Battery degradation path determination method and device based on large language model

Abstract

The invention discloses a battery degradation path determination method and device based on a large language model. The method is applied to a rechargeable lithium battery, the rechargeable lithium battery comprises at least one battery cell, and the method comprises the following steps: for at least one battery cell, acquiring battery related parameters of the battery cell in a current cycle period; according to the likelihood function corresponding to each degradation mode and the prior probability, processing the feature vectors corresponding to the battery related parameters to obtain posterior probabilities corresponding to various degradation modes; determining a target degradation mode according to the posterior probability corresponding to each degradation mode, and processing the target degradation mode, a target degradation path corresponding to the mode, a feature contribution rate corresponding to each index in the feature vector and the feature vector based on a target language model, and obtaining the interpretable content and the corresponding confidence corresponding to each piece of degradation information in the target degradation path for judging whether the interpretable content is displayed or not. And the accuracy and interpretability of battery degradation path prediction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery management, in particular to a battery degradation path determination method and device based on a large language model. BACKGROUND

[0002] With the acceleration of the global energy structure transformation to renewable energy, as a key technology to balance supply and demand and improve the stability of the power grid, the reliability and life management of energy storage batteries have become the focus of the industry. Battery degradation path prediction technology aims to identify performance degradation trends early, optimize maintenance strategies, reduce operation and maintenance costs, and prevent potential safety risks.

[0003] Currently, a physical model method based on electrochemical mechanism can be used to construct differential equation models of degradation mechanisms such as SEI film growth, lithium dendrite generation, and electrode material phase change, such as the P2D model, but parameter calibration relies on a large number of experiments, and the model is complex, making it difficult to adapt to the scenario of multiple parallel batteries in energy storage systems and dynamic changes in working conditions, and real-time and generalization are insufficient. Traditional experience and statistical models, such as the Arrhenius model, Weibull distribution model, and Kalman filter, rely on key parameters such as temperature and cycle number to fit the degradation law, although the calculation is simple, but it is difficult to capture nonlinear degradation under multiple stress coupling, and the prediction accuracy is easily affected by environmental fluctuations, suitable for rough estimation under simple working conditions. Therefore, the existing battery degradation path prediction method cannot meet the accuracy and interpretability requirements. SUMMARY

[0004] The present application provides a battery degradation path determination method and device based on a large language model to solve the problem that the existing battery degradation path prediction method cannot meet the accuracy and interpretability requirements.

[0005] According to an aspect of the present application, a battery degradation path determination method based on a large language model is provided, which is applied to a rechargeable lithium battery including at least one battery cell, the method comprising:

[0006] For at least one battery cell, obtain battery-related parameters of the battery cell in the current cycle period, wherein the battery-related parameters include at least voltage sequence, current sequence, temperature sequence, environmental sequence parameters of the operating environment of the battery cell, maximum available capacity of the current cycle period, internal resistance of the battery cell, and charge-discharge rate of the battery cell;

[0007] According to the likelihood function corresponding to each degradation mode and the prior probability, the feature vector corresponding to the battery-related parameters is processed to obtain the posterior probability corresponding to each degradation mode.

[0008] determine a target degradation mode according to the posterior probability corresponding to at least one degradation mode, and process the target degradation mode, a target degradation path corresponding to the target degradation mode, a feature contribution rate corresponding to each index in the feature vector, and the feature vector based on a target language model, to obtain an explainability content corresponding to each degradation information in the target degradation path and a corresponding confidence level;

[0009] determine whether to display the explainability content based on the posterior probability and the confidence level corresponding to the target degradation mode.

[0010] Optionally, after obtaining the battery-related parameters, the method further includes: obtaining a feature vector of the battery cell by processing the battery-related parameters; and obtaining the feature vector of the battery cell by processing the battery-related parameters, including: obtaining a maximum available capacity in the battery-related parameters, and determining a capacity attenuation rate of the battery cell based on the maximum available capacity and a maximum available capacity of the battery cell in an initial state; determining an internal resistance increment of the battery cell according to an internal resistance in the battery-related parameters and an initial internal resistance of the battery cell in the initial state; determining a differential voltage peak shift parameter and a differential current peak shift parameter according to a voltage sequence and a current sequence in the battery-related parameters; determining a first change curve between voltage and capacity according to the voltage sequence and capacity change information in the battery-related parameters; and determining the feature vector of the battery cell by structurally processing the capacity attenuation rate, the internal resistance increment, the differential voltage peak shift parameter, the differential current peak shift parameter, the first change curve, a charge-discharge rate, an environmental sequence parameter, and a temperature sequence.

[0011] Optionally, the feature vector corresponding to the battery-related parameters is processed to obtain a posterior probability corresponding to each degradation mode according to a likelihood function corresponding to each degradation mode and a prior probability, including: for each degradation mode, calling the likelihood function corresponding to the degradation mode, and substituting the feature vector into the likelihood function to obtain a likelihood probability of the battery cell in a current cycle, wherein the likelihood function includes a feature mean value and a feature covariance corresponding to each index associated with the degradation mode, and the feature mean value and the feature covariance are determined based on a feature value under the same index in the current cycle and historical cycles before the current cycle; and determining the posterior probability corresponding to the degradation mode according to the prior probability and the likelihood probability corresponding to the degradation mode, wherein the prior probability is determined based on a sample number corresponding to the degradation mode and all total sample data.

[0012] Optionally, the target degradation mode is determined according to the posterior probability corresponding to at least one degradation mode, including: taking the degradation mode with the largest posterior probability as the target degradation mode; wherein the target degradation mode includes a preset degradation path corresponding to the battery cell, and the preset degradation path includes at least one degradation information.

[0013] Optionally, the feature contribution rate corresponding to each index in the feature vector is determined in the following manner: the feature influence intensity information of the index to which each feature vector belongs is determined by calculating the posterior probability of the target degradation mode and the partial derivative corresponding to each feature vector; and the feature contribution rate corresponding to each index in the feature vector is determined according to the feature influence intensity information of each index and the feature influence intensity information of all indexes.

[0014] Optionally, whether the explainability content is displayed is determined based on the posterior probability and the confidence degree corresponding to the target degradation mode, including: obtaining the confidence degree corresponding to the explainability content of each degradation information; fusing the mean value of all confidence degrees and the posterior probability to obtain the displayable parameter of all explainability contents; and generating the target content based on each degradation information and the corresponding explainability content and displaying the target content in the case where the displayable parameter meets a preset condition.

[0015] Optionally, the method further includes: inputting the battery-related parameters of the battery cell into the pre-trained classification model to output the predicted mode category and the probability information corresponding to each predicted mode category; determining the mode set corresponding to the battery cell according to the probability information of the battery cell; determining whether the target degradation mode is in the corresponding mode set after obtaining the target degradation mode corresponding to each battery cell; and determining the accurate attribute of the target degradation mode of each battery cell according to the number of battery cells in the corresponding mode set and the total number of battery cells.

[0016] According to another aspect of the present application, a battery degradation path determination device based on a large language model is provided, including:

[0017] The battery-related parameter acquisition module is configured to acquire, for at least one battery cell, battery-related parameters of the battery cell in a current cycle, wherein the battery-related parameters at least include a voltage sequence, a current sequence, a temperature sequence, an environment sequence parameter of an operating environment of the battery cell, a maximum available capacity of the current cycle, an internal resistance of the battery cell, and a charge-discharge rate of the battery cell.

[0018] The posterior probability determination module is configured to process the feature vector corresponding to the battery-related parameters according to the likelihood function corresponding to each degradation mode and the prior probability to obtain the posterior probability corresponding to each degradation mode.

[0019] The explainability content and confidence degree determination module is configured to determine the target degradation mode according to the posterior probability corresponding to the at least one degradation mode, and process the target degradation mode, the target degradation path corresponding to the target degradation mode, the feature contribution rate corresponding to each index in the feature vector, and the feature vector based on the target language model to obtain the explainability content corresponding to each degradation information in the target degradation path and the corresponding confidence degree.

[0020] The explainability content display module is configured to determine whether to display the explainability content based on the posterior probability and the confidence degree corresponding to the target degradation mode.

[0021] According to another aspect of the present application, an electronic device is provided, the electronic device comprising:

[0022] at least one processor; and

[0023] a memory connected to the at least one processor in communication; wherein,

[0024] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the battery degradation path determination method based on the large language model according to any one of the embodiments of the present application.

[0025] According to another aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium stores computer instructions for enabling a processor to implement the battery degradation path determination method based on the large language model according to any one of the embodiments of the present application when executed by the processor.

[0026] The technical scheme of the embodiment of the application obtains battery related parameters of the battery cell in the current cycle period, wherein the battery related parameters at least include a voltage sequence, a current sequence, a temperature sequence, an environment sequence parameter of an operating environment of the battery cell, a maximum available capacity of the current cycle period, an internal resistance of the battery cell, and a charge-discharge rate of the battery cell; the feature vector corresponding to the battery related parameters is processed according to the likelihood function corresponding to each degradation mode and the prior probability, to obtain the posterior probability corresponding to each degradation mode; the target degradation mode is determined according to the posterior probability corresponding to at least one degradation mode, and the target degradation mode, the target degradation path corresponding to the target degradation mode, the feature contribution rate corresponding to each index in the feature vector, and the feature vector are processed based on the target language model, to obtain the explainability content corresponding to each degradation information in the target degradation path and the corresponding confidence; whether the explainability content is displayed is determined based on the posterior probability and the confidence corresponding to the target degradation mode. The scheme provides solid data support for accurate identification of degradation modes by using multi-dimensional battery related parameters, improves the scientificity and reliability of target degradation mode identification by combining prior knowledge and real-time data based on the mode determination method of probability statistics, and avoids the black box limitation of traditional degradation prediction models by introducing the feature contribution rate and generating the explainability content, so that the prediction result has clear explainability content. At the same time, through the double check of the posterior probability and the confidence, a high-confidence analysis conclusion can be screened out, invalid or misleading information output is avoided, more instructive basis is provided for operation and maintenance decision of the energy storage battery, the problem that the existing battery degradation path prediction method cannot meet the accuracy and explainability requirements is solved, and the accuracy and explainability of battery degradation path prediction are improved.

[0027] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the application, nor is it intended to limit the scope of the application. Other features of the application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0029] Figure 1 is a flowchart of a battery degradation path determination method based on a large language model provided by the first embodiment of the application;

[0030] Figure 2is a flowchart of a battery degradation path determination method based on a large language model provided by Embodiment Two of the present application;

[0031] Figure 3 is a structural schematic diagram of a battery degradation path determination device based on a large language model provided by Embodiment Three of the present application;

[0032] Figure 4 is a structural schematic diagram of an electronic device implementing the battery degradation path determination method based on a large language model of the present application. DETAILED DESCRIPTION

[0033] In order 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 described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0034] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0035] Embodiment One

[0036] Figure 1 is a flowchart of a battery degradation path determination method based on a large language model provided by Embodiment One of the present application. The present embodiment can be applicable to the case of predicting the battery degradation path. The method can be executed by a battery degradation path determination device based on a large language model. The battery degradation path determination device based on a large language model can be realized in the form of hardware and / or software. The battery degradation path determination device based on a large language model can be configured in an electronic device such as a computer and a server. The method is applied to a rechargeable lithium battery. The rechargeable lithium battery includes at least one battery cell, as shown in the figure. The method includes: Figure 1

[0037] ​S110, for at least one battery cell, obtain battery-related parameters of the battery cell in the current cycle period, wherein the battery-related parameters at least include voltage sequence, current sequence, temperature sequence, environment sequence parameters of the running environment where the battery cell is located, maximum available capacity of the current cycle period, internal resistance of the battery cell, and charge-discharge rate of the battery cell.

[0038] The battery-related parameters can be understood as a multi-dimensional index set representing the running state, performance level and degradation trend of the battery cell, including time sequence parameters dynamically changing over time during the charging and discharging process, such as voltage sequence, current sequence reflecting the electrical characteristics of the battery cell, temperature sequence reflecting the thermal state, and environment sequence parameters affecting the working environment of the battery cell, such as environmental temperature and humidity; and static characteristic parameters that can directly measure the core performance of the battery cell, such as the maximum available capacity directly reflecting the degree of energy storage capacity attenuation of the battery cell in the current cycle period, the internal resistance related to the internal charge transfer efficiency and heating characteristics of the battery cell, and the charge-discharge rate reflecting the charging and discharging rate of the battery cell, which is closely related to the service life and power output of the battery cell. The corresponding key information can be pre-set according to the actual battery degradation path prediction requirements, so that the key information can be matched in the real-time or historical running data of the battery cell when predicting the battery degradation path, so as to represent the working state of the battery cell from multiple dimensions, and provide core data support for subsequent degradation mode recognition, degradation path prediction and health state evaluation.

[0039] Specifically, for at least one target battery cell in the energy storage system, the voltage sequence, current sequence, temperature sequence and other dynamic time sequence parameters of the battery cell in the current cycle period are synchronously collected according to the battery management system (BMS) and the matching voltage, current and temperature sensors. At the same time, the environment sequence parameters of the running environment where the battery cell is located are obtained by combining the environmental monitoring equipment, and the maximum available capacity and internal resistance of the battery cell in the cycle period are determined by means of capacity test and alternating current impedance test. The corresponding charge-discharge rate is extracted combined with the charge-discharge control strategy, and finally the multi-dimensional and multi-type parameters are preprocessed such as normalization and time sequence alignment to form a complete battery cell state feature data set, so as to obtain the battery-related parameters of the battery cell in the current cycle period.

[0040] In one embodiment, the data can be obtained by BMS sampling log, test bench acquisition system or battery test equipment, and each data has a unified time stamp t. For any battery cell in the rechargeable lithium battery, the corresponding battery-related parameters are obtained, including multiple time sequence data as follows: (1) voltage sequence: , is the terminal voltage at time t, and T is the time sequence length; (2) current sequence: , is the working current at time t, and the charge can be set as positive and the discharge as negative; (3) temperature sequence: , (3) Cell temperature; (4) Environmental sequence data of the operating environment of the cell: For example, ambient temperature and humidity; (5) Cell operating condition information This includes the maximum available capacity for the current cycle, the internal resistance of the cell, and the charge / discharge rate of the cell. It is understandable that, since different signals may have different sampling frequencies (e.g., the BMS records data every 1 second, while environmental data is recorded every 10 seconds), synchronization can be achieved using timestamp alignment and resampling. The specific processing includes: constructing a unified time grid based on the time resolution of the voltage / current sequence. Environmental records Perform interpolation or step preservation to make each Each corresponds to an environment record. For missing data points, appropriate interpolation (such as linear interpolation) or value preservation is used; obviously abnormal data (such as voltage exceeding limits or temperature exceeding range) are filtered or marked. After the above preprocessing, the system obtains an aligned multidimensional time series.

[0041] ;

[0042] This provides a unified input for subsequent feature engineering and degradation pattern reasoning.

[0043] In this embodiment, the collected parameters cover both the dynamic electrical and thermal characteristics time-series data of the cell itself during the charging and discharging process and the influencing factors of the external environment. They also include core performance indicators such as capacity, internal resistance, and rate capability, thus achieving a multi-dimensional characterization of the cell's operating status. This provides a rich and accurate data foundation for subsequent degradation mode identification and degradation path prediction, effectively avoiding the one-sidedness of assessing the cell's health status with a single parameter or static parameter, and improving the accuracy and reliability of subsequent degradation analysis.

[0044] S120. Based on the likelihood function and prior probability corresponding to each degradation mode, process the feature vectors corresponding to the battery-related parameters to obtain the posterior probability corresponding to each degradation mode.

[0045] It should be noted that in the technical system for predicting battery degradation paths, degradation modes can be specifically understood as typical mechanistic events triggered by different electrochemical or physical mechanisms during the cyclic charging and discharging, storage, and operation of the battery cell. Degradation modes can be set according to electrochemical mechanisms. For lithium-ion battery systems, the set of degradation modes can be defined as: For example, if K is 9, the meanings of each mode are as follows: SEI membrane growth (Solid Electrolyte Interphase Growth); : Loss of Active Lithium (LAM-Li); : Loss of Active Material (LAM-Cathode); : Anode active material decay (LAM-Anode); : Particle Cracking and Structural Damage. Electrolyte Decomposition; Lithium dendrite formation. Current collector corrosion. Side reactions related to transition metals, each with unique cell performance degradation patterns and parameter variation characteristics. Prior probability can be understood as the probability of various degradation modes occurring without current cell state information, based on battery electrochemical mechanisms, historical degradation data, or expert experience; it reflects the initial judgment of the likelihood of different degradation modes occurring. The likelihood function can be understood as the probability of the currently collected cell feature vector appearing given a specific degradation mode; it is used to quantify the degree of matching between the degradation mode and real-time parameter characteristics. The expression is as follows:

[0046] ;

[0047] Here, z is the feature vector, which consists of features such as capacity decay rate, internal resistance increment, differential voltage peak shift, high-temperature residence rate, and high-rate cycling rate. The feature vector is obtained by structuring the relevant battery parameters. ; Let i be the i-th degradation mode; the set of modes is denoted as . d is the dimension of the feature vector, that is, the number of components in z; Indicates in pattern The eigenmean vector below: ; Indicates in pattern The characteristic covariance matrix (d×d real symmetric positive definite matrix) is as follows: ; Covariance matrix determinant of the matrix of the features, used for the normalization factor . is the inverse of the covariance matrix, used to measure the "Mahalanobis distance" of a feature from the mean. is the square of the Mahalanobis distance, indicating the degree of deviation of the feature vector z from the mode .

[0048] The posterior probability can be specifically understood as the probability of the actual occurrence of each degradation mode under the condition that the current battery feature vector is known, which is calculated by combining the prior probability and the likelihood function through the Bayes theorem, and aims to correct the initial judgment of the degradation mode based on real-time data, and finally determine the dominant degradation reason of the current battery. The calculation formula of the posterior probability is as follows:

[0049] ;

[0050] wherein, represents the prior probability under the mode; k represents the index of the degradation mode.

[0051] Specifically, based on the battery electrochemical mechanism and historical degradation data, the various degradation modes of the battery can be determined, a corresponding degradation mode set can be formed, the prior probability corresponding to each degradation mode can be determined, and the likelihood function between each degradation mode and the feature vector can be constructed. Then, the feature vector representing the current state of the battery is substituted into the likelihood function corresponding to each degradation mode to obtain the likelihood probability, and then the prior probability model is calculated through the Bayesian theorem and other probability statistical methods, and finally the posterior probability of each degradation mode under the current state of the battery is obtained.

[0052] In the embodiment, the prior mechanism knowledge is combined with the real-time feature vector data instead of relying solely on data fitting, which improves the scientificity and interpretability of the degradation mode determination, and at the same time, the occurrence probability of each degradation mode is determined through the probabilistic quantization method, which provides accurate and quantitative judgment basis for subsequent screening of the dominant target degradation mode, and effectively reduces the misjudgment risk caused by single feature or experience judgment.

[0053] Optionally, after obtaining the battery-related parameters, the method further comprises: obtaining a feature vector of the battery cell by processing the battery-related parameters; and obtaining the feature vector of the battery cell by processing the battery-related parameters comprises: obtaining a maximum available capacity in the battery-related parameters, and determining a capacity attenuation rate of the battery cell based on the maximum available capacity and a maximum available capacity of the battery cell in an initial state; determining an internal resistance increment of the battery cell according to an internal resistance in the battery-related parameters and an initial internal resistance of the battery cell in the initial state; determining a differential voltage peak shift parameter and a differential current peak shift parameter according to a voltage sequence and a current sequence in the battery-related parameters; determining a first change curve between the voltage and the capacity according to the voltage sequence and capacity change information in the battery-related parameters; and determining the feature vector of the battery cell by structurally processing the capacity attenuation rate, the internal resistance increment, the differential voltage peak shift parameter, the differential current peak shift parameter, the first change curve, a charge-discharge rate, an environmental sequence parameter, and a temperature sequence.

[0054] Specifically, after obtaining the multi-dimensional battery-related parameters of the battery cell, the maximum available capacity in the initial state is taken as a reference to calculate the capacity attenuation rate corresponding to the maximum available capacity in the current cycle period, and the capacity attenuation rate The calculation formula of the capacity attenuation rate is as follows:

[0055] ;

[0056] wherein, represents the maximum available capacity in the initial state, which is generally taken as the average value of the 0th cycle or the previous several cycles and is used as a capacity reference, represents the maximum available capacity in the current cycle period, is the discharge capacity, which is obtained by integrating the current in a complete constant-current discharge process, The calculation formula of the capacity attenuation rate is as follows:

[0057] ;

[0058] The internal resistance increment is calculated in combination with the initial internal resistance and the current internal resistance, and the internal resistance increment The calculation formula of the internal resistance increment is as follows:

[0059] ;

[0060] wherein, represents the initial internal resistance, which is generally measured by HPPC (Hybrid Pulse Power Characterization) test or current pulse method in the initial stage of the life test; represents the internal resistance value measured in the current life stage, which can be obtained by calculation in the following manner:

[0061] ;

[0062] in, This represents the change in terminal voltage at the instant of the current pulse. This represents the pulse amplitude.

[0063] Subsequently, differential voltage peak shift parameters and differential current peak shift parameters were extracted based on the voltage and current sequences. A voltage-capacity first change curve was then plotted based on the voltage sequence and cell capacity change information. It should be noted that the differential voltage peak shift is defined as:

[0064] ;

[0065] in, Differential voltage curve at the beginning of the lifespan (cycle=0) The location of the peak voltage in the middle; This refers to the voltage position of the same peak in the differential voltage curve for the current cycle. Differential voltage characteristics can be pre-constructed. or its equivalent form To determine the differential voltage characteristics, first acquire the voltage V and capacity Q curves during the constant current charging / discharging process, smooth the discrete data (e.g., using Savitzky-Golay filtering), and then calculate the discrete differential:

[0066] ;

[0067] Record the peak positions of all peaks at the beginning of the lifespan. Record the corresponding peak position in the current cycle. ; Calculate peak position shift:

[0068] .

[0069] Finally, the core indicators such as capacitance decay rate, internal resistance increment, differential voltage peak shift parameter, differential current peak shift parameter, first change curve, charge / discharge rate, environmental sequence parameters, and temperature sequence are structured and integrated to form a feature vector that can comprehensively characterize the cell state.

[0070] In the embodiment, by processing and refining the original parameters, the dispersed time sequence parameters and static parameters are converted into feature indicators with clear physical meaning, which not only retains the core degradation representation of cell capacity attenuation, internal resistance change, etc., but also incorporates the key information reflecting the electrochemical characteristics such as differential peak shift, voltage-capacity curve, etc., while taking into account the external influencing factors such as charging and discharging rate, environment and temperature, etc. The constructed feature vector has comprehensiveness and pertinence, can accurately depict the actual state of the cell, provides high-quality data support for subsequent degradation mode recognition and degradation path prediction, and effectively improves the analysis accuracy and reliability of the subsequent algorithm model.

[0071] S130, determining a target degradation mode according to the posterior probability corresponding to at least one degradation mode, and processing the target degradation mode, the target degradation path corresponding to the target degradation mode, the feature contribution rate corresponding to each index in the feature vector, and the feature vector based on the target language model, to obtain the explainability content corresponding to each degradation information in the target degradation path and the corresponding confidence.

[0072] It should be noted that in the battery degradation path prediction system, the target degradation path can be specifically understood as a typical performance degradation evolution trajectory corresponding to the current dominant target degradation mode of the cell, which is summarized based on electrochemical mechanism and a large number of historical degradation data, and is a sequential causal chain sequence composed of observable operating features, potential mechanism events and degradation modes, used to describe the whole process of gradually producing internal material changes of the cell under external stress and finally showing performance degradation. For example, the degradation path is formalized as: wherein, represents a causal dependence or conditional probability relationship. The feature contribution rate can be understood as a quantitative weight measuring the influence degree of each parameter index in the feature vector on the target degradation mode, reflecting the contribution size of different parameters (such as temperature, charging and discharging rate, internal resistance increment, etc.) in driving the development process of the degradation path. The higher the contribution rate of a parameter, the more significant the impact on degradation.

[0073] The explainability content refers to the analysis text generated based on the target language model, combined with the target degradation mode, the target degradation path, the feature contribution rate and the feature vector, which has clear causal logic, and its content covers the core causes of degradation, key influencing parameters, evolution mechanism at each stage and potential risks, etc. The confidence is a quantitative probability value output with the explainability content, which comprehensively considers the posterior probability of the target degradation mode, the matching degree of the feature vector and the degradation path, the stability of the model prediction, etc., and is used to represent the reliability of the explainability content. The higher the confidence, the higher the degree of fit between the analysis conclusion and the actual degradation state of the cell.

[0074] Specifically, according to the posterior probability of each type of degradation mode, one or more modes with the highest probability or meeting the preset threshold are selected as the target degradation mode, and then the typical degradation path corresponding to the target degradation mode is retrieved. Combined with the feature contribution rate (i.e., the influence weight of each feature component on the degradation mode) of each index in the feature vector and the feature vector representing the current state of the battery cell, the target language model is input for joint processing. Based on the built-in electrochemical mechanism knowledge base and language generation logic of the model, the explainability content corresponding to each stage of the degradation information in the target degradation path is output, such as the degradation inducement, development trend, key impact parameters, etc. At the same time, the model's probability output mechanism gives these contents corresponding confidence.

[0075] For example, when the target degradation mode is determined to be After that, the structured information is determined, including: Top-1 mode and its posterior probability ; the corresponding degradation path P, such as high temperature SEI thickening capacity attenuation; the list of main contribution factors and their contribution rates ; the key values in the feature vector z, such as , , . The obtained structured information is transmitted to the preset target language model, which generates explainability content under the constraint of the causal chain template, such as: {“Long-term operation in high temperature area causes SEI film to continuously thicken; internal resistance growth is consistent with SEI thickening; differential voltage peak drifts by 25mV; further confirms SEI dominant degradation mode.” It should be noted that when the target language model generates explainability content based on the target degradation path, the output token sequence , where T is the number of tokens in the entire explanation paragraph. When generating each token , the target language model outputs its conditional probability , where is the model input information, i.e., the structured information. In this embodiment, token refers to the smallest processing unit of the target language model when generating text, which can be an English word fragment (subword), Chinese character or word piece (according to BPE or SentencePiece segmentation rules), punctuation, number, unit symbol, etc. It is a basic symbol unit used for encoding and generation inside the model; the paragraph average token confidence refers to the average of all token probabilities generated by the model, which is the confidence of the entire explanation :

[0076] ;

[0077] If the internal reasoning of the explanation paragraph is consistent and the model confidence is high, all Larger, therefore, Larger; if the model is uncertain or produces contradictory inferences in part steps, the probability of the corresponding token is lower, thereby pulling down the overall average; because the average of all tokens is taken, it can reflect the confidence consistency of the entire explanation at the "sentence level", rather than the accidental confidence of a few tokens, and can comprehensively reflect the internal consistency of the entire paragraph on the semantic chain, so it can be called the paragraph average token confidence.

[0078] In the embodiment, the target degradation mode screening method based on posterior probability ensures the accuracy and objectivity of pattern recognition, and the introduction of feature contribution rate and the application of target language model break through the black box limitation of traditional degradation prediction model, which not only can determine the current dominant degradation mode and development path of the battery cell, but also can generate interpretable content with causal logic, so that the technical personnel can clearly know the key driving factors and evolution law of degradation, and the confidence degree also provides a quantitative basis for judging the reliability of the explanation content, which greatly improves the guidance of the degradation analysis result to the battery operation and maintenance decision.

[0079] Optionally, the target degradation mode is determined according to the posterior probability corresponding to the at least one degradation mode, including: taking the degradation mode with the maximum posterior probability as the target degradation mode; wherein the target degradation mode includes a preset degradation path corresponding to the battery cell, and the preset degradation path includes at least one degradation information.

[0080] It should be noted that in the battery degradation path prediction system, the degradation information can be understood as the key node and stage feature data and mechanism description that characterize the performance degradation process of the battery cell in the preset degradation path, which is extracted according to the electrochemical mechanism and a large number of historical degradation data, and contains the core performance parameter threshold of the battery cell at different stages under a certain degradation mode, such as the node of capacity decay to 80% of the initial value and the critical point of internal resistance growth exceeding 20%, and also covers the degradation performance and internal cause at the corresponding stage, such as the slow capacity decay feature at the SEI film growth stage and the internal resistance mutation mechanism at the lithium dendrite generation stage. These information are arranged in order according to the degradation process, which constitutes the core content of the target degradation path and is an important basis for generating interpretable analysis text and predicting the performance evolution trend of the battery cell.

[0081] Specifically, after obtaining the posterior probabilities corresponding to various battery degradation modes, the degradation mode with the maximum posterior probability is directly selected as the target degradation mode by comparing the sizes of the probability values, and the preset degradation path of the target degradation mode is retrieved, which is constructed based on the electrochemical mechanism and historical degradation data. The path contains at least one key degradation information of the performance degradation of the battery cell under the degradation mode, such as the capacity decay threshold at a certain stage and the internal resistance mutation node.

[0082] In the embodiment, by setting the maximum posterior probability as the basis for determining the target degradation mode, the quantitative logic of probability statistics is followed, the dominant degradation cause of the current battery cell is quickly locked, misjudgment caused by multi-mode confusion is avoided, the matching retrieval of the preset degradation path can directly provide a standardized evolution trajectory reference for subsequent degradation analysis, the key degradation information built in the path also lays a foundation for accurately positioning the current degradation stage of the battery cell and predicting the subsequent performance change trend, and the entire determination process is simple and efficient, with clear logic, and is suitable for engineering application scenarios.

[0083] Optionally, the feature contribution rate corresponding to each index in the feature vector is determined in the following manner: the feature influence intensity information of the index to which each feature vector belongs is determined by calculating the posterior probability of the target degradation mode and the partial derivative corresponding to each feature vector; and the feature contribution rate corresponding to each index in the feature vector is determined according to the feature influence intensity information of each index and the feature influence intensity information of all indexes. In the embodiment, in order to measure the influence degree of each feature component on the posterior probability of the mode, the gradient sensitivity and the normalized contribution rate are defined as follows: The posterior probability of the mode The posterior probability of the mode

[0084] ;

[0085] wherein, represents the jth component of the feature vector z, such as the capacity decay rate, the internal resistance increment, a certain differential voltage peak shift or a certain temperature residence ratio, etc. is the posterior probability of the mode . is the partial derivative of the posterior probability with respect to the feature component , indicating the influence direction and size of a slight change in the feature component on . is the absolute value of the gradient sensitivity of the feature component to the mode , indicating the influence intensity, and the greater the value, the more important the feature to the determination of the mode . is the normalized result of in the feature dimension, satisfying: , which can be directly sorted and displayed as the contribution rate

[0086] Specifically, after determining the target degradation mode, the partial derivative of the posterior probability to each index in the feature vector is calculated based on the posterior probability function of the mode, and the feature influence intensity information of each index is obtained according to the numerical size and positive and negative of the partial derivative, which represents the influence degree and direction of the change of a single index on the posterior probability. Then, the feature influence intensity of all indexes is normalized, and the feature contribution rate of each index in the feature vector is finally determined by calculating the ratio of the influence intensity of a single index to the total influence intensity of all indexes.

[0087] In this embodiment, the influence of the index on the degradation mode is quantified by the mathematical properties of the partial derivative, which makes the calculation of the feature contribution rate have rigorous mathematical logic support, rather than relying on empirical assignment, thereby improving the objectivity and accuracy of the results. Meanwhile, the normalization processing makes the index contribution rates of different dimensions comparable, which can clearly define the key indexes and secondary indexes driving the target degradation mode, provide quantitative basis for subsequent explainable content generation, and further enhance the scientificity and persuasiveness of the entire degradation prediction scheme.

[0088] S140, determining whether to display the explainable content based on the posterior probability and the confidence degree corresponding to the target degradation mode.

[0089] Specifically, the double threshold standards of the posterior probability and the confidence degree are preset, and the posterior probability corresponding to the target degradation mode and the confidence degree corresponding to the explainable content are compared with the corresponding threshold respectively. Only when both of them meet or at least meet the preset determination condition, the generated explainable content is output and displayed, otherwise it is not displayed.

[0090] In this embodiment, through the double check of the posterior probability and the confidence degree, the degradation analysis results with low posterior probability and insufficient confidence can be effectively filtered out, avoiding misleading the operation and maintenance decision with inaccurate or doubtful information. At the same time, the determination mechanism of the double threshold takes into account the reliability of the target degradation mode and the effectiveness of the explainable content, so that the final displayed analysis conclusion has scientificity and rigor, and the practical value and decision reference significance of the entire degradation prediction scheme are improved.

[0091] Optionally, whether to display the explainable content is determined based on the posterior probability and the confidence degree corresponding to the target degradation mode, including: obtaining the confidence degree corresponding to the explainable content of each degradation information; fusing all confidence degrees after mean processing with the posterior probability to obtain the displayable parameters of all explainable contents; and generating target content based on each degradation information and the corresponding explainable content and displaying the target content in the case that the displayable parameters meet the preset condition.

[0092] Specifically, the confidence of the explainability content corresponding to each degradation information in the target degradation path is extracted, the mean of the confidences is calculated to obtain an overall confidence level, the mean is fused with the posterior probability corresponding to the target degradation mode to obtain a displayable parameter measuring the overall reliability of the explainability content, and then the parameter is compared with a preset determination condition. If the parameter meets the condition, each degradation information and the corresponding explainability content are integrated to generate and display the target content. The expression of the fusion operation is as follows:

[0093] ;

[0094] Wherein, conf represents the displayable parameter, and a is a preset weight coefficient. For example, a can be 0.7. If conf is greater than or equal to a first threshold value, it is determined that the displayable parameter meets the preset condition, and the explainability content can be automatically pushed and displayed. If conf is less than the first threshold value and greater than or equal to a second threshold value, it is determined that the displayable parameter does not meet the preset condition, and the explainability content can be marked as “further review”. If conf is less than the second threshold value, it is determined that the displayable parameter does not meet the preset condition, and the explainability content can be deleted, and the explainability content and the corresponding confidence are re-determined. The first threshold value is greater than the second threshold value, the first threshold value can be set to 0.8, and the second threshold value can be set to 0.6. It should be noted that the thresholds can be set according to requirements, and are not limited herein.

[0095] In the embodiment, the mean of the confidences of the degradation information is processed to comprehensively reflect the overall reliability of the explainability content, and the posterior probability of the target degradation mode is fused for determination, rather than single-dimensional verification. The displayable parameter has both the reliability of mode recognition and the reliability of content generation, effectively avoiding misjudgment caused by fluctuations in the confidence of individual degradation information. The target content is generated and displayed based on the parameter meeting the condition, which can guarantee the rigor and practicality of the final output information and provide high-quality reference for battery operation and maintenance decisions.

[0096] On the basis of the above embodiment, the method further comprises: inputting the battery-related parameters of the battery cell into a pre-trained classification model to output a predicted mode category and probability information corresponding to each predicted mode category; determining a mode set corresponding to the battery cell according to the probability information of the battery cell; determining whether the target degradation mode is in the corresponding mode set after obtaining the target degradation mode corresponding to each battery cell; and determining the accurate attribute of the target degradation mode of each battery cell according to the number of battery cells in the corresponding mode set and the total number of battery cells.

[0097] Specifically, the collected battery cell related parameters are input into a classification model pre-trained based on a large amount of battery degradation data, the prediction mode category corresponding to the battery cell and the probability information corresponding to each category are output by the model, then the category meeting the preset threshold is selected according to the probability information, and a mode set of the battery cell is constructed. The classification model can be a model constructed based on a convolutional neural network, including but not limited to ResNet, VGG and 1DCNN. After determining the target degradation mode of each battery cell, it is checked whether the target degradation mode exists in the corresponding mode set, and finally the number of battery cells with the target degradation mode in the corresponding mode set is counted, and the ratio of the number to the total number of battery cells is calculated to determine the accuracy of the target degradation mode.

[0098] For example, there are N battery cells in a rechargeable lithium battery. The classification model determines the prediction mode category meeting the preset threshold corresponding to each of the N battery cells, and the real degradation mode set of the i-th battery cell is denoted as The first two degradation mode sets output by the model in descending order of posterior probability are denoted as Then, it is determined whether the first two degradation mode sets of each battery cell hit by using the following function, and the hit degradation mode determination function is:

[0099] ;

[0100] The Top-2 hit rate of the degradation mode is defined as:

[0101] ;

[0102] If , the accuracy of the target degradation mode meets the preset accuracy requirement.

[0103] In this embodiment, the pre-prediction by the classification model and the construction of the mode set provide an additional checking dimension for the target degradation mode, avoiding the limitations of single method determination. At the same time, the accuracy of the target degradation mode is quantified by statistical analysis of the group of battery cells, which not only improves the reliability of single battery cell degradation mode recognition, but also reflects the general degree of the degradation mode in the group of battery cells, providing a decision basis with individual precision and group reference for batch operation and degradation prevention of the battery pack.

[0104] The technical scheme of the embodiment is characterized in that, for at least one battery cell, battery-related parameters of the battery cell in a current cycle period are acquired, wherein the battery-related parameters at least include a voltage sequence, a current sequence, a temperature sequence, an environment sequence parameter of an operating environment of the battery cell, a maximum available capacity of the current cycle period, an internal resistance of the battery cell, and a charge-discharge rate of the battery cell; a feature vector corresponding to the battery-related parameters is processed according to a likelihood function corresponding to each degradation mode and a prior probability, to obtain a posterior probability corresponding to each degradation mode; a target degradation mode is determined according to the posterior probability corresponding to at least one degradation mode, and the target degradation mode, a target degradation path corresponding to the target degradation mode, a feature contribution rate corresponding to each index in the feature vector, and the feature vector are processed based on a target language model, to obtain an explainability content corresponding to each degradation information in the target degradation path and a corresponding confidence degree; and whether the explainability content is displayed is determined based on the posterior probability corresponding to the target degradation mode and the confidence degree. The scheme provides solid data support for accurate identification of degradation modes by using multi-dimensional battery-related parameters, improves the scientificity and reliability of target degradation mode identification by combining prior knowledge and real-time data based on a mode determination method of probability statistics, avoids the black box limitation of a traditional degradation prediction model by introducing a feature contribution rate and generating an explainability content, so that the prediction result has clear explainability content, and through double checking of the posterior probability and the confidence degree, a high-confidence analysis conclusion can be screened out, invalid or misleading information output is avoided, more instructive basis is provided for operation and maintenance decision of energy storage batteries, and the problem that existing battery degradation path prediction methods cannot meet the accuracy and explainability requirements is solved, and the accuracy and explainability of battery degradation path prediction are improved.

[0105] Embodiment two

[0106] Figure 2 is a flowchart of a battery degradation path determination method based on a large language model provided by the embodiment two of the application. The method of the embodiment is a further optimization of the method of the above-mentioned embodiment. Optionally, for each degradation mode, a likelihood function corresponding to the degradation mode is called, and the feature vector is substituted into the likelihood function to obtain a likelihood probability of the battery cell in the current cycle period. The likelihood function includes a feature mean value and a feature covariance corresponding to each index associated with the degradation mode. The feature mean value and the feature covariance are determined based on the feature values in the current cycle period and the same index in the historical cycle period before the current cycle period. The posterior probability corresponding to the degradation mode is determined according to the prior probability and the likelihood probability corresponding to the degradation mode. The prior probability is determined based on the sample quantity corresponding to the degradation mode and all total sample data. Figure 2 As shown in the figure, the method includes:

[0107] S210, for at least one battery cell, obtain battery-related parameters of the battery cell in the current cycle period, wherein the battery-related parameters at least include voltage sequence, current sequence, temperature sequence, environmental sequence parameters of the operating environment of the battery cell, maximum available capacity of the current cycle period, internal resistance of the battery cell, and charge-discharge rate of the battery cell.

[0108] S220, for each degradation mode, call the likelihood function corresponding to the degradation mode, and substitute the feature vector into the likelihood function to obtain the likelihood probability of the battery cell in the current cycle period, wherein the likelihood function includes the feature mean and feature covariance corresponding to each index associated with the degradation mode, and the feature mean and feature covariance are determined based on the feature values of the same index in the current cycle period and the historical cycle period before the current cycle period.

[0109] Specifically, for each preset degradation mode of the battery cell, the exclusive likelihood function previously constructed is called first, which is embedded with the feature mean and feature covariance calculated from the feature values of the index associated with the degradation mode in the current cycle period and the historical cycle period before that, and then the feature vector representing the current state of the battery cell is substituted into the likelihood function for operation, and finally the likelihood probability of the battery cell in the current cycle period corresponding to the degradation mode is obtained.

[0110] In this embodiment, the core parameters (feature mean and feature covariance) of the likelihood function are determined according to the historical and current data of the whole cycle period, rather than relying on empirical values only, so that the function can accurately fit the real parameter distribution characteristics of the degradation mode. At the same time, the design of matching exclusive likelihood function for each degradation mode greatly improves the matching accuracy between the feature vector and the degradation mode, thereby providing more reliable basic data for the subsequent calculation of posterior probability and enhancing the scientificity and accuracy of the whole degradation mode recognition process.

[0111] S230, determine the posterior probability corresponding to the degradation mode according to the prior probability and the likelihood probability corresponding to the degradation mode; wherein the prior probability is determined based on the sample number corresponding to the degradation mode and the total sample data of all degradation modes.

[0112] Specifically, the number of historical samples corresponding to each type of degradation mode and the total sample data of all degradation modes are counted, and the prior probability of each degradation mode is determined by calculating the ratio of the two. The prior probability calculation formula is as follows: In combination with the corresponding likelihood probability obtained by substituting the feature vector into the exclusive likelihood function, the posterior probability of each degradation mode is finally obtained by joint operation according to Bayes' theorem.

[0113] In the embodiment, the prior probability is obtained based on the statistics of real historical sample data instead of subjective experience assignment, ensuring the objectivity and rationality of the prior probability. Meanwhile, the posterior probability is derived through rigorous Bayes theorem combining the likelihood probability, realizing the organic fusion of the prior knowledge and the current battery state data, enabling the posterior probability to more accurately quantify the possibility of the current battery corresponding to each type of degradation mode, and providing a scientific and reliable quantitative basis for the subsequent determination of the target degradation mode.

[0114] S240, determining the target degradation mode according to the posterior probability corresponding to at least one degradation mode, and processing the target degradation mode, the target degradation path corresponding to the target degradation mode, the feature contribution rate corresponding to each index in the feature vector, and the feature vector based on the target language model, to obtain the explainability content corresponding to each degradation information in the target degradation path and the corresponding confidence.

[0115] S250, determining whether to display the explainability content based on the posterior probability and the confidence corresponding to the target degradation mode.

[0116] The technical scheme of the embodiment is characterized in that, for at least one battery cell, the voltage sequence, the current sequence, the temperature sequence, the operating environment sequence parameter, the maximum available capacity, the internal resistance and the charge-discharge rate and other multi-dimensional battery-related parameters in the current cycle are collected and integrated into a feature vector; for each preset degradation mode, the exclusive likelihood function (the feature mean and the feature covariance of the function are calculated from the same index feature value of the current and historical cycles) is called, and the feature vector is substituted into the function to obtain the corresponding likelihood probability; then, according to the prior probability determined by the proportion of the number of historical samples, the posterior probability of each degradation mode is calculated by Bayes' theorem in combination with the likelihood probability; then, the target degradation mode is determined according to the posterior probability, and the target language model is used to fuse the target degradation mode, the corresponding target degradation path, the feature contribution rate of each index of the feature vector and the feature vector itself, to generate the explainability content and the confidence of each degradation information in the target degradation path; finally, whether to display the explainability content is determined based on the posterior probability and the confidence of the target degradation mode. The scheme uses multi-dimensional battery-related parameters to provide solid data support for accurate identification of degradation modes, and the mode determination method based on probability statistics combines prior knowledge and real-time data, thereby improving the scientificity and reliability of target degradation mode identification. The introduction of the feature contribution rate and the generation of the explainability content avoid the black box limitation of the traditional degradation prediction model, so that the prediction result has clear explainability content. Through the double check of the posterior probability and the confidence, a high-confidence analysis conclusion can be screened out, invalid or misleading information output is avoided, more instructive basis is provided for operation and maintenance decision of the energy storage battery, and the problem that the existing battery degradation path prediction method cannot meet the accuracy and explainability requirements is solved, thereby improving the accuracy and explainability of battery degradation path prediction.

[0117] Embodiment three

[0118] Figure 3 is a structural schematic diagram of a battery degradation path determination device based on a large language model provided by the embodiment three of the application. As shown in the figure, Figure 3 the device comprises:

[0119] The battery-related parameter acquisition module 310 is configured to acquire, for at least one battery cell, the battery-related parameters of the battery cell in the current cycle, wherein the battery-related parameters at least include the voltage sequence, the current sequence, the temperature sequence, the environment sequence parameter of the operating environment of the battery cell, the maximum available capacity of the current cycle, the internal resistance of the battery cell and the charge-discharge rate of the battery cell.

[0120] The posterior probability determination module 320 is configured to process the feature vector corresponding to the battery-related parameters according to the likelihood function corresponding to each degradation mode and the prior probability, to obtain the posterior probability corresponding to each degradation mode.

[0121] The explainable content and confidence determination module 330 is configured to determine a target degradation mode according to a posterior probability corresponding to at least one degradation mode, and process the target degradation mode, a target degradation path corresponding to the target degradation mode, a feature contribution rate corresponding to each index in the feature vector, and the feature vector based on a target language model, to obtain explainable content corresponding to each degradation information in the target degradation path and a corresponding confidence.

[0122] The explainable content display module 340 is configured to determine whether to display the explainable content based on the posterior probability and the confidence corresponding to the target degradation mode.

[0123] The technical scheme of the embodiment is configured to obtain, by the battery-related parameter acquisition module, battery-related parameters of the battery cell in the current cycle period, where the battery-related parameters at least include a voltage sequence, a current sequence, a temperature sequence, an environment sequence parameter of an operating environment of the battery cell, a maximum available capacity of the current cycle period, an internal resistance of the battery cell, and a charge-discharge rate of the battery cell; the posterior probability determination module processes a feature vector corresponding to the battery-related parameters according to a likelihood function corresponding to each degradation mode and a prior probability, to obtain a posterior probability corresponding to each degradation mode; the explainable content and confidence determination module determines a target degradation mode according to the posterior probability corresponding to at least one degradation mode, and processes the target degradation mode, a target degradation path corresponding to the target degradation mode, a feature contribution rate corresponding to each index in the feature vector, and the feature vector based on a target language model, to obtain explainable content corresponding to each degradation information in the target degradation path and a corresponding confidence; and the explainable content display module determines whether to display the explainable content based on the posterior probability and the confidence corresponding to the target degradation mode. The scheme provides solid data support for accurate identification of degradation modes by using multi-dimensional battery-related parameters, improves the scientificity and reliability of target degradation mode identification by combining a mode determination method based on probability statistics with prior knowledge and real-time data, avoids the black box limitation of traditional degradation prediction models by introducing a feature contribution rate and generating explainable content, so that the prediction result has clear explainable content, and through double verification of the posterior probability and the confidence, an analysis conclusion with high credibility can be screened out to avoid invalid or misleading information output, provide a more instructive basis for operation and maintenance decisions of energy storage batteries, and solve the problem that existing battery degradation path prediction methods cannot meet the accuracy and explainability requirements, thereby improving the accuracy and explainability of battery degradation path prediction.

[0124] On the basis of the above-mentioned embodiments, optionally, the posterior probability determination module 320 comprises a feature vector determination unit. After obtaining the battery-related parameters, the feature vector determination unit is configured to obtain the feature vector of the battery cell by processing the battery-related parameters. The feature vector determination unit is specifically configured to obtain the maximum available capacity in the battery-related parameters, and determine the capacity attenuation rate of the battery cell based on the maximum available capacity and the maximum available capacity of the battery cell in the initial state; determine the internal resistance increment of the battery cell according to the internal resistance in the battery-related parameters and the initial internal resistance of the battery cell in the initial state; determine the differential voltage peak shift parameter and the differential current peak shift parameter according to the voltage sequence and the current sequence in the battery-related parameters; determine the first change curve between the voltage and the capacity according to the voltage sequence in the battery-related parameters and the capacity change information of the battery cell; and determine the feature vector of the battery cell by structurally processing the capacity attenuation rate, the internal resistance increment, the differential voltage peak shift parameter, the differential current peak shift parameter, the first change curve, the charge-discharge rate, the environmental sequence parameter, and the temperature sequence.

[0125] Optionally, the posterior probability determination module 320 comprises a posterior probability determination unit, which is configured to, for each degradation mode, call the likelihood function corresponding to the degradation mode, and substitute the feature vector into the likelihood function to obtain the likelihood probability of the battery cell in the current cycle period, wherein the likelihood function comprises the feature mean and the feature covariance corresponding to each index associated with the degradation mode, and the feature mean and the feature covariance are determined based on the feature values of the same index in the current cycle period and the historical cycle periods before the current cycle period; determine the posterior probability corresponding to the degradation mode according to the prior probability corresponding to the degradation mode and the likelihood probability; wherein the prior probability is determined based on the sample number corresponding to the degradation mode and all total sample data.

[0126] Optionally, the explainability content and confidence determination module 330 comprises a target degradation mode determination unit and an explainability content and confidence determination unit. The target degradation mode determination unit is configured to take the degradation mode with the maximum posterior probability as the target degradation mode; wherein the target degradation mode comprises a preset degradation path corresponding to the battery cell, and the preset degradation path comprises at least one degradation information. The explainability content and confidence determination unit is configured to determine the feature influence intensity information of each index by calculating the posterior probability of the target degradation mode and the partial derivative corresponding to each feature vector; and determine the feature contribution rate corresponding to each index in the feature vector according to the feature influence intensity information of each index and the feature influence intensity information of all indexes.

[0127] Optionally, the explainable content display module 340 is specifically configured to acquire a confidence degree corresponding to each explainable content of the degradation information; perform mean processing on all confidence degrees, and fuse the mean processed confidence degrees with a posterior probability to obtain displayable parameters of all explainable contents; and generate target content based on each degradation information and the corresponding explainable content and display the target content in a case where the displayable parameters meet preset conditions.

[0128] Optionally, the device is further configured to input the battery related parameters of the battery cell into the pre-trained classification model to output a prediction mode category and probability information corresponding to each prediction mode category; determine a mode set corresponding to the battery cell according to the probability information of the battery cell; determine whether the target degradation mode is in the corresponding mode set after obtaining the target degradation mode corresponding to each battery cell; and determine the accurate attribute of the target degradation mode of each battery cell according to the number of battery cells in the corresponding mode set and the total number of battery cells.

[0129] The battery degradation path determination device based on the large language model provided by the embodiment of the application can execute the battery degradation path determination method based on the large language model provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of the execution method.

[0130] Embodiment four

[0131] Figure 4 is a structural schematic diagram of an electronic device provided by Embodiment Four of the application. The electronic device 10 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are merely examples and are not intended to limit the implementations of the application described and / or claimed herein.

[0132] As Figure 4As shown, the electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., communicatively connected to the at least one processor 11, where the memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0133] Various components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc., an output unit 17, such as various types of displays, a speaker, etc., a storage unit 18, such as a magnetic disk, an optical disk, etc., and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0134] The processor 11 can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the battery degradation path determination method based on a large language model.

[0135] In some embodiments, the battery degradation path determination method based on a large language model can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the battery degradation path determination method based on a large language model described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the battery degradation path determination method based on a large language model by any other appropriate means, e.g., by means of firmware.

[0136] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0137] Computer programs used to implement the battery degradation path determination method based on large language model of the present application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the computer program, when executed by the processor, implements the functions / operations specified in the flow diagrams and / or the block diagrams. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a standalone software package and partially on a remote machine, or entirely on a remote machine or server.

[0138] Embodiment five

[0139] Embodiment five of the present application also provides a computer readable storage medium, which stores computer instructions for causing a processor to execute a battery degradation path determination method based on a large language model, the method comprising:

[0140] For at least one battery cell, obtain battery-related parameters of the battery cell in the current cycle, wherein the battery-related parameters at least include voltage sequence, current sequence, temperature sequence, environmental sequence parameters of the running environment where the battery cell is located, maximum available capacity of the current cycle, internal resistance of the battery cell, and charge-discharge rate of the battery cell;

[0141] According to the likelihood function corresponding to each degradation mode and the prior probability, the feature vector corresponding to the battery-related parameters is processed to obtain the posterior probability corresponding to each degradation mode;

[0142] According to the posterior probability corresponding to at least one degradation mode, a target degradation mode is determined, and the target degradation mode, a target degradation path corresponding to the target degradation mode, a feature contribution rate corresponding to each index in the feature vector, and the feature vector are processed based on a target language model to obtain an explainability content corresponding to each degradation information in the target degradation path and a corresponding confidence degree;

[0143] Based on the posterior probability and the confidence degree corresponding to the target degradation mode, it is determined whether the explainability content is displayed.

[0144] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disc read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0145] To provide for interaction with the application, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the application, and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the application can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with the application; for example, feedback provided to the application can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the application can be received in any form, including acoustic, speech, or tactile input.

[0146] The systems and techniques described herein can be implemented in a computing system that includes a back end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front end component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described herein, or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0147] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0148] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the scope of the present disclosure. For example, the steps recited in the present disclosure can be executed in parallel, executed in series, or executed in different orders, as long as the desired results of the technical solutions of the present disclosure can be achieved, and the present disclosure is not limited herein.

[0149] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for determining battery degradation paths based on a large language model, characterized in that, Applied to a rechargeable lithium battery, the rechargeable lithium battery including at least one cell, the method includes: For the at least one battery cell, obtain the battery-related parameters of the battery cell in the current cycle, wherein the battery-related parameters include at least the voltage sequence, current sequence, temperature sequence, environmental sequence parameters of the operating environment of the battery cell, the maximum available capacity of the current cycle, the internal resistance of the battery cell, and the charge / discharge rate of the battery cell; Based on the likelihood function and prior probability corresponding to each degradation mode, the feature vectors corresponding to the battery-related parameters are processed to obtain the posterior probability corresponding to each degradation mode. Based on the posterior probability corresponding to at least one degradation mode, the target degradation mode is determined, and the target degradation mode, the target degradation path corresponding to the target degradation mode, the feature contribution rate corresponding to each index in the feature vector, and the feature vector are processed based on the target language model to obtain the interpretability content and corresponding confidence level corresponding to each degradation information in the target degradation path. Based on the posterior probability corresponding to the target degradation mode and the confidence level, it is determined whether the interpretable content should be displayed.

2. The method according to claim 1, characterized in that, After obtaining the battery-related parameters, the method further includes: By processing the battery-related parameters, the feature vector of the battery cell is obtained; The process of processing the battery-related parameters to obtain the feature vector of the battery cell includes: Obtain the maximum usable capacity from the battery-related parameters, and determine the capacity decay rate of the battery cell based on the maximum usable capacity and the maximum usable capacity of the battery cell in the initial state; Based on the internal resistance in the battery-related parameters and the initial internal resistance of the cell in the initial state, determine the internal resistance increment of the cell; Based on the voltage and current sequences in the battery-related parameters, determine the differential voltage peak shift parameter and the differential current peak shift parameter; Based on the voltage sequence and cell capacity change information in the battery-related parameters, determine the first change curve between voltage and capacity; The characteristic vector of the battery cell is determined by structuring the capacitance decay rate, the internal resistance increment, the differential voltage peak shift parameter, the differential current peak shift parameter, the first variation curve, the charge / discharge rate, the environmental sequence parameters, and the temperature sequence.

3. The method according to claim 1, characterized in that, The step of processing the feature vectors corresponding to the battery-related parameters based on the likelihood function and prior probability corresponding to each degradation mode to obtain the posterior probability corresponding to each degradation mode includes: For each degradation mode, the likelihood function corresponding to the degradation mode is retrieved, and the feature vector is substituted into the likelihood function to obtain the likelihood probability of the battery cell in the current cycle. The likelihood function includes the feature mean and feature covariance corresponding to each indicator associated with the degradation mode. The feature mean and feature covariance are determined based on the feature values ​​of the same indicator in the current cycle and the historical cycle before the current cycle. Based on the prior probability and the likelihood probability corresponding to the degradation mode, determine the posterior probability corresponding to the degradation mode; The prior probability is determined based on the number of samples corresponding to the degradation mode and all total sample data.

4. The method according to claim 1, characterized in that, Determining the target degradation mode based on the posterior probability corresponding to the at least one degradation mode includes: The degradation pattern with the highest posterior probability is taken as the target degradation pattern; The target degradation mode includes a preset degradation path corresponding to the battery cell, and the preset degradation path includes at least one degradation information.

5. The method according to claim 1, characterized in that, The feature contribution rate corresponding to each indicator in the feature vector is determined in the following way: By calculating the posterior probability of the target degradation mode and the partial derivative corresponding to each feature vector, the feature influence intensity information of the index to which each feature vector belongs is determined; Based on the characteristic influence intensity information of each indicator and the characteristic influence intensity information of all indicators, the characteristic contribution rate corresponding to each indicator in the characteristic vector is determined.

6. The method according to claim 1, characterized in that, The step of determining whether to display the interpretable content based on the posterior probability corresponding to the target degradation mode and the confidence level includes: Obtain the confidence level corresponding to the interpretability of each piece of degraded information; After processing the mean of all confidence scores, they are fused with the posterior probability to obtain the displayable parameters of all the interpretable content. When the displayable parameters meet the preset conditions, target content is generated and displayed based on each degradation information and the corresponding interpretable content.

7. The method according to claim 1, characterized in that, The method further includes: The battery-related parameters of the battery cell are input into a pre-trained classification model, which outputs the predicted mode category and the probability information corresponding to each predicted mode category. Based on the probability information of the battery cell, determine the set of patterns corresponding to the battery cell; After obtaining the target degradation mode corresponding to each battery cell, it is determined whether the target degradation mode is in the corresponding mode set; Based on the number of cells in the corresponding mode set and the total number of cells, determine the accurate attributes of the target degradation mode for each cell.

8. A battery degradation path determination device based on a large language model, characterized in that, include: A battery-related parameter acquisition module is used to acquire, for at least one battery cell, battery-related parameters of the battery cell in the current cycle, wherein the battery-related parameters include at least voltage sequence, current sequence, temperature sequence, environmental sequence parameters of the operating environment of the battery cell, maximum available capacity of the current cycle, internal resistance of the battery cell, and charge / discharge rate of the battery cell; The posterior probability determination module is used to process the feature vectors corresponding to the battery-related parameters based on the likelihood function and prior probability corresponding to each degradation mode, so as to obtain the posterior probability corresponding to each degradation mode. The interpretability content and confidence determination module is used to determine the target degradation mode based on the posterior probability corresponding to at least one degradation mode, and to process the target degradation mode, the target degradation path corresponding to the target degradation mode, the feature contribution rate corresponding to each index in the feature vector, and the feature vector based on the target language model to obtain the interpretability content and corresponding confidence corresponding to each degradation information in the target degradation path. The interpretable content display module is used to determine whether to display the interpretable content based on the posterior probability corresponding to the target degradation mode and the confidence level.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the battery degradation path determination method based on a large language model as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the battery degradation path determination method based on a large language model as described in any one of claims 1-7.

Images