate value for the current sampling period by using the ampere-hour integration method when the open-circuit voltage change does not meet the estimation start condition, in order to avoid estimation errors caused by the AEKF algorithm being mistakenly started and to ensure the continuity of SOC estimation. This method is simple to calculate, consumes little computing power, and can quickly output SOC results, which is suitable for the real-time calculation requirements of the vehicle BMS.

[0202] S106. In response to the open-circuit voltage change not meeting the estimated start-up condition, the target state of charge estimate is determined based on the actual operating data using a preset ampere-hour integration method.

[0203] In this embodiment, the core principle of the ampere-hour integration method is to calculate the change in SOC by integrating the battery charging and discharging current of the current sampling period, combined with the battery rated capacity, charging and discharging efficiency and sampling period, and then obtain the estimated SOC value of the current sampling period based on the SOC value of the previous sampling period. The calculation formula is as follows.

[0204] SOC(k) = SOC(k-1) - (η·T s / C n )·I(k)×100%.

[0205] Where SOC(k) is the estimated SOC value for the current sampling period, SOC(k-1) is the SOC value for the previous sampling period, η is the charge / discharge efficiency (η≥1 during charging, η≤1 during discharging, and is taken as 1 in this embodiment), and T s For the sampling period, C n I(k) represents the battery's rated capacity, and I(k) represents the actual current value during the current sampling period (positive for charging and negative for discharging).

[0206] Thus, the method described in this application embodiment has significant advantages over the prior art, and can effectively solve the technical pain points of existing vehicle battery SOC estimation methods, such as poor adaptability throughout the entire life cycle, narrow operating condition adaptability, low estimation accuracy, easy accumulation of errors, and high computing power consumption, as detailed below: Firstly, it achieves high-precision estimation of SOC throughout the battery's entire life cycle by dynamically updating model parameters through particle swarm optimization, adapting in real time to the battery's aging state (SOH decrease), ensuring that the model parameters always match the actual electrochemical characteristics of the battery, avoiding the decrease in estimation accuracy caused by fixed parameters, and covering the full life cycle requirements of the battery from the state of new (BOL) to the state of aged (EOL).

[0207] Secondly, it adapts to all operating conditions. For the three core operating conditions of AC charging, DC charging and discharging, it identifies and updates the corresponding model parameters respectively, which solves the problem that the existing technology can only be used in the discharging condition and cannot accurately estimate the SOC during the charging process, thus meeting the diverse operating needs of vehicle batteries.

[0208] Third, it improves estimation accuracy and suppresses error accumulation. By intelligently controlling the start and stop of the AEKF algorithm, it avoids the plateau region of the SOC-OCV curve of lithium iron phosphate batteries, thus avoiding errors caused by the algorithm being turned on in the plateau region. At the same time, it utilizes the adaptive noise covariance update mechanism of the AEKF algorithm to resist sensor noise and model errors. Combined with the ampere-hour integration method as a backup, it balances estimation accuracy and continuity.

[0209] Fourth, it reduces the computational load of the vehicle-mounted BMS. The model parameters adopt an "on-demand update" mode, and particle swarm identification is triggered only when the error exceeds the threshold. Furthermore, the AEKF algorithm and the ampere-hour integration method can be flexibly switched to simplify the calculation process, adapt to the computing power requirements of the vehicle-mounted embedded system, and improve the practicality and applicability of the algorithm.

[0210] In summary, the method described in this application embodiment can achieve high-precision real-time estimation of SOC of vehicle batteries throughout their entire life cycle and under all operating conditions, providing reliable core parameter support for the BMS of new energy vehicles, helping to improve the accuracy of mileage prediction, power performance and battery life of new energy vehicles, and has high engineering application value.

[0211] The method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle provided in this application collects current charging and discharging conditions and actual operating data of the vehicle battery. It uses the particle swarm optimization method to determine and dynamically update the parameters of the discretized second-order equivalent circuit model. Combining the discrete state equation and the linearized observation equation, it determines the estimated SOC value of the vehicle battery by starting and stopping the intelligent control adaptive extended Kalman filter algorithm and using the ampere-hour integral method as a fallback. This method solves the technical defects of existing vehicle battery SOC estimation methods, such as poor adaptability throughout the entire lifecycle, narrow operating condition adaptability, low estimation accuracy, and easy accumulation of errors. It achieves high-precision real-time estimation of the SOC of the battery throughout its entire lifecycle and under all operating conditions. At the same time, it reduces the computational load of the on-board battery management system and improves the practicality and feasibility of the algorithm.

[0212] Please see Figure 3 , Figure 4 , Figure 3 This is one of the structural schematic diagrams of a vehicle battery state-of-charge estimation device for the entire life cycle provided in this application embodiment. Figure 4 This is a second schematic diagram of a device for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle, provided in an embodiment of this application. Figure 3 As shown, the state of charge estimation device 300 includes: The data acquisition module 310 is used to acquire the actual operating data of the target vehicle battery under the current charging and discharging conditions in response to obtaining the charging and discharging conditions of the target vehicle battery in the current sampling period, and to determine the actual terminal voltage data of the target vehicle battery based on the actual operating data. The model parameter module 320 is used to determine the initial model parameters of the target vehicle battery under the charging and discharging conditions based on the actual terminal voltage data, the test state data of the target vehicle battery under the charging and discharging conditions, and the preset discretized second-order equivalent circuit model, using a preset particle swarm optimization method. The parameter update module 330 is used to update the initial model parameters using the particle swarm optimization method based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold, so as to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. The equation parameter module 340 is used to determine the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation based on the target model parameters. The data estimation module 350 is used to determine the target state of charge (SBC) estimate of the target vehicle battery in the current sampling period based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, using a preset adaptive extended Kalman filter method.

[0213] Furthermore, when the model parameter module 320 is used to determine the initial model parameters of the target vehicle battery under the charging and discharging conditions based on the actual terminal voltage data, the test state data of the target vehicle battery under the charging and discharging conditions, and the preset discretized second-order equivalent circuit model, using a preset particle swarm optimization method, the model parameter module 320 is used to: Based on the test state data of the target vehicle battery under the charging and discharging conditions, the preset particle swarm method is initialized to set particle parameter information; wherein, the particle parameter information includes particle weight, learning factor, iteration number, number of particles, particle position information and velocity information, and the particle position information is the model parameter to be identified corresponding to the preset discretized second-order equivalent circuit model; Based on the particle parameter information, the model terminal voltage data is determined using the discretized second-order equivalent circuit model, and the error between the model terminal voltage data and the actual terminal voltage data is used as the optimization target to construct an offline identification fitness function for battery parameters. The particle position information is substituted into the offline identification fitness function of the battery parameters to calculate the fitness value of each particle, and based on the fitness value of each particle, the individual extreme value and the group extreme value are determined respectively; wherein, the individual extreme value includes the position information and fitness value corresponding to the current particle, and the group extreme value includes the velocity information and position information corresponding to the particle with the target fitness value; Based on the individual extreme value and the group extreme value, the velocity information of the particles is iteratively updated using a preset particle velocity update formula, and the particle position information is iteratively updated using a preset position update formula. The fitness value is recalculated based on the updated velocity and position information of the particles, and the individual extreme value and the group extreme value are updated to determine whether the preset number of iterations has been reached. If the preset number of iterations is reached, the iteration stops and the initial model parameters of the target vehicle battery under the charging and discharging conditions are output; wherein, the initial model parameters include at least the initial ohmic internal resistance value, the initial polarization internal resistance value, and the initial polarization capacitance value.

[0214] Furthermore, when the parameter update module 330 updates the initial model parameters using the particle swarm optimization method based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold, to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging condition, the parameter update module 330 is used to: Based on the discretized second-order equivalent circuit model and the test current value in the test state data, determine the model terminal voltage data; The mean absolute error between the actual terminal voltage data and the model terminal voltage data is determined using a preset voltage loss function. The average absolute error value is compared with a preset voltage error threshold to obtain the error comparison result; If the mean absolute error value is less than the preset voltage error threshold, then the initial model parameters are determined as the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. If the mean absolute error value is greater than or equal to the preset voltage error threshold, the particle swarm optimization method is triggered to identify and update the initial model parameters online to obtain the target model parameters.

[0215] Furthermore, when the equation parameter module 340 is used to pre-set the discrete state equation, the equation parameter module 340 is used for: The preset battery state of charge calculation formula is discretized using the ampere-hour integral method to obtain the initial discrete state equation, and the system state variables are defined as battery state of charge, low-order polarization voltage and high-order polarization voltage. Based on the discretized second-order equivalent circuit model, the first coefficient matrix and the second coefficient matrix are constructed respectively. The discrete state equation is determined based on the system state variables, the first coefficient matrix, the second coefficient matrix, and the experimental state data, with the process noise covariance matrix introduced.

[0216] Furthermore, when the equation parameter module 340 is used to pre-set the linearized observation equation, the equation parameter module 340 is used to: Based on the discretized second-order equivalent circuit model, a nonlinear observation equation for the terminal voltage is established. A first-order Taylor expansion of the open-circuit voltage in the terminal voltage nonlinear observation equation is performed with respect to the state of charge to linearize the terminal voltage nonlinear observation equation, resulting in an initial linearized observation equation. Taking the partial derivative of the initial linearized observation equation, we obtain the Jacobian matrix and the observation matrix; Based on the Jacobian matrix, the observation matrix, and the system state variables, and by configuring the corresponding measurement noise covariance matrix, the linearized observation equation is constructed.

[0217] Furthermore, when determining whether the open-circuit voltage change corresponding to the historical state-of-charge estimation value determined in the previous sampling period meets the preset estimation start condition, the data estimation module 350 is used to: Based on the actual terminal voltage data and the target model parameters, determine the open-circuit voltage corresponding to the current sampling period; Based on the historical open-circuit voltage value corresponding to the historical state of charge estimation value determined in the previous sampling period and the open-circuit voltage, the difference between adjacent open-circuit voltages is calculated to obtain the open-circuit voltage change, and the open-circuit voltage change corresponding to a preset number of time windows is continuously acquired. When the open-circuit voltage change corresponding to the preset number of time windows is greater than or equal to the preset open-circuit voltage threshold, it is determined that the preset estimated opening condition is met. If the change in open-circuit voltage is less than the preset open-circuit voltage threshold for any time window, it is determined that the estimated opening condition is not met.

[0218] Furthermore, such as Figure 4 As shown, the state of charge estimation device 300 further includes an ampere-hour integration module 360, which is used for: In response to the open-circuit voltage change not meeting the estimated start-up conditions, the target state of charge estimate is determined based on the actual operating data using a preset ampere-hour integration method.

[0219] Furthermore, when the data estimation module 350 determines the target state of charge estimate of the target vehicle battery in the current sampling period based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation using a preset adaptive extended Kalman filter method, the data estimation module 350 is used to: Based on the actual current value in the actual operating data, the target model parameters and the historical state vector corresponding to the previous iteration cycle, the prior state estimation is performed using the discrete state equation to obtain the prior state vector corresponding to the current iteration cycle. Based on the state equation parameters and the historical error covariance matrix corresponding to the previous iteration cycle, determine the prior error covariance matrix corresponding to the current iteration cycle. Based on the observation matrix in the observation equation parameters, the ohmic internal resistance value in the target model parameters, the actual current value, and the prior state vector, the model prediction terminal voltage is determined using the linearized observation equation, and the difference between the actual voltage value in the actual operating data and the model prediction terminal voltage is determined as the innovation error. Based on the prior error covariance matrix, the measurement noise covariance matrix in the observation equation parameters, and the observation matrix, the Kalman gain matrix is ​​determined, and based on the innovation error, the measurement noise covariance matrix and the process noise covariance matrix in the state equation parameters are adaptively updated. The prior state vector is corrected a posteriori by using the Kalman gain matrix and the innovation error to obtain the target state vector corresponding to the current iteration period, and the target state of charge estimate of the target vehicle battery in the current sampling period is determined in the target state vector. Based on the Kalman gain matrix and the observation matrix, the prior error covariance matrix is ​​updated to obtain the posterior error covariance matrix, which is then used to calculate the next iteration cycle.

[0220] The vehicle battery state-of-charge (SOC) estimation device provided in this application collects the current charging and discharging conditions and actual operating data of the vehicle battery. It uses the particle swarm optimization method to determine and dynamically update the parameters of the discretized second-order equivalent circuit model. Combining the discrete state equation and the linearized observation equation, it determines the estimated SOC value of the vehicle battery by intelligent control of the adaptive extended Kalman filter algorithm and a fallback method of ampere-hour integration. This solves the technical defects of existing vehicle battery SOC estimation methods, such as poor adaptability throughout the entire life cycle, narrow operating condition adaptation range, low estimation accuracy, and easy accumulation of errors. It achieves high-precision real-time estimation of SOC throughout the entire battery life cycle and under all operating conditions, while reducing the computational load of the vehicle battery management system and improving the practicality and feasibility of the algorithm.

[0221] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 500 includes a processor 510, a memory 520, and a bus 530.

[0222] The memory 520 stores machine-readable instructions executable by the processor 510. When the electronic device 500 is running, the processor 510 and the memory 520 communicate via the bus 530. When the machine-readable instructions are executed by the processor 510, they can perform the operations described above. Figure 1 as well as Figure 2The steps of the method for estimating the state of charge of a vehicle battery throughout its entire life cycle in the illustrated method embodiment can be found in the method embodiment for specific implementation methods, which will not be repeated here.

[0223] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 as well as Figure 2 The steps of the method for estimating the state of charge of a vehicle battery throughout its entire life cycle in the illustrated method embodiment can be found in the method embodiment for specific implementation methods, which will not be repeated here.

[0224] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0225] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0226] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0227] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0228] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0229] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle, characterized in that, The method includes: In response to obtaining the charging and discharging conditions of the target vehicle battery in the current sampling period, the actual operating data of the target vehicle battery under the charging and discharging conditions is collected, and the actual terminal voltage data of the target vehicle battery is determined based on the actual operating data. Based on the actual terminal voltage data, the test state data of the target vehicle battery under the charge and discharge conditions, and the preset discretized second-order equivalent circuit model, the initial model parameters of the target vehicle battery under the charge and discharge conditions are determined using the preset particle swarm method. Based on the comparison between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and the preset voltage error threshold, the initial model parameters are updated using the particle swarm method to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. Based on the target model parameters, the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation are determined respectively. In response to the open-circuit voltage change corresponding to the historical state of charge estimation value determined in the previous sampling period satisfying the preset estimation start condition, based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, the target state of charge estimation value of the target vehicle battery in the current sampling period is determined using a preset adaptive extended Kalman filter method.

2. The method according to claim 1, characterized in that, The charging and discharging conditions include at least AC charging conditions, DC charging conditions, and discharging conditions; the actual operating data includes at least actual current values, actual voltage values, and actual temperature values. The test status data includes at least the test current value, test voltage value, and test temperature value.

3. The method according to claim 1, characterized in that, Based on the actual terminal voltage data, the test state data of the target vehicle battery under the charge-discharge condition, and a preset discretized second-order equivalent circuit model, the initial model parameters corresponding to the target vehicle battery under the charge-discharge condition are determined using a preset particle swarm optimization method, including: Based on the test state data of the target vehicle battery under the charging and discharging conditions, the preset particle swarm method is initialized to set particle parameter information; wherein, the particle parameter information includes particle weight, learning factor, iteration number, number of particles, particle position information and velocity information, and the particle position information is the model parameter to be identified corresponding to the preset discretized second-order equivalent circuit model; Based on the particle parameter information, the model terminal voltage data is determined using the discretized second-order equivalent circuit model, and the error between the model terminal voltage data and the actual terminal voltage data is used as the optimization target to construct an offline identification fitness function for battery parameters. The particle position information is substituted into the offline identification fitness function of the battery parameters to calculate the fitness value of each particle, and based on the fitness value of each particle, the individual extreme value and the group extreme value are determined respectively; wherein, the individual extreme value includes the position information and fitness value corresponding to the current particle, and the group extreme value includes the velocity information and position information corresponding to the particle with the target fitness value; Based on the individual extreme value and the group extreme value, the velocity information of the particles is iteratively updated using a preset particle velocity update formula, and the particle position information is iteratively updated using a preset position update formula. The fitness value is recalculated based on the updated velocity and position information of the particles, and the individual extreme value and the group extreme value are updated to determine whether the preset number of iterations has been reached. If the preset number of iterations is reached, the iteration stops and the initial model parameters of the target vehicle battery under the charging and discharging conditions are output; wherein, the initial model parameters include at least the initial ohmic internal resistance value, the initial polarization internal resistance value, and the initial polarization capacitance value.

4. The method according to claim 1, characterized in that, The step of updating the initial model parameters using the particle swarm optimization method based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold, to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging condition, includes: Based on the discretized second-order equivalent circuit model and the test current value in the test state data, determine the model terminal voltage data; The mean absolute error between the actual terminal voltage data and the model terminal voltage data is determined using a preset voltage loss function. The average absolute error value is compared with a preset voltage error threshold to obtain the error comparison result; If the mean absolute error value is less than the preset voltage error threshold, then the initial model parameters are determined as the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. If the mean absolute error value is greater than or equal to the preset voltage error threshold, the particle swarm optimization method is triggered to identify and update the initial model parameters online to obtain the target model parameters.

5. The method according to claim 1, characterized in that, The discrete state equation is pre-defined through the following steps: The preset battery state of charge calculation formula is discretized using the ampere-hour integral method to obtain the initial discrete state equation, and the system state variables are defined as battery state of charge, low-order polarization voltage and high-order polarization voltage. Based on the discretized second-order equivalent circuit model, the first coefficient matrix and the second coefficient matrix are constructed respectively. The discrete state equation is determined based on the system state variables, the first coefficient matrix, the second coefficient matrix, and the experimental state data, with the process noise covariance matrix introduced.

6. The method according to claim 5, characterized in that, The linearized observation equation is pre-defined through the following steps: Based on the discretized second-order equivalent circuit model, a nonlinear observation equation for the terminal voltage is established. A first-order Taylor expansion of the open-circuit voltage in the terminal voltage nonlinear observation equation is performed with respect to the state of charge to linearize the terminal voltage nonlinear observation equation, resulting in an initial linearized observation equation. Taking the partial derivative of the initial linearized observation equation, we obtain the Jacobian matrix and the observation matrix; Based on the Jacobian matrix, the observation matrix, and the system state variables, and by configuring the corresponding measurement noise covariance matrix, the linearized observation equation is constructed.

7. The method according to claim 1, characterized in that, The following steps are used to determine whether the change in open-circuit voltage corresponding to the historical state of charge estimation value determined in the previous sampling period meets the preset estimation activation conditions: Based on the actual terminal voltage data and the target model parameters, determine the open-circuit voltage corresponding to the current sampling period; Based on the historical open-circuit voltage value corresponding to the historical state of charge estimation value determined in the previous sampling period and the open-circuit voltage, the difference between adjacent open-circuit voltages is calculated to obtain the open-circuit voltage change, and the open-circuit voltage change corresponding to a preset number of time windows is continuously acquired. When the open-circuit voltage change corresponding to the preset number of time windows is greater than or equal to the preset open-circuit voltage threshold, it is determined that the preset estimated opening condition is met. If the change in open-circuit voltage is less than the preset open-circuit voltage threshold for any time window, it is determined that the estimated opening condition is not met.

8. The method according to claim 1, characterized in that, The method further includes: In response to the open-circuit voltage change not meeting the estimated start-up conditions, the target state of charge estimate is determined based on the actual operating data using a preset ampere-hour integration method.

9. The method according to claim 1, characterized in that, The step of determining the target state of charge (SOC) estimate of the target vehicle battery in the current sampling period based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, using a preset adaptive extended Kalman filter method, includes: Based on the actual current value in the actual operating data, the target model parameters and the historical state vector corresponding to the previous iteration cycle, the prior state estimation is performed using the discrete state equation to obtain the prior state vector corresponding to the current iteration cycle. Based on the state equation parameters and the historical error covariance matrix corresponding to the previous iteration cycle, determine the prior error covariance matrix corresponding to the current iteration cycle. Based on the observation matrix in the observation equation parameters, the ohmic internal resistance value in the target model parameters, the actual current value, and the prior state vector, the model prediction terminal voltage is determined using the linearized observation equation, and the difference between the actual voltage value in the actual operating data and the model prediction terminal voltage is determined as the innovation error. Based on the prior error covariance matrix, the measurement noise covariance matrix in the observation equation parameters, and the observation matrix, the Kalman gain matrix is ​​determined, and based on the innovation error, the measurement noise covariance matrix and the process noise covariance matrix in the state equation parameters are adaptively updated. The prior state vector is corrected a posteriori by using the Kalman gain matrix and the innovation error to obtain the target state vector corresponding to the current iteration period, and the target state of charge estimate of the target vehicle battery in the current sampling period is determined in the target state vector. Based on the Kalman gain matrix and the observation matrix, the prior error covariance matrix is ​​updated to obtain the posterior error covariance matrix, which is then used to calculate the next iteration cycle.

10. A device for estimating the state of charge (SOC) of a vehicle battery throughout its entire lifecycle, characterized in that, The state of charge estimation device includes: The data acquisition module is used to respond to the acquisition of the charging and discharging conditions of the target vehicle battery in the current sampling period, collect the actual operating data of the target vehicle battery under the charging and discharging conditions, and determine the actual terminal voltage data of the target vehicle battery based on the actual operating data. The model parameter module is used to determine the initial model parameters of the target vehicle battery under the charging and discharging conditions based on the actual terminal voltage data, the test state data of the target vehicle battery under the charging and discharging conditions, and the preset discretized second-order equivalent circuit model, using a preset particle swarm optimization method. The parameter update module is used to update the initial model parameters using the particle swarm optimization method based on the comparison result between the error value between the actual terminal voltage data and the model terminal voltage data corresponding to the discretized second-order equivalent circuit model and a preset voltage error threshold, so as to obtain the target model parameters corresponding to the target vehicle battery under the charging and discharging conditions. The equation parameter module is used to determine the state equation parameters corresponding to the preset discrete state equation and the observation equation parameters corresponding to the preset linearized observation equation based on the target model parameters. The data estimation module is used to determine the target state of charge (SBC) estimate of the target vehicle battery in the current sampling period based on the actual operating data, the target model parameters, the state equation parameters, the observation equation parameters, the discrete state equation, and the linearized observation equation, using a preset adaptive extended Kalman filter method.

11. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the method for estimating the state of charge of a vehicle battery throughout its entire life cycle as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method for estimating the state of charge of a vehicle battery throughout its entire life cycle as described in any one of claims 1 to 9.

Images