Method for constructing state prediction model of battery system and related device

Abstract

The invention provides a method for constructing a state prediction model of a battery system and a related device, and the method comprises the steps: constructing a first prediction model according to the accelerated degradation experiment data of the battery system and the data of the battery system, and enabling the first prediction model to be used for predicting the internal state parameters of the battery system; a second prediction model is constructed according to charging characteristic data, time sequence characteristic data and internal state characteristic data of the battery system, the internal state characteristic data is determined according to the internal state parameters, and the second prediction model is used for predicting a health index sequence and an uncertainty range of the battery system. The health index sequence is used for reflecting the life attenuation trend of the battery system, and the uncertainty range is used for reflecting the standard deviation of each health index in the health index sequence. The thermodynamic physical model can be used as a priori condition, and the constraint condition is embedded into the neural network model, so that the state prediction accuracy of the battery system is greatly improved.

Description

Technical Field

[0001] This application relates to the field of battery state prediction technology, and in particular to a method and related apparatus for constructing a state prediction model for a battery system. Background Technology

[0002] Currently, battery system condition prediction includes remaining service life prediction, which refers to the length of time a device or system can maintain its expected function under specific operating conditions. Prediction methods are mainly divided into two categories: data-driven and model-driven. Deep learning models capture degradation characteristics by analyzing sensor time-series data; statistical models establish probability distributions based on degradation trajectories.

[0003] In the field of lithium batteries, research focuses on modeling the correlation between capacity decay and charging characteristics, and employs methods such as improved support vector regression to enhance prediction accuracy. Existing technologies generally face challenges such as adaptability to complex operating conditions and efficiency in extracting degradation features. Summary of the Invention

[0004] In view of this, this application provides a method and related apparatus for constructing a state prediction model of a battery system, which uses a thermodynamic physical model as a priori condition and embeds the constraint conditions into a neural network model, thereby greatly improving the accuracy of state prediction of the battery system.

[0005] In a first aspect, embodiments of this application provide a method for constructing a state prediction model for a battery system, the method comprising: A first prediction model is constructed based on accelerated degradation experimental data of the battery system and battery system data. The first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined according to the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the life decay trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

[0006] In one possible embodiment, constructing the first prediction model based on accelerated degradation experimental data of the battery system and battery system data includes: Acquire observed degradation data of the battery system under multiple environmental stresses; A parametric model is used to fit the observed degradation data to obtain pseudo lifetime data when the battery system degrades to a preset failure threshold. An equivalent lifetime model is constructed based on the pseudo lifetime data and the multiple environmental stresses. The equivalent lifetime model is used to reflect the mapping relationship between the pseudo lifetime data and the multiple environmental stresses. The equivalent lifetime model is updated with parameters based on the battery system data and filtering algorithm to obtain the first prediction model.

[0007] In one possible embodiment, constructing an equivalent lifetime model based on the pseudo-lifetime data and the plurality of environmental stresses includes: The characteristic lifetime is determined based on the pseudo-lifetime data; Convert the characteristic lifetime into a position parameter; The baseline environmental stress is converted into baseline parameters, and the plurality of environmental stresses are converted into a plurality of influence parameters, each influence parameter including an influence weight, and each influence weight is obtained by fitting the accelerated degradation experimental data; The location parameter is used as the result value, and the baseline parameter is summed with the plurality of influence parameters to construct the equivalent lifetime model.

[0008] In one possible embodiment, constructing the second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system includes: The charging characteristic data are determined based on the charging curve data of the battery system; The timing feature data is determined based on the timing data of the battery system; The internal state feature data is determined based on the internal state parameters; Feature vector data is determined based on the charging feature data, the timing feature data, and the internal state feature data; The feature vector data is input into a preset prediction model, and the preset prediction model is trained using a hybrid loss function to obtain the second prediction model.

[0009] In one possible embodiment, the hybrid loss function includes a first loss function, a second loss function, and a third loss function; the method further includes: The first loss function is determined based on the first error data between the output data of the preset prediction model and the actual observation data. The second loss function is determined based on the second error data of the output data and the derived data of the preset prediction model, wherein the derived data is determined based on the internal state parameters output by the first prediction model. The weighted regularization loss function is normalized to obtain the third loss function; The first loss function, the second loss function, and the third loss function are weighted and summed to obtain the hybrid loss function.

[0010] In one possible embodiment, after constructing the second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system, the method further includes: If the health status data of the battery system is higher than the first health status threshold but lower than the second health status threshold, the health index sequence and the uncertainty range are used as prior data to construct the Wiener process degradation model. Kalman filtering and Bayesian updates are performed on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time.

[0011] In one possible embodiment, constructing a Wiener process degradation model using the health index sequence and the uncertainty range as prior data includes: The Wiener process equation is obtained by integrating the time-varying drift rate, the initial health index, the health index at the corresponding time, the Wiener process random fluctuation term, and the observation noise term. The Wiener process equations are discretized to obtain the state equations and observation equations.

[0012] In one possible embodiment, after performing Kalman filtering and Bayesian updates on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time, the method further includes: The predicted remaining lifetime, first degradation rate, and first uncertainty at the current moment are determined by the output of the second prediction model. The inverse Gaussian probability density function corresponding to the predicted remaining lifetime is determined based on the first degradation rate and the first uncertainty. The second degradation rate and the second uncertainty are determined by the output of the Wiener process degradation model; The posterior probability density function corresponding to the predicted remaining lifetime is determined based on the second degradation rate and the second uncertainty. The probability density function corresponding to the predicted remaining lifetime is determined based on the inverse Gaussian distribution probability density function and the posterior distribution probability density function.

[0013] Secondly, embodiments of this application provide an apparatus for constructing a state prediction model for a battery system, the apparatus comprising: The first construction unit is used to construct a first prediction model based on the accelerated degradation experimental data of the battery system and the battery system data. The first prediction model is used to predict the internal state parameters of the battery system. The second construction unit is used to construct a second prediction model based on the charging characteristic data, time-series characteristic data and internal state characteristic data of the battery system. The internal state characteristic data is determined according to the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the life decay trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

[0014] Thirdly, embodiments of this application provide an electronic device, including a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for performing steps in any method of the first aspect of this application.

[0015] Fourthly, embodiments of this application provide a computer storage medium, wherein the computer storage medium stores a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in any method of the first aspect of embodiments of this application.

[0016] Fifthly, embodiments of this application provide a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in any method of the first aspect of this application. The computer program product may be a software installation package.

[0017] As can be seen, through the aforementioned method and apparatus for constructing a state prediction model for the battery system, a first prediction model is constructed based on accelerated degradation experimental data and battery system data. This first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined based on the internal state parameters. This second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence reflects the lifespan degradation trend of the battery system, and the uncertainty range reflects the standard deviation of each health index in the health index sequence. Thermodynamic physical models can be used as prior conditions, and constraints can be embedded into neural network models, significantly improving the accuracy of battery system state prediction. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A system architecture diagram of a method for constructing a state prediction model for a battery system provided in an embodiment of this application; Figure 2 A flowchart illustrating a method for constructing a state prediction model for a battery system provided in an embodiment of this application; Figure 3 A flowchart illustrating another method for constructing a state prediction model for a battery system provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 5 A functional unit block diagram of a device for constructing a state prediction model of a battery system provided in an embodiment of this application; Figure 6 A block diagram of the functional units of an apparatus for constructing a state prediction model of a battery system provided in an embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0021] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0022] It should be understood that the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document indicates that the preceding and following related objects are in an "or" relationship. In the embodiments of this application, "multiple" refers to two or more.

[0023] In the embodiments of this application, "at least one item" or its similar expression refers to any combination of these items, including any combination of a single item or a plurality of items. "One or more" means one or more, while "multiple" means two or more. For example, "at least one item" of a, b, or c can represent the following seven cases: a, b, c; a and b; a and c; b and c; a, b, and c. Each of a, b, and c can be an element or a set containing one or more elements.

[0024] In this application, the term "connection" refers to various connection methods, such as direct connection or indirect connection, to achieve communication between devices. This application does not impose any limitations on this.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] Please see Figure 1 , Figure 1 The system architecture diagram of a method for constructing a state prediction model of a battery system provided in this application embodiment includes a battery system 110 and a processing module 120, wherein the battery system 110 can be communicatively connected to the processing module 120.

[0027] The battery system 110 may include multiple cells 111. The health status of the battery system 110 needs to take into account the inconsistency of cell capacity. That is, theoretically, the capacity decay rate of the entire battery system 110 is determined by the cell with the fastest decay among all the cells 111 it includes.

[0028] The processing module 120 is responsible for monitoring and managing the charging and discharging process of the battery system 110, and monitoring the voltage, current, and temperature of the battery cells 111 in real time. It collects data such as voltage, current, and temperature of the battery cells 111 through various sensors, and analyzes and processes this data. The processing module 120 can construct a first prediction model, a second prediction model, etc., to accurately predict the State of Health (SOH) and Remaining Useful Life (RUL) of each battery cell 111 in the battery system 110. The processing module 120 can be a server, cloud server, etc., and is not specifically limited here.

[0029] Please see Figure 2 , Figure 2 A flowchart illustrating a method for constructing a state prediction model for a battery system, as provided in this application embodiment, specifically includes the following steps: Step 201: Construct a first prediction model based on the accelerated degradation experimental data of the battery system and the battery system data.

[0030] The first prediction model is used to predict the internal state parameters of the battery system.

[0031] In one possible embodiment, observed degradation data of the battery system under multiple environmental stresses can be obtained; a parametric model can be used to fit the observed degradation data to obtain pseudo-lifetime data when the battery system degrades to a preset failure threshold; an equivalent lifetime model can be constructed based on the pseudo-lifetime data and the multiple environmental stresses, the equivalent lifetime model being used to reflect the mapping relationship between the pseudo-lifetime data and the multiple environmental stresses; the parameters of the equivalent lifetime model can be updated based on the battery system data and a filtering algorithm to obtain the first prediction model.

[0032] The first prediction model can be a simplified physical model whose predicted internal state parameters can be used to describe the aging process of the battery system under complex operating conditions. This first prediction model does not involve deep learning; its essence remains a physical model constructed from mathematical equations. Environmental stress can be a factor leading to battery system aging, including thermal stress (such as temperature, as high temperatures drastically accelerate all chemical side reactions within the battery, leading to rapid capacity decay); electrical stress (such as power, current, state of charge window, and charging cutoff voltage, as high-rate charging and discharging generate high current densities within the battery, exacerbating electrode stress, generating heat, and potentially triggering harmful processes such as lithium dendrite formation; prolonged operation at excessively high or low states of charge, or deep charge-discharge cycles, will cause stress on the electrode material structure; excessively high charging voltage will exacerbate the structural degradation of the cathode material and electrolyte oxidation); and temporal stress, as the battery will age under certain environmental conditions even when not in use. By setting multiple environmental stresses, the observed degradation of the battery under different environmental stresses can be obtained, enabling the construction of a first prediction model that can predict the internal state parameters of the battery under real complex operating conditions. It is understandable that in order to quickly obtain data on the aging process of batteries under normal conditions, which can last for years or even decades, environmental stresses can be more severe than normal usage conditions, such as higher temperatures and higher charge / discharge rates, so as to quickly obtain data on the amount of degradation observed.

[0033] The parameterized model can be a Wiener process, an exponential model, etc. The preset failure threshold can be a pre-configured critical value for determining the amount of degradation that causes battery failure. The time required for the battery to reach the preset failure threshold is the pseudo-lifetime. For example, if the preset capacity decay to 80% of the initial value is considered failure, and the battery capacity drops to 79% of the initial value after 120 days of cycling, then its pseudo-lifetime is 120 days.

[0034] For example, under environmental stress Next, for the first Degradation of each sample Observations were conducted. A parametric model was used for fitting:

[0035] in, The degradation rate is related to environmental stress. It is a time-scale function; Let be a random variable, representing fluctuations and measurement noise; when First time reaching the preset failure threshold At that time, the corresponding moment This is the pseudo-lifetime of the sample under that environmental stress. Collecting the pseudo-lifetimes under all environmental stresses yields the pseudo-lifetime data. The timescale function can be a function derived from electrochemical mechanisms to describe the inherent trend of lithium-ion loss or internal resistance growth. Its essence is physical information, while pseudo-lifetime data can reflect the impact of environmental stress on the remaining life of the battery, which facilitates the subsequent coupling of physical information and data-driven processes.

[0036] Furthermore, in one possible embodiment, constructing an equivalent lifetime model based on the pseudo-lifetime data and the plurality of environmental stresses includes: determining a characteristic lifetime based on the pseudo-lifetime data; converting the characteristic lifetime into a location parameter; converting a reference environmental stress into a reference parameter; and converting the plurality of environmental stresses into a plurality of influence parameters, each influence parameter including an influence weight, each influence weight being fitted based on the accelerated degradation experimental data; using the location parameter as the result value, and summing the reference parameter with the plurality of influence parameters to construct the equivalent lifetime model.

[0037] Characteristic lifetime, such as the median lifetime obtained from pseudo-lifetime data, is not the individual lifetime of a single battery sample, but rather a statistical characteristic value of the lifetime of a batch of batteries of the same type under the same environmental stress, reflecting the overall aging level of the batch. For example, under a fixed environmental stress, such as 45°C, a batch of batteries is tested. When 50% of the batteries in this batch experience a capacity decay to 80% of their initial capacity, the corresponding time is the characteristic lifetime under this environmental stress. Accelerated degradation experiments yield the failure time of each sample under different environmental stresses, which are discrete data points. Assuming that these discrete points follow a location-scale family distribution, such as a Weibull distribution or a log-normal distribution, the representative characteristic lifetime of the distribution can be determined from the discrete data points. Determining the characteristic lifetime under different environmental stresses provides a data foundation for subsequently establishing a model of the relationship between environmental stress and pseudo-lifetime. The characteristic lifetime can be logarithmically converted into a location parameter. For example, under stress... Below, its position parameters (e.g., scale parameters) and characteristic lifetimes of pseudo-lifetimes (e.g., median lifetime). Existence relationship, that is:

[0038] The relationship between characteristic lifetime and stress is described using a generalized linear acceleration model:

[0039] In the formula, The baseline parameter represents the baseline value of the logarithmic characteristic life under the baseline environmental stress, and serves as the starting point for model fitting. Environmental stress The known function is the j-th influencing parameter; different environmental stresses correspond to different influencing parameters. To influence the weights; to address the limitations of a single acceleration model, it is necessary to construct an equivalent lifetime model that couples multiple factors, including ambient temperature ( ), power (P), SOC window ( SOC, charging cutoff voltage ( To establish an equivalent lifetime model, we need to consider multiple dimensions, including:

[0040] in, k is the parameter affecting temperature. B The constant is Boltzmann's constant, and its influence weight is usually an integer. As the ambient temperature T increases, Reduced, thus leading to The reduction, or shortened lifespan, aligns with the physical characteristic of accelerated aging at high temperatures.

[0041] in, These are the parameters affecting charge / discharge rates, used to describe the impact of charge / discharge current on battery life. The relationship between battery life and charge / discharge rate is exponential or power-law related, not a simple linear one. It is usually a negative number. When C increases, the charging and discharging speeds up. It also increases, due to It is a negative number, thus leading to The reduction in lifespan is consistent with the physical characteristic of accelerated aging caused by high current.

[0042] in, This describes the impact of varying states of charge (SOCs) on battery life. For example, always cycling between 20% and 80% has a drastically different impact on battery life compared to deep charging and discharging between 0% and 100%. The impact of charging cutoff voltage on battery life is discussed. Excessively high charging voltage forces excessive lithium-ion intercalation and deintercalation, leading to damage to the cathode material structure and oxidative decomposition of the electrolyte. and It is a function obtained by fitting electrochemical mechanisms or experimental data, and the influence weight of this function. DOE data can be obtained by designing accelerated degradation experiments and fitting them using the least squares method.

[0043] In one possible embodiment, parameters affecting environmental stresses, such as discharge cutoff voltage, ambient humidity, and settling time, may also be added, without specific limitations.

[0044] Furthermore, using battery system data such as real-time voltage, real-time current, and real-time temperature, Kalman filtering or particle filtering is applied to analyze key aging parameters in the equivalent lifetime model, such as the diffusion coefficients of the positive and negative electrodes. , reaction rate constant k, internal resistance Real-time joint state and parameter estimation is then performed. The process noise covariance matrix of the filter can be adaptively adjusted based on the aging trend predicted by the equivalent lifetime model, thereby obtaining the first prediction model.

[0045] It is evident that by constructing the first prediction model based on the accelerated degradation experimental data of the battery system and the battery system data, an online adaptive thermodynamic physical field model can be built, providing a reliable physical interpretability input for the subsequent construction of the neural network model.

[0046] Step 202: Construct a second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system.

[0047] The internal state characteristic data is determined based on the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the lifespan degradation trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

[0048] In one possible embodiment, the charging characteristic data can be determined based on the charging curve data of the battery system; the time-series characteristic data can be determined based on the time-series data of the battery system; the internal state characteristic data can be determined based on the internal state parameters; feature vector data can be determined based on the charging characteristic data, the time-series characteristic data, and the internal state characteristic data; the feature vector data can be input into a preset prediction model, and the preset prediction model can be trained using a hybrid loss function to obtain the second prediction model.

[0049] The charging curve data can include curves showing the change of battery voltage over time and the change of charged capacity over time. It can be represented as a voltage-capacity curve. By differentiating the original voltage-capacity curve, the change in charged or discharged capacity corresponding to each small change in voltage can be obtained, thus revealing easily identifiable change characteristics. For example, peak voltage reflects changes in the battery's internal thermodynamic equilibrium potential. As the battery ages, the peak voltage shifts, usually towards higher voltages. Monitoring the shift in peak voltage can directly determine the aging process. Peak height is directly related to the battery's reversible capacity. As the battery ages, active materials are consumed, and the amount of material participating in phase transition reactions decreases, leading to a decrease in peak height. Monitoring the rate of peak height decay can determine the battery capacity decay rate. Peak area is positively correlated with the amount of charge participating in a specific electrochemical reaction. The decay of the peak area directly reflects the amount of active material lost corresponding to that electrochemical reaction.

[0050] Based on time-series data, time-series characteristic data can be statistically extracted, including voltage relaxation slope, constant current charge / discharge efficiency, and temperature rise rate. The voltage relaxation segment is the recovery phase after charging or discharging stops. As the battery ages, internal resistance increases, diffusion slows, and polarization intensifies. Therefore, the voltage drops faster after charging stops, or the voltage recovers more slowly after resting. The magnitude and direction of the slope directly reflect the severity of internal polarization and the relaxation rate. The constant current charge / discharge efficiency is negatively correlated with the degree of battery aging; the temperature rise rate is positively correlated with the degree of battery aging. Power station batteries rarely undergo a complete cycle from 0% to 100%. However, these characteristics can be extracted from any operational segment, such as a two-hour constant power discharge process, or the resting period after charging, without requiring large amounts of continuous data.

[0051] The internal state characteristic data can be obtained based on the internal state parameters obtained from the first prediction model. The internal state characteristic data can include the ratio of the currently estimated reaction rate constant to the nominal value, where the nominal value is the baseline reaction rate constant. The reaction rate constant can describe the ease and speed of the electrochemical reaction of lithium ions on the electrode surface. A ratio of 1 indicates that the battery's reaction activity is like that of a brand new battery. As the battery ages, this ratio will gradually decrease.

[0052] The internal state characteristic data can include the solid-phase diffusion time constant, which represents the diffusion rate of lithium ions within the electrode solid material particles. This is a key bottleneck limiting the battery's fast-charging capability and power characteristics. This process can be described by the solid-phase diffusion time constant, whose value is positively correlated with the degree of battery aging. By using the internal state characteristic data, a first prediction model can be introduced, making the obtained feature vector data more realistic and improving accuracy.

[0053] The feature vector data HFV can be in the following form:

[0054] In the formula, For charging characteristic data, from partial charging curves , In the process, the incremental capacity dQ / dV curve is calculated, and the peak value is extracted. Features such as peak area.

[0055] For time-series characteristic data, extract statistical features from the time-series data, such as the slope of the voltage relaxation segment. Constant current charge / discharge efficiency η, temperature rise rate, etc.

[0056] For internal state feature data, extract internal state features from the online adaptive physics model, such as the currently estimated reaction rate constant. Compared with nominal value The ratio of solid-phase diffusion time constant wait.

[0057] Furthermore, the hybrid loss function includes a first loss function, a second loss function, and a third loss function; the method further includes: determining a first loss function based on a first error data between the output data of the preset prediction model and the actual observed data; determining a second loss function based on a second error data between the output data of the preset prediction model and the derived data, wherein the derived data is determined based on the internal state parameters output by the first prediction model; normalizing the weighted regularization loss function to obtain the third loss function; and weighting and summing the first loss function, the second loss function, and the third loss function to obtain the hybrid loss function.

[0058] For example, the mixed loss function is expressed by the following formula:

[0059] In the formula, The first loss function is the mean squared error between the predicted health index (such as capacity retention rate) and the observed value. The second loss function is the physical consistency impairment, for example, forcing network predictions. With internal state characteristics Calculated from Error correction; The third loss function is used for weight regularization to prevent overfitting. The weighting coefficients adjust the contribution ratio of each loss function in the mixed loss function. The first loss function not only ensures that the predicted values ​​of the second prediction model are as close as possible to the actual observed values, but also ensures that the prediction results of the second prediction model conform to physical laws. Simultaneously, the third loss function prevents the second prediction model from becoming overly complex or overfitting.

[0060] As can be seen, the above-described method for constructing a state prediction model for a battery system involves building a first prediction model based on accelerated degradation experimental data and battery system data. This first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is then built based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined based on the internal state parameters. This second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence reflects the lifespan degradation trend of the battery system, and the uncertainty range reflects the standard deviation of each health index in the health index sequence. By using a thermodynamic physical model as a priori condition and embedding constraints into a neural network model, the accuracy of battery system state prediction is significantly improved.

[0061] Please see Figure 3 , Figure 3 A flowchart illustrating another method for constructing a state prediction model for a battery system provided in this application embodiment is shown, specifically including the following steps: Step 301: Construct a first prediction model based on the accelerated degradation experimental data of the battery system and the battery system data.

[0062] Step 302: Construct a second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system.

[0063] Step 303: If the health status data of the battery system is higher than the first health status threshold but lower than the second health status threshold, the health index sequence and the uncertainty range are used as prior data to construct the Wiener process degradation model.

[0064] The battery lifecycle is divided into three stages: the initial stable period, the mid-term degradation period, and the final accelerated degradation period. The initial stable period is characterized by slow capacity decay, with the State of Health (SOH) maintained above 90%, and stable system performance. During the mid-term degradation period, the capacity decay rate accelerates, with the SOH ranging between 70% and 90%, requiring enhanced monitoring and maintenance. In the final accelerated degradation period, the SOH falls below 70%, resulting in significant capacity loss until the 60% lifespan endpoint is reached. At this point, system reliability drops drastically, necessitating replacement or deep repair. Therefore, a mathematical model using a two-stage prediction framework is employed for SOH > 70%. The first health state threshold is 70%, and the second health state threshold is 90%.

[0065] In the first stage, when SOH > 90%, the data output by the second prediction model is used directly. For example, the health index prediction sequence and uncertainty for the next L time points output by the second prediction model are expressed by the following formula:

[0066] In the second stage, where 90% > SOH > 70%, adaptive Bayesian and Wiener processes are needed for micro-level updates. Assuming a health index... The degradation follows a Wiener process with a time-varying drift rate:

[0067] in, The time-varying drift rate is modeled as a stochastic process: ; It is the process diffusion coefficient. This is the observation error. Discretizing it to form a state-space model facilitates Kalman filtering.

[0068] Observation equation:

[0069] Among them, the process noise covariance matrix and and Related to; observation noise variance .

[0070] Step 304: Perform Kalman filtering and Bayesian update on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time.

[0071] The parameter estimation and Bayesian update may include the following steps: Initialization: Drift rate It can be set as the initial slope of the predicted sequence, using the future trend predicted by the AI ​​model to initialize the state vector and covariance matrix.

[0072] Online updates: When new observational data becomes available At that time, the Kalman filter prediction and update steps are run to obtain the posterior state estimate. , Its covariance matrix enables online noise reduction and parameter adaptation.

[0073] Small-sample adaptation: Initial model parameters for new battery batches Unknown. We can define a prior distribution (such as a conjugate prior) and then use a small amount of field data to perform posterior sampling using the Markov Chain Monte Carlo (MCMC) method. Calculate the posterior probability distribution: For the i-th cycle, the Metropolis-Hastings method is used for the i-th iteration. The acceptance probability is calculated as follows:

[0074] With probability accept ,otherwise

[0075] Use the samples from cycle i+1 to calculate the posterior distribution.

[0076] In one possible embodiment, after performing Kalman filtering and Bayesian updates on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time, the method further includes: The predicted remaining lifetime, first degradation rate, and first uncertainty at the current moment are determined by the output of the second prediction model. The inverse Gaussian probability density function corresponding to the predicted remaining lifetime is determined based on the first degradation rate and the first uncertainty. The second degradation rate and the second uncertainty are determined by the output of the Wiener process degradation model; The posterior probability density function corresponding to the predicted remaining lifetime is determined based on the second degradation rate and the second uncertainty. The probability density function corresponding to the predicted remaining lifetime is determined based on the inverse Gaussian distribution probability density function and the posterior distribution probability density function.

[0077] For example, the remaining useful life (RUL) distribution can be derived over time. The current degradation level is known. The failure threshold is Remaining service life .

[0078] Given the current state estimate (E[ ], E[ Under conditions of uncertainty, The probability density function (PDF) can be approximately derived for the linear drift case:

[0079] in, (·) is the PDF of the inverse Gaussian distribution. This integral is obtained by sampling from the KF posterior state distribution using the MCMC method described above. The sample distribution.

[0080] As can be seen, this study focuses on multi-physics coupled modeling of energy storage batteries for RUL prediction: It investigates thermodynamic physical field models suitable for online adaptation. The key challenge is rapid modeling of critical internal states of batteries under complex operating conditions (wide temperature range, flexible operating conditions); and extraction techniques for key physical field features reflecting thermodynamic / kinetic degradation, providing physically interpretable input for AI models. Based on multi-source heterogeneous operating data, the study investigates feature extraction algorithms to reconstruct the battery state of health (SOH) from partial charge-discharge segments, addressing the pain points of fragmented and incomplete BMS data in power plants. It integrates time-series data such as voltage, current, and temperature, as well as internal state features calculated based on physical models, to construct a high-dimensional health feature vector. Finally, it constructs a coupled AI prediction model driven by physical fields and data: using thermodynamic physical models as prior conditions and embedding constraints into deep learning networks, adaptively selecting a multi-model architecture of convolutional neural networks + long-term series data + Transformer. Adaptive Model Based on Data-Driven and Bayesian-Wiener Process: This model is applicable to the adaptive selection of multi-stage degradation models throughout the entire lifecycle of lithium-ion batteries. It investigates small-sample / zero-sample adaptive prediction algorithms across operating conditions and battery batches, enabling the model to quickly migrate to new operating environments using laboratory data and limited field data. It also provides adaptive Wiener models with measurement errors and various algorithms such as Kalman filtering to denoise power plant data. Online Adaptive Calibration and Uncertainty Quantification of Remaining Life Prediction Results: This research investigates dynamic calibration techniques for prediction models based on Bayesian updates or online learning frameworks, continuously correcting the prediction trajectory using real-time incoming data. It establishes a method for quantifying the uncertainty of prediction results, providing confidence intervals for operation and maintenance decisions. This achieves a unity of mechanism interpretability and data adaptability. Physical / chemical equations describing the essence of battery aging (such as degradation trajectory models for capacity decay and Wiener processes reflecting stress acceleration effects) are deeply embedded in the training and inference processes of neural networks (such as Transformers) as core constraints or prior knowledge. This overcomes the inherent defects of pure data models, such as "black box" decision-making, poor extrapolation, and the inability to apply high-fidelity physical model calculations online. It possesses early and accurate prediction capabilities and small-sample adaptive capabilities. The model can utilize limited data from the early stages of the battery's lifespan (the first 100 cycles) and combine this data with physical laws to infer long-term degradation trajectories, enabling early health warnings. Simultaneously, when battery operating conditions or batches change, the model can quickly calibrate through online learning, significantly improving its generalization ability and robustness across different application scenarios. It reduces reliance on complete historical failure data, better reflecting the fragmented data reality of actual energy storage power station operation. The constructed fusion model boasts high computational efficiency and can be deployed on BMS or edge computing devices to achieve real-time online state estimation and closed-loop lifetime management, providing direct evidence for predictive maintenance.

[0081] As can be seen, the above-described method for constructing a state prediction model for a battery system involves building a first prediction model based on accelerated degradation experimental data and battery system data. This first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is then built based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined based on the internal state parameters. This second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence reflects the lifespan degradation trend of the battery system, and the uncertainty range reflects the standard deviation of each health index in the health index sequence. By using a thermodynamic physical model as a priori condition and embedding constraints into a neural network model, the accuracy of battery system state prediction is significantly improved.

[0082] For steps not detailed above, please refer to Figure 2 The steps described in the instructions will not be repeated here.

[0083] The following is combined with Figure 4 An electronic device according to an embodiment of this application will be described. Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. The electronic device 400 includes a processor 401, a memory 402, and a communication bus 403 for connecting the processor 401 and the memory 402.

[0084] In some possible implementations, memory 402 includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or compact discread-only memory (CD-ROM), which is used to store program code executed by electronic device 400 and data transmitted.

[0085] In some possible implementations, the electronic device 400 also includes a communication interface for receiving and sending data.

[0086] In some possible implementations, processor 401 may be one or more central processing units (CPUs). If processor 401 is a central processing unit (CPU), the central processing unit (CPU) may be a single-core central processing unit (CPU) or a multi-core central processing unit (CPU).

[0087] In some possible implementations, processor 401 may be a baseband chip, a chip, a central processing unit (CPU), a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof.

[0088] In specific implementation, the processor 401 in the electronic device 400 executes the program instructions 421 stored in the memory 402 to perform the following operations: A first prediction model is constructed based on accelerated degradation experimental data of the battery system and battery system data. The first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined according to the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the life decay trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

[0089] In one possible embodiment, constructing the first prediction model based on accelerated degradation experimental data of the battery system and battery system data includes: Acquire observed degradation data of the battery system under multiple environmental stresses; A parametric model is used to fit the observed degradation data to obtain pseudo lifetime data when the battery system degrades to a preset failure threshold. An equivalent lifetime model is constructed based on the pseudo lifetime data and the multiple environmental stresses. The equivalent lifetime model is used to reflect the mapping relationship between the pseudo lifetime data and the multiple environmental stresses. The equivalent lifetime model is updated with parameters based on the battery system data and filtering algorithm to obtain the first prediction model.

[0090] In one possible embodiment, constructing an equivalent lifetime model based on the pseudo-lifetime data and the plurality of environmental stresses includes: The characteristic lifetime is determined based on the pseudo-lifetime data; Convert the characteristic lifetime into a position parameter; The baseline environmental stress is converted into baseline parameters, and the plurality of environmental stresses are converted into a plurality of influence parameters, each influence parameter including an influence weight, and each influence weight is obtained by fitting the accelerated degradation experimental data; The location parameter is used as the result value, and the baseline parameter is summed with the plurality of influence parameters to construct the equivalent lifetime model.

[0091] In one possible embodiment, constructing the second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system includes: The charging characteristic data are determined based on the charging curve data of the battery system; The timing feature data is determined based on the timing data of the battery system; The internal state feature data is determined based on the internal state parameters; Feature vector data is determined based on the charging feature data, the timing feature data, and the internal state feature data; The feature vector data is input into a preset prediction model, and the preset prediction model is trained using a hybrid loss function to obtain the second prediction model.

[0092] In one possible embodiment, the hybrid loss function includes a first loss function, a second loss function, and a third loss function; the method further includes: The first loss function is determined based on the first error data between the output data of the preset prediction model and the actual observation data. The second loss function is determined based on the second error data of the output data and the derived data of the preset prediction model, wherein the derived data is determined based on the internal state parameters output by the first prediction model. The weighted regularization loss function is normalized to obtain the third loss function; The first loss function, the second loss function, and the third loss function are weighted and summed to obtain the hybrid loss function.

[0093] In one possible embodiment, after constructing the second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system, the method further includes: If the health status data of the battery system is higher than the first health status threshold but lower than the second health status threshold, the health index sequence and the uncertainty range are used as prior data to construct the Wiener process degradation model. Kalman filtering and Bayesian updates are performed on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time.

[0094] In one possible embodiment, constructing a Wiener process degradation model using the health index sequence and the uncertainty range as prior data includes: The Wiener process equation is obtained by integrating the time-varying drift rate, the initial health index, the health index at the corresponding time, the Wiener process random fluctuation term, and the observation noise term. The Wiener process equations are discretized to obtain the state equations and observation equations.

[0095] In one possible embodiment, performing Kalman filtering and Bayesian updates on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time includes: Initialize the time-varying drift rate, state vector, and covariance matrix of the Wiener process degradation model; Kalman filtering and Bayesian update are performed based on the battery system data to obtain the posterior state estimate and the corresponding covariance matrix. The posterior probability distribution is determined using the Markov chain Monte Carlo algorithm.

[0096] As can be seen, through the aforementioned method and apparatus for constructing a state prediction model for the battery system, a first prediction model is constructed based on accelerated degradation experimental data and battery system data. This first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined based on the internal state parameters. This second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence reflects the lifespan degradation trend of the battery system, and the uncertainty range reflects the standard deviation of each health index in the health index sequence. Thermodynamic physical models can be used as prior conditions, and constraints can be embedded into neural network models, significantly improving the accuracy of battery system state prediction.

[0097] It should be noted that the specific implementation of each operation can be described in the corresponding description of the method embodiments shown above. The electronic device 400 can be used to execute the method embodiments of this application, and will not be described again here.

[0098] The above primarily describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, the electronic device includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0099] This application embodiment can divide the electronic device into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0100] When dividing each function into modules according to its corresponding function. Figure 5 This is a functional unit block diagram of a battery system state prediction model construction device 500 provided in an embodiment of this application. The device includes: The first construction unit 510 is used to construct a first prediction model based on the accelerated degradation experimental data of the battery system and the battery system data. The first prediction model is used to predict the internal state parameters of the battery system. The second construction unit 520 is used to construct a second prediction model based on the charging characteristic data, time-series characteristic data and internal state characteristic data of the battery system. The internal state characteristic data is determined according to the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the life decay trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

[0101] As can be seen, through the aforementioned method and apparatus for constructing a state prediction model for the battery system, a first prediction model is constructed based on accelerated degradation experimental data and battery system data. This first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined based on the internal state parameters. This second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence reflects the lifespan degradation trend of the battery system, and the uncertainty range reflects the standard deviation of each health index in the health index sequence. Thermodynamic physical models can be used as prior conditions, and constraints can be embedded into neural network models, significantly improving the accuracy of battery system state prediction.

[0102] It should be noted that the specific implementation of each operation can be described in the corresponding description of the method embodiments shown above. The battery system state prediction model construction device 500 can be used to execute the above method embodiments of this application, and will not be described again here.

[0103] When using integrated units, the following is combined with Figure 6 A detailed description of another battery system state prediction model construction apparatus 600 in the embodiments of this application is provided. The battery system state prediction model construction apparatus 600 includes a processing unit 601 and a communication unit 602. The processing unit 601 is used to perform any step as described in the above method embodiments, and when performing data transmission such as sending, the communication unit 602 can be selectively invoked to complete the corresponding operation.

[0104] The battery system state prediction model construction device 600 may further include a storage unit 603 for storing program code and data. The processing unit 601 may be a processor, the communication unit 602 may be a wireless communication module, and the storage unit 603 may be a memory; the processing unit 601 is specifically used for: A first prediction model is constructed based on accelerated degradation experimental data of the battery system and battery system data. The first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined according to the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the life decay trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

[0105] As can be seen, through the aforementioned method and apparatus for constructing a state prediction model for the battery system, a first prediction model is constructed based on accelerated degradation experimental data and battery system data. This first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined based on the internal state parameters. This second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence reflects the lifespan degradation trend of the battery system, and the uncertainty range reflects the standard deviation of each health index in the health index sequence. Thermodynamic physical models can be used as prior conditions, and constraints can be embedded into neural network models, significantly improving the accuracy of battery system state prediction.

[0106] It should be noted that the specific implementation of each operation can be described in the corresponding description of the method embodiments shown above. The battery system state prediction model construction device 600 can be used to execute the above method embodiments of this application, and will not be described again here.

[0107] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the methods described in the above method embodiments, wherein the computer includes an electronic device.

[0108] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments. The computer program product may be a software installation package, and the computer may include an electronic device.

[0109] It should be noted that, for the sake of simplicity, the above embodiments are all described as a series of actions. Those skilled in the art should understand that this application is not limited to the described order of actions, as some steps in the embodiments of this application can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions, steps, modules, or units involved are not necessarily essential to the embodiments of this application.

[0110] In the above embodiments, the descriptions of each embodiment in this application have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0111] The steps of the methods or algorithms described in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in RAM, flash memory, ROM, EPROM, electrically erasable programmable read-only memory (EEPROM), registers, hard disk, portable hard disk, read-only optical disk (CD-ROM), or any other form of storage medium well known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Furthermore, the ASIC can reside in a terminal device or management device. Alternatively, the processor and storage medium can exist as discrete components in the terminal device or management device.

[0112] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in the embodiments of this application can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0113] The modules / units included in the various devices and products described in the above embodiments can be software modules / units, hardware modules / units, or a combination of both. For example, for devices and products applied to or integrated into a chip, all modules / units can be implemented using hardware methods such as circuits, or at least some modules / units can be implemented using software programs that run on a processor integrated within the chip, while the remaining (if any) modules / units can be implemented using hardware methods such as circuits. For devices and products applied to or integrated into a chip module, all modules / units can be implemented using hardware methods such as circuits. Different modules / units can be located in the same component (e.g., chip, circuit module, etc.) or different components of the chip module, or at least some modules / units can be implemented using hardware methods such as circuits. The implementation is achieved through a software program that runs on the processor integrated within the chip module. The remaining modules / units (if any) can be implemented using hardware methods such as circuits. For various devices and products applied to or integrated into terminal equipment, each of their modules / units can be implemented using hardware methods such as circuits. Different modules / units can be located in the same component (e.g., chip, circuit module, etc.) or different components within the terminal equipment. Alternatively, at least some modules / units can be implemented through a software program that runs on the processor integrated within the terminal equipment, while the remaining modules / units (if any) can be implemented using hardware methods such as circuits.

[0114] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the embodiments of this application. It should be understood that the above descriptions are merely specific embodiments of the embodiments of this application and are not intended to limit the protection scope of the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solutions of the embodiments of this application should be included within the protection scope of the embodiments of this application.

Claims

1. A method for constructing a state prediction model for a battery system, characterized in that, The method includes: A first prediction model is constructed based on accelerated degradation experimental data of the battery system and battery system data. The first prediction model is used to predict the internal state parameters of the battery system. A second prediction model is constructed based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system. The internal state characteristic data is determined according to the internal state parameters. The second prediction model is used to predict the health index sequence and uncertainty range of the battery system. The health index sequence is used to reflect the life decay trend of the battery system, and the uncertainty range is used to reflect the standard deviation of each health index in the health index sequence.

2. The method according to claim 1, characterized in that, The construction of the first prediction model based on accelerated degradation experimental data of the battery system and battery system data includes: Acquire observed degradation data of the battery system under multiple environmental stresses; A parametric model is used to fit the observed degradation data to obtain pseudo lifetime data when the battery system degrades to a preset failure threshold. An equivalent lifetime model is constructed based on the pseudo lifetime data and the multiple environmental stresses. The equivalent lifetime model is used to reflect the mapping relationship between the pseudo lifetime data and the multiple environmental stresses. The equivalent lifetime model is updated with parameters based on the battery system data and filtering algorithm to obtain the first prediction model.

3. The method according to claim 2, characterized in that, The step of constructing an equivalent lifetime model based on the pseudo-lifetime data and the multiple environmental stresses includes: The characteristic lifetime is determined based on the pseudo-lifetime data; Convert the characteristic lifetime into a position parameter; The baseline environmental stress is converted into baseline parameters, and the plurality of environmental stresses are converted into a plurality of influence parameters, each influence parameter including an influence weight, and each influence weight is obtained by fitting the accelerated degradation experimental data; The location parameter is used as the result value, and the baseline parameter is summed with the plurality of influence parameters to construct the equivalent lifetime model.

4. The method according to any one of claims 1-3, characterized in that, The construction of the second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system includes: The charging characteristic data are determined based on the charging curve data of the battery system; The timing feature data is determined based on the timing data of the battery system; The internal state feature data is determined based on the internal state parameters; Feature vector data is determined based on the charging feature data, the timing feature data, and the internal state feature data; The feature vector data is input into a preset prediction model, and the preset prediction model is trained using a hybrid loss function to obtain the second prediction model.

5. The method according to claim 4, characterized in that, The hybrid loss function includes a first loss function, a second loss function, and a third loss function; the method further includes: The first loss function is determined based on the first error data between the output data of the preset prediction model and the actual observation data. The second loss function is determined based on the second error data of the output data and the derived data of the preset prediction model, wherein the derived data is determined based on the internal state parameters output by the first prediction model. The weighted regularization loss function is normalized to obtain the third loss function; The first loss function, the second loss function, and the third loss function are weighted and summed to obtain the hybrid loss function.

6. The method according to any one of claims 1-5, characterized in that, After constructing the second prediction model based on the charging characteristic data, time-series characteristic data, and internal state characteristic data of the battery system, the method further includes: If the health status data of the battery system is higher than the first health status threshold but lower than the second health status threshold, the health index sequence and the uncertainty range are used as prior data to construct the Wiener process degradation model. Kalman filtering and Bayesian updates are performed on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time.

7. The method according to claim 6, characterized in that, The step of constructing a Wiener process degradation model using the health index sequence and the uncertainty range as prior data includes: The Wiener process equation is obtained by integrating the time-varying drift rate, the initial health index, the health index at the corresponding time, the Wiener process random fluctuation term, and the observation noise term. The Wiener process equations are discretized to obtain the state equations and observation equations.

8. The method according to claim 6, characterized in that, After performing Kalman filtering and Bayesian updates on the Wiener process degradation model based on the battery system data to adjust the Wiener process degradation model in real time, the method further includes: The predicted remaining lifetime, first degradation rate, and first uncertainty at the current moment are determined by the output of the second prediction model. The inverse Gaussian probability density function corresponding to the predicted remaining lifetime is determined based on the first degradation rate and the first uncertainty. The second degradation rate and the second uncertainty are determined by the output of the Wiener process degradation model; The posterior probability density function corresponding to the predicted remaining lifetime is determined based on the second degradation rate and the second uncertainty. The probability density function corresponding to the predicted remaining lifetime is determined based on the inverse Gaussian distribution probability density function and the posterior distribution probability density function.

9. An electronic device, characterized in that, include: A processor, a memory, and one or more programs; the one or more programs are stored in the memory and configured to be executed by the processor, the programs including instructions for performing the steps of the method as described in any one of claims 1-8.

10. A computer storage medium, characterized in that, The computer storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-8.

Images