A power battery lithium precipitation online detection method and device based on cloud big data

Abstract

The application provides a power battery lithium precipitation online detection method and device based on cloud big data, which comprises the following steps: collecting multi-dimensional original data in the charging process of the power battery and preprocessing, extracting capacity increment IC curve, electric parameter and temperature characteristics to construct a core feature set; constructing a two-step hybrid machine learning model combining support vector machine and random forest to realize lithium precipitation identification and reversible / irreversible lithium precipitation type judgment; transmitting data from the vehicle end BMS to the cloud end in real time during the charging process to complete the detection and feed back the results to the vehicle end to perform graded closed-loop processing; and periodically iterating and optimizing the model based on the accumulated full data in the cloud. The application can realize in-situ real-time accurate detection and active intervention of lithium precipitation without additional detection hardware, effectively improves the safety and cycle life of the power battery, and has strong model generalization.

Description

Technical Field

[0001] This application relates to the field of battery testing technology, and in particular to an online detection method and device for lithium plating in power batteries based on cloud-based big data. Background Technology

[0002] As a core energy storage component of new energy vehicles, lithium-ion power batteries face a critical safety hazard during the charging phase: lithium deposition at the negative electrode. The metallic lithium deposited at the negative electrode readily forms lithium dendrites, which can lead to loss of active lithium and capacity decay, or even puncture the separator, causing internal short circuits and potentially resulting in battery fires or explosions. Currently, lithium deposition detection for power batteries employs two main technical approaches: offline and online. Offline detection relies on battery disassembly, X-ray diffraction, and three-electrode testing, requiring system shutdown and sample removal. This not only fails to meet the real-time monitoring needs of vehicles during charging and discharging but also causes permanent damage to the tested cells. Conventional online detection largely relies on single physical parameters such as voltage and gas production to indirectly deduce the lithium deposition state. This is susceptible to fluctuations in ambient temperature, charging rate, and battery aging, resulting in a high rate of missed and false detections, and making it difficult to differentiate between reversible and irreversible lithium deposition.

[0003] The existing lithium plating control system for power batteries lacks a closed-loop architecture that links detection and charging control. Traditional diagnostic models mostly rely on small sample data from individual cells or single vehicle models for modeling, lacking the support of massive amounts of real-world vehicle operating condition data. This results in weak model generalization performance and a significant decline in detection accuracy across battery types and operating scenarios. Conventional detection processes only determine lithium plating faults and cannot dynamically adjust the vehicle's BMS charging parameters based on the type of lithium plating. They cannot achieve lithium re-intercalation repair through operating condition optimization for early reversible lithium plating, nor can they simultaneously push maintenance reminders for irreversible lithium plating. They can only passively detect faults and cannot curb the continued deterioration of lithium plating from the source of charging control, failing to meet the actual usage needs of power battery safety protection and life extension management. Summary of the Invention

[0004] In view of this, the embodiments of this application provide a method and device for online detection of lithium plating in power batteries based on cloud big data, which can realize in-situ real-time accurate detection and active intervention of lithium plating, effectively improving the safety and cycle life of power batteries.

[0005] The technical solution of this application embodiment is implemented as follows: In a first aspect, embodiments of this application provide an online detection method for lithium plating in power batteries based on cloud-based big data, comprising the following steps: Collect multi-dimensional raw data during the charging process of new energy vehicle power batteries, and perform outlier removal, missing value imputation and data standardization in sequence to obtain standardized data; Extract a core feature set related to lithium plating from the standardized data. The core feature set includes capacity increment IC curve features, electrical parameter features, and temperature features. A hybrid machine learning model is constructed based on a lithium plating identification sub-model using Support Vector Machine (SVM) and a lithium plating type judgment sub-model using Random Forest. The hybrid model is then trained and optimized based on the core feature set. During the charging process of new energy vehicles, the BMS system transmits the multi-dimensional data collected in real time to the cloud. After processing, the data is input into the trained hybrid machine learning model. The lithium plating identification sub-model determines whether lithium plating has occurred. If lithium plating is determined to have occurred, the lithium plating type identification sub-model distinguishes between reversible and irreversible lithium plating. The judgment result is fed back to the vehicle-side BMS system, and the corresponding processing strategy is executed according to different judgment results. If there is weak lithium plating, normal charging is maintained. If reversible lithium plating occurs, the charging strategy is adjusted to promote lithium dendrite re-intercalation. If irreversible lithium plating occurs, charging is stopped and an early warning is issued.

[0006] Secondly, embodiments of this application also provide an online detection device for lithium plating in power batteries based on cloud-based big data, the device comprising: The data acquisition module is used to collect multi-dimensional raw data during the charging process of new energy vehicle power batteries, and then perform outlier removal, missing value imputation and data standardization to obtain standardized data. The extraction module is used to extract a core feature set related to lithium plating from the standardized data. The core feature set includes capacity increment IC curve features, electrical parameter features, and temperature features. The module is used to build a hybrid machine learning model based on a lithium plating identification sub-model using support vector machine (SVM) and a lithium plating type judgment sub-model using random forest, and to train and optimize the hybrid model based on the core feature set. The judgment module is used during the charging process of new energy vehicles. It transmits multi-dimensional data collected in real time to the cloud through the BMS system. After processing, the data is input into the trained hybrid machine learning model. The lithium plating identification sub-model determines whether lithium plating has occurred. If lithium plating is determined to have occurred, the lithium plating type judgment sub-model distinguishes between reversible and irreversible lithium plating. The feedback module is used to send the judgment results back to the vehicle-side BMS system. Based on different judgment results, the corresponding processing strategy is executed. If there is weak lithium plating, normal charging is maintained. If reversible lithium plating occurs, the charging strategy is adjusted to promote lithium dendrite re-intercalation. If irreversible lithium plating occurs, charging is stopped and an early warning is issued.

[0007] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the online detection method for lithium plating of power batteries based on cloud big data as described in any of the first aspects.

[0008] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the online detection method for lithium plating of power batteries based on cloud big data as described in any one of the first aspects.

[0009] The embodiments of this application have the following beneficial effects: Through multi-source data preprocessing, multi-dimensional feature extraction, combined SVM and random forest modeling, real-time online analysis, hierarchical closed-loop control, and periodic model iteration, lithium plating can be identified in situ and accurately classified into reversible and irreversible lithium plating categories during vehicle charging. Relying on native BMS data eliminates the need for external detection hardware, reducing detection costs. Closed-loop control can optimize charging conditions for reversible lithium plating to achieve lithium re-intercalation repair and delay battery degradation, while providing timely warnings and mitigation for irreversible lithium plating to avoid the risk of thermal runaway caused by lithium dendrites puncturing the separator. With quarterly iterative optimization to continuously expand the sample and update the model, the model's generalization detection capability across battery models and operating conditions is effectively improved. This ensures the safety of vehicle batteries while extending the cycle life of power batteries. The accumulated full monitoring data can also provide a data foundation for refined battery health management. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating steps S101-S105 provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the cloud-based big data-based online detection device for lithium plating in power batteries provided in this application embodiment; Figure 3 This is a schematic diagram of the composition structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0013] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0014] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0015] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0016] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application and is not intended to limit this application.

[0018] See Figure 1 , Figure 1This is a flowchart illustrating steps S101-S105 of the online lithium plating detection method for power batteries based on cloud-based big data provided in this application embodiment. Figure 1 Steps S101-S105 are explained below.

[0019] In step S101, multi-dimensional raw data during the charging process of the power battery of the new energy vehicle are collected, and outlier removal, missing value filling and data standardization are performed in sequence to obtain standardized data; In step S102, a core feature set related to lithium plating is extracted from the standardized data. The core feature set includes capacity increment IC curve features, electrical parameter features, and temperature features. In step S103, a hybrid machine learning model is constructed based on the lithium plating identification sub-model of support vector machine (SVM) and the lithium plating type judgment sub-model of support random forest, and the hybrid model is trained and optimized based on the core feature set. In step S104, during the charging process of the new energy vehicle, the multi-dimensional data collected in real time is transmitted to the cloud through the BMS system. After processing, the data is input into the trained hybrid machine learning model. The lithium plating identification sub-model determines whether lithium plating has occurred. If lithium plating is determined to have occurred, the lithium plating type identification sub-model distinguishes between reversible and irreversible lithium plating. In step S105, the judgment result is fed back to the vehicle-side BMS system. The corresponding processing strategy is executed according to different judgment results. If there is weak lithium plating, normal charging is maintained. If reversible lithium plating occurs, the charging strategy is adjusted to promote lithium dendrite re-intercalation. If irreversible lithium plating occurs, charging is stopped and an early warning is issued.

[0020] Data preprocessing involves systematically cleaning and standardizing the raw data collected from the vehicle, eliminating noise, outliers, missing values, and dimensional differences. This provides high-quality, highly consistent data input for subsequent feature extraction and model training. The quality of the raw data directly determines the upper limit of the machine learning model. Unprocessed raw data contains a large amount of interfering information, which can cause the model to learn incorrect patterns and severely reduce detection accuracy. Multi-dimensional feature extraction is then performed, transforming high-dimensional, continuous time-series data into discrete feature vectors with clear physical meaning that can be effectively recognized and processed by the machine learning model.

[0021] This embodiment constructs a two-step hybrid model architecture by fusing two different types of machine learning algorithms, Support Vector Machine (SVM) and Random Forest, to separately identify the presence or absence of lithium plating and distinguish the type of lithium plating. This hybrid architecture fully leverages the advantages of both algorithms, avoiding the limitations of a single algorithm in handling complex classification problems, while significantly improving the model's generalization ability and robustness. The real-time detection and judgment step enables online and real-time lithium plating detection. Through a low-latency communication network between the vehicle and the cloud, the charging data collected in real time by the vehicle's BMS is transmitted to the cloud for processing and inference, and the judgment results are fed back to the vehicle in real time, ensuring that anomalies can be detected in the early stages of lithium plating, thus gaining valuable time for subsequent closed-loop processing.

[0022] Differentiated treatment strategies are adopted for different types of lithium plating. For reversible lithium plating, the charging parameters are adjusted to promote the re-intercalation of lithium dendrites. For irreversible lithium plating, charging is stopped in time and an early warning is issued. This fundamentally solves the problems of shortened battery life and safety hazards caused by lithium plating.

[0023] In some embodiments, the acquisition frequency of the multi-dimensional raw data is 10Hz, including: Electrical parameters: charging current I, charging voltage U, battery terminal voltage Ucell, SOC, coulombic efficiency η; Environmental and temperature parameters: cell temperature (Tcell), ambient temperature (Tenvironment); Basic battery parameters: battery model, production batch, number of cycles N, aging degree S, 0≤S≤1, S=1 is a brand new battery, S=0 is a scrapped battery; Charging parameters: charging rate C, charging mode.

[0024] The raw data collected in this embodiment is divided into four categories and 12 specific dimensions, comprehensively covering all key factors affecting the occurrence and development of lithium plating. The first category is electrical parameters, including charging current I(t), charging voltage U(t), battery terminal voltage Ucell(t), SOC(t), and coulombic efficiency η(t). Charging current and voltage are the most basic electrical parameters, directly reflecting the external electrical characteristics of the battery during charging; battery terminal voltage refers to the voltage of a single cell, which more accurately reflects the internal state of a single cell than the total voltage of the battery pack; SOC (State of Charge) represents the current remaining charge of the battery, and lithium plating usually occurs in the charging stage when SOC is high; coulombic efficiency is the ratio of the amount of charge released by the battery during discharge to the amount of charge input during charging, which can directly reflect the degree of side reactions inside the battery.

[0025] The second category consists of environmental and temperature parameters, including cell temperature (Tcell(t)) and ambient temperature (Tenvironment(t)). Temperature is one of the most important factors affecting the performance and lithium plating behavior of lithium-ion batteries. The lithium-ion insertion rate at the negative electrode, the electrolyte conductivity, and the rate of side reactions are all closely related to temperature. Cell temperature directly reflects the internal heat generation of the battery. During lithium plating, the increased heat generation from side reactions and Joule heating can lead to an abnormal rise in cell temperature. Ambient temperature is the external condition for battery operation. Low temperatures significantly reduce the diffusion rate of lithium ions, which is one of the main causes of lithium plating.

[0026] The third category consists of basic battery parameters, including battery model, production batch, cycle count (N), and aging level (S). Different battery models exhibit significant differences in lithium plating characteristics due to variations in positive and negative electrode materials, electrolyte formulations, and structural designs. Even batteries of the same model from different production batches may show performance differences due to fluctuations in the manufacturing process. Cycle count and aging level directly reflect the battery's health status. As the number of cycles increases, the battery gradually ages, the number of lithium insertion sites on the negative electrode decreases, internal resistance increases, and the risk of lithium plating significantly increases. Aging level (S) is a dimensionless parameter between 0 and 1, where S=1 represents a brand-new battery and S=0 represents a completely unusable battery. It integrates multiple factors such as cycle count, capacity decay, and increased internal resistance, providing a more comprehensive reflection of the battery's aging state.

[0027] The fourth category is charging parameters, including the charging rate C(t) and charging mode. The charging rate refers to the ratio of the battery charging current to its rated capacity. High charging rates cause the deposition rate of lithium ions on the negative electrode surface to exceed the insertion rate, thus triggering lithium plating. Charging modes mainly include constant current charging, constant voltage charging, and various combinations thereof. Different charging modes have different effects on lithium plating; the risk of lithium plating is generally higher during constant current charging than during constant voltage charging. By simultaneously collecting these four categories of data, this embodiment can comprehensively and three-dimensionally understand the battery's charging and health status from multiple perspectives, providing sufficient and comprehensive information support for lithium plating detection.

[0028] In some embodiments, outlier removal adopts the 3σ criterion. For any dimension of data X, if X exceeds the range of [μ-3σ,μ+3σ], it is determined to be an outlier and removed. μ is the data mean and σ is the standard deviation. Missing value imputation adopts linear interpolation. Data standardization adopts the Min-Max standardization method to map the data to the interval [0,1].

[0029] Outlier removal employs the 3σ criterion (Raida criterion), which assumes the data follows a normal distribution. In this embodiment, the mean μ and standard deviation σ are calculated for each dimension of time-series data. For each data point, it is determined whether the data point falls within the range of [μ-3σ, μ+3σ] to avoid misjudgments caused by variations in the overall data distribution.

[0030] Missing value imputation uses linear interpolation. During vehicle-side data acquisition, short-term data loss often occurs due to unstable network transmission, typically lasting from a few hundred milliseconds to a few seconds. Linear interpolation uses two adjacent valid data points before and after the missing data point to calculate an estimated value for the missing data point through linear fitting. Its basic assumption is that the battery's state changes linearly over a short period. This assumption holds true for the charging process of a power battery, as parameters such as voltage, current, and temperature change slowly and continuously during normal charging. For continuous data loss lasting more than several seconds, this embodiment marks this data segment as invalid and excludes it from subsequent feature extraction and model inference, avoiding misjudgments due to excessive interpolation errors.

[0031] Data standardization employs the Min-Max standardization method, also known as deviation standardization, which linearly maps the original data to the [0,1] interval. Its calculation formula is X_norm = (X - X_min) / (X_max - X_min), where X is the original data, and X_min and X_max are the minimum and maximum values ​​of that dimension of data throughout the entire charging process, respectively. The purpose of data standardization is to eliminate differences in dimensions and numerical ranges between different data dimensions. For example, the numerical range of charging voltage is typically between 3V and 4.2V, while the numerical range of charging current may be between 0A and 100A. Without standardization, features with larger numerical ranges will dominate during model training, causing the model to ignore features with smaller numerical ranges that are equally important. The Min-Max standardization method is computationally simple and highly efficient, making it very suitable for the rapid processing of massive real-time data on cloud platforms. Simultaneously, it preserves the linear relationship of the original data and does not alter the data distribution characteristics.

[0032] In some embodiments, the capacity increment IC curve features include five maximum peaks P1-P5, two minimum peaks P6-P7, and the battery terminal voltage value corresponding to the minimum peak P7. IC curve analysis is a non-destructive testing technique based on the relationship between battery voltage and capacity, which can intuitively reflect the process of lithium-ion intercalation into the negative electrode. The horizontal axis of the IC curve represents the battery terminal voltage Ucell, and the vertical axis represents the capacity increment IC. The formula for calculating the IC value is IC = ΔQ / ΔU, where ΔQ is the change in charging capacity per unit time, and ΔU is the corresponding change in battery terminal voltage. During normal battery charging, lithium ions can smoothly intercalate into the layered structure of the graphite negative electrode, and the voltage rises steadily with increasing capacity. Therefore, the IC curve exhibits distinct peaks and valleys and a regular shape. Each peak corresponds to a specific lithium intercalation stage in the graphite negative electrode, and the position and height of each peak are relatively stable.

[0033] When lithium plating occurs during charging, metallic lithium is deposited on the surface of the negative electrode. This deposited lithium metal occupies the lithium intercalation sites on the negative electrode surface, increasing the resistance to lithium ion intercalation. To achieve the same capacity increment ΔQ, the battery terminal voltage needs to rise at a faster rate, resulting in a significant increase in voltage increment ΔU. According to the IC value calculation formula, the increase in ΔU will directly lead to a decrease in the IC value, thus causing significant distortion of the IC curve. Specifically, this manifests as: a significant decrease in the height of the main peak of the IC curve, with the peak position shifting towards higher voltage; a significant increase and shallowing of the valley values, with a reduction in the difference between peaks and valleys; and the overall curve becoming flatter, losing its original regular shape. This embodiment specifically extracts the five maximum peaks (P1-P5), two minimum peaks (P6-P7), and the battery terminal voltage value corresponding to the minimum peak P7 from the IC curve. These features can accurately reflect the degree of distortion of the IC curve and are used to determine whether lithium plating has occurred.

[0034] Electrical parameters include the charging current fluctuation coefficient KI, voltage fluctuation coefficient KU, and coulombic efficiency η. The charging current fluctuation coefficient and voltage fluctuation coefficient are the ratios of the standard deviation to the average value of the charging current and voltage, respectively, reflecting the stability of the current and voltage during charging. During normal battery charging, the lithium-ion intercalation process is smooth and stable, and the polarization changes within the battery are gradual, resulting in very stable charging current and voltage with minimal fluctuation coefficients. When lithium plating occurs, the lithium metal deposited on the negative electrode surface leads to uneven distribution of lithium intercalation resistance, making the lithium-ion intercalation process unstable and causing abnormal fluctuations in charging current and voltage, with a significantly increased fluctuation coefficient. Coulombic efficiency refers to the ratio of the amount of charge released during battery discharge to the amount of charge input during charging; it directly reflects the degree of side reactions within the battery. During lithium plating, the deposited metallic lithium reacts with the electrolyte to generate irreversible products, leading to a decrease in coulombic efficiency. In reversible lithium plating, the side reactions are mild, the decrease in coulombic efficiency is small, and it can recover; in irreversible lithium plating, the side reactions are severe and continuous, resulting in a significant and irrecoverable decrease in coulombic efficiency.

[0035] The temperature characteristic is the difference between the cell temperature and the ambient temperature, ΔT = Tcell - Temperature. This characteristic, by subtracting the ambient temperature, eliminates the influence of ambient temperature changes on the battery temperature, accurately reflecting the temperature rise inside the battery due to electrochemical reactions and Joule heating. During normal battery charging, the lithium-ion insertion / extraction process is smooth and gentle, generating only uniform and trace amounts of reaction heat and Joule heat. Therefore, the cell temperature rises slowly and uniformly, and ΔT remains within a small, stable range. When lithium plating occurs, the deposited metallic lithium undergoes an exothermic side reaction with the electrolyte. Simultaneously, the increase in battery internal resistance also leads to increased Joule heat, significantly accelerating the battery's heat generation rate and causing an abnormally high ΔT. In reversible lithium plating, the increase in heat generation is small, and the heat generation rate gradually returns to normal as lithium plating recovers. In irreversible lithium plating, the increase in heat generation is large and lasts for a long time, with the battery temperature recovering slowly after charging.

[0036] In some embodiments, the lithium plating identification sub-model is a support vector machine (SVM) model, used to achieve binary classification judgment between lithium plating-free and lithium plating-present conditions; The lithium plating type judgment sub-model is a random forest model, which is used to classify samples judged to have lithium plating as either reversible or irreversible lithium plating. During model training, the dataset is divided into training set, validation set and test set in a ratio of 7:2:1. The sample labels are divided into three categories: 0 - no lithium plating, 1 - reversible lithium plating, and 2 - irreversible lithium plating.

[0037] The lithium plating identification sub-model employs a Support Vector Machine (SVM) model. The SVM model exhibits superior performance when processing high-dimensional, small-sample data. It possesses non-linear fitting capabilities, effectively handling complex relationships between features, and demonstrates good generalization ability, effectively avoiding overfitting. In this embodiment, the lithium plating identification sub-model's task is to classify samples into two categories: "no lithium plating" and "lithium plating present." In actual operation, the vast majority of charging processes are normal, with no lithium plating samples accounting for over 90%. The SVM model can quickly and accurately identify these normal samples without lithium plating, eliminating the need for subsequent lithium plating type determination steps. This significantly improves the overall system's detection efficiency and reduces the computational load on the cloud platform.

[0038] The lithium plating type determination sub-model employs a random forest model. Random forest models offer advantages such as fast training speed, strong resistance to overfitting, ability to handle multi-class classification problems, and automatic assessment of feature importance, making them highly suitable for the complex classification task of distinguishing between reversible and irreversible lithium plating. The differences in features between reversible and irreversible lithium plating are subtle and multi-dimensional, and a single decision tree struggles to accurately capture these differences. Random forests, by integrating the predictions of multiple decision trees, significantly improve classification accuracy and robustness.

[0039] During model training, the massive amount of preprocessed feature data accumulated on the cloud platform is first divided into training, validation, and test sets in a 7:2:1 ratio. This ensures sufficient model training while effectively evaluating the model's generalization ability. The training set is used for parameter training, the validation set is used for adjusting hyperparameters, and the test set is used for final performance evaluation. The sample labels in the dataset are divided into three categories: 0 indicates no lithium plating, 1 indicates reversible lithium plating, and 2 indicates irreversible lithium plating.

[0040] In some embodiments, the random forest model generates multiple decision trees through bootstrap sampling, with each decision tree having a maximum depth of 8. The Gini coefficient is used to measure feature importance and select core features. The penalty factor C of the SVM model is 10. The model is optimized using 5-fold cross-validation.

[0041] In this embodiment, the random forest model generates multiple decision trees through bootstrap sampling. Bootstrap draws samples with replacement from the original training set, with the same size as the original training set, as the training set for each decision tree. This sampling method ensures that the training sets of each decision tree have some differences, thereby improving model diversity and reducing the risk of overfitting. Generally, the more decision trees there are, the higher the model's accuracy, but the computational complexity and memory consumption also increase significantly. Experiments in this embodiment show that when the number of decision trees increases from 100 to 150, the model's classification accuracy improves significantly; when the number of decision trees exceeds 150, the improvement in model accuracy is very limited, almost plateauing, but the inference time increases linearly. Therefore, this embodiment selects 150 decision trees.

[0042] The maximum depth of each decision tree is set to 8. Decision trees that are too deep can learn very subtle features in the training set, including noise, leading to overfitting and a decrease in the model's generalization ability on unknown data. Decision trees that are too shallow, on the other hand, cannot fully learn the complex patterns in the data, resulting in underfitting and insufficient model accuracy. This embodiment uses a 5-fold cross-validation experiment to test the model's performance when the maximum depth ranges from 5 to 15. The results show that when the maximum depth is 8, the model achieves the highest accuracy on the validation set and the lowest risk of overfitting. Furthermore, during the training process of the random forest model, when each node of each decision tree splits, it randomly selects the square root of the total number of features as candidate features, which further improves the model's diversity and resistance to overfitting.

[0043] Random forest models use the Gini coefficient to measure feature importance. The Gini coefficient is an indicator of node impurity; a smaller Gini coefficient indicates that the node contains a more homogeneous set of sample classes, resulting in higher purity. By calculating the change in the Gini coefficient before and after a feature splits a node, the contribution of that feature to the classification result can be quantified; a larger change indicates a more important feature. This embodiment utilizes this characteristic to automatically select the most critical features for determining lithium plating type, including coulombic efficiency, IC curve peak value P1, and temperature difference ΔT. By removing unimportant redundant features, not only can the computational load of the model be reduced and the inference speed improved, but the interference of redundant features on the model can also be avoided, further enhancing the model's classification accuracy.

[0044] The penalty factor C of the SVM model is set to 10. The penalty factor C is the most important hyperparameter of the SVM model, used to control the intensity of the penalty for misclassified samples. A larger C value results in a heavier penalty for misclassified samples, attempting to correctly classify all training samples, which can easily lead to overfitting. A smaller C value results in a higher tolerance for misclassified samples, seeking a wider classification margin, which can easily lead to underfitting. This embodiment tested multiple values ​​of C between 0.1 and 100 using grid search and 5-fold cross-validation. The results show that when C=10, the SVM model exhibits the best generalization performance on the lithium plating identification task, maintaining a high detection rate while keeping the false detection rate low. The SVM model uses a radial basis function (RBF) as its kernel function. This kernel function can transform a nonlinearly separable problem in a low-dimensional feature space into a linearly separable problem in a high-dimensional feature space, exhibiting good nonlinear fitting ability.

[0045] The model is optimized using 5-fold cross-validation. 5-fold cross-validation divides the training set into five non-overlapping subsets. Each time, four subsets are used as the training set, and the remaining subset is used as the validation set. This process is repeated five times, and the average of the five validation results is taken as the model's final performance. 5-fold cross-validation makes full use of limited training data, more accurately evaluates the model's generalization ability, and effectively avoids overfitting. Through 5-fold cross-validation, this embodiment can find the optimal combination of model hyperparameters, ensuring that the trained model maintains high detection accuracy even on unknown real-world data.

[0046] In some embodiments, the adjustment of the charging strategy for reversible lithium plating includes: Adjust the current charging rate C to 50%-70% of the original rate; The temperature of individual battery cells should be controlled within the range of 25℃-35℃; Switch to constant voltage charging mode after SOC ≥ 80%; After adjustment, lithium plating is re-evaluated every first preset time. If no lithium plating is detected multiple times in a row, the normal charging strategy is restored. If reversible lithium plating is still detected after the second preset time, a maintenance reminder is triggered.

[0047] The first adjustment measure is to reduce the charging rate, adjusting the current charging rate C(t) to 50%-70% of the original rate. High charging rates are one of the main causes of lithium plating. When the charging rate is too high, the rate at which lithium ions escape from the positive electrode exceeds their rate of insertion into the negative electrode, leading to a large accumulation and deposition of lithium ions on the negative electrode surface. Reducing the charging rate directly slows down the migration and deposition rate of lithium ions, reducing the deposition of new lithium metal, while giving the already deposited lithium metal sufficient time to redissolve and re-insert into the negative electrode. Experiments have shown that reducing the charging rate to 50%-70% of the original rate is the optimal adjustment range. If the reduction is too small, it will not effectively suppress the continued development of lithium plating; if the reduction is too large, it will significantly prolong the charging time, affecting the user's charging experience.

[0048] The second adjustment measure is to control the temperature of the individual battery cells within the range of 25℃-35℃. Temperature has a decisive influence on the diffusion rate of lithium ions in the graphite anode. Within a certain range, increased temperature significantly accelerates the diffusion rate of lithium ions, thereby improving the re-intercalation efficiency of lithium dendrites. 25℃-35℃ is the optimal operating temperature range for lithium-ion batteries. Within this range, the diffusion rate of lithium ions is the fastest, the conductivity of the electrolyte is the highest, and the rate of side reactions is relatively low. If the battery temperature is below 25℃, the diffusion rate of lithium ions will drop sharply, and the re-intercalation efficiency of lithium dendrites will be very low. If the battery temperature is above 35℃, although the diffusion rate of lithium ions will further increase, the rate of side reactions such as electrolyte decomposition will also increase significantly, accelerating battery aging. Therefore, this method uses the vehicle-side thermal management system to precisely control the battery temperature within the range of 25℃-35℃. If the temperature is below 25℃, the battery heating system is activated; if the temperature is above 35℃, the battery cooling system is activated.

[0049] The third adjustment measure is to switch to constant voltage charging mode after the SOC reaches ≥80%. The charging process of a power battery typically consists of two stages: constant current charging and constant voltage charging. In the constant current charging stage, the charging current remains constant while the battery voltage gradually increases; in the constant voltage charging stage, the battery voltage remains constant while the charging current gradually decreases. Lithium plating usually occurs in the later stages of constant current charging, i.e., when the SOC is high. When the SOC ≥80%, most of the lithium intercalation sites on the negative electrode are already occupied, and continuing to use high-current constant current charging will significantly increase the risk of lithium plating. After switching to constant voltage charging mode, the charging current gradually decreases, effectively avoiding the aggravated lithium plating caused by high current. Simultaneously, the slower charging current promotes the uniform insertion of lithium ions into the remaining lithium intercalation sites on the negative electrode and also facilitates the re-intercalation of already deposited lithium metal.

[0050] After adjusting the charging strategy, the system continuously monitors the battery status, collecting data every 5 minutes and inputting it into the machine learning model for re-evaluation of lithium plating. The 5-minute detection interval allows sufficient time for the charging strategy adjustment to take effect, as the re-intercalation of lithium dendrites is an electrochemical process that requires time. Excessive detection is not only unnecessary but also increases computational burden. If two consecutive evaluations show no lithium plating, it indicates that the deposited lithium metal has been largely re-intercalated, and the battery has returned to normal. The system will then automatically revert to the original charging strategy and continue charging. If reversible lithium plating is still detected after 30 minutes, it indicates severe lithium plating or other abnormalities in the battery that cannot be effectively resolved simply by adjusting the charging strategy. In this case, the system will determine it as suspected irreversible lithium plating, triggering a maintenance reminder. Users are advised to have the battery thoroughly inspected and maintained by a professional repair shop as soon as possible.

[0051] In summary, the embodiments of this application have the following beneficial effects: By integrating multi-dimensional charging data through a cloud-based big data platform and fusing a high-sensitivity feature set of capacity increment IC curves, electrical parameters, and temperature differences, a two-step hybrid machine learning model combining SVM and random forest is employed. This enables in-situ real-time detection of lithium plating during charging and accurate differentiation between reversible and irreversible lithium plating, effectively solving the problems of offline damage, poor anti-interference performance of single parameters, and inability to distinguish lithium types in traditional methods. Furthermore, for reversible lithium plating, the charging strategy is dynamically adjusted to promote lithium dendrite re-intercalation, extending battery cycle life. For irreversible lithium plating, charging is stopped promptly with warnings to prevent short circuits and fires. This solution requires no additional detection hardware, relying solely on data collected from the existing BMS, significantly reducing detection costs. The model trained on massive cloud-based data possesses excellent generalization capabilities, allowing for batch adaptation to different vehicle models, types, and aging levels of power batteries. In addition, the continuously accumulated full lifecycle data in the cloud provides data support for charging strategy iteration, battery health management, and next-generation battery technology development, comprehensively improving the safety, reliability, and economy of new energy vehicle power batteries.

[0052] Based on the same inventive concept, this application also provides a cloud-based big data-based power battery lithium plating online detection device, which corresponds to the cloud-based big data-based power battery lithium plating online detection method in the first embodiment. Since the principle of the device in this application is similar to the above-mentioned cloud-based big data-based power battery lithium plating online detection method, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0053] like Figure 2 As shown, Figure 2This is a schematic diagram of the structure of the cloud-based big data-based online lithium plating detection device 200 for power batteries provided in this application embodiment. The cloud-based big data-based online lithium plating detection device 200 for power batteries includes: The data acquisition module 201 is used to collect multi-dimensional raw data during the charging process of the power battery of new energy vehicles, and sequentially perform outlier removal, missing value filling and data standardization to obtain standardized data. Extraction module 202 is used to extract a core feature set related to lithium plating from the standardized data. The core feature set includes capacity increment IC curve features, electrical parameter features, and temperature features. Module 203 is used to construct a hybrid machine learning model based on a lithium plating identification sub-model using support vector machine (SVM) and a lithium plating type judgment sub-model using random forest, and to train and optimize the hybrid model based on the core feature set. The judgment module 204 is used to transmit multi-dimensional data collected in real time to the cloud through the BMS system during the charging process of new energy vehicles. After processing, the data is input into the trained hybrid machine learning model. The lithium plating identification sub-model determines whether lithium plating has occurred. If lithium plating is determined to have occurred, the lithium plating type judgment sub-model distinguishes between reversible and irreversible lithium plating. The feedback module 205 is used to feed the judgment result back to the vehicle-side BMS system. Based on different judgment results, the corresponding processing strategy is executed. If there is weak lithium plating, normal charging is maintained. If reversible lithium plating occurs, the charging strategy is adjusted to promote lithium dendrite re-intercalation. If irreversible lithium plating occurs, charging is stopped and an early warning is issued.

[0054] Those skilled in the art should understand that Figure 2 The functions of each unit in the cloud-based big data-based power battery lithium plating online detection device 200 shown can be understood by referring to the relevant description of the cloud-based big data-based power battery lithium plating online detection method. Figure 2 The functions of each unit in the cloud-based big data-driven power battery lithium plating online detection device 200 shown can be implemented through a program running on a processor or through specific logic circuits.

[0055] In one possible implementation, module 20* includes: In one possible implementation, module 20* further includes: The aforementioned cloud-based big data-driven online lithium plating detection device for power batteries integrates multi-dimensional charging data through a cloud-based big data platform. It incorporates a high-sensitivity feature set of capacity increment IC curves, electrical parameters, and temperature differences. Employing a two-step hybrid machine learning model combining SVM and random forest, it achieves in-situ real-time detection of lithium plating during charging and accurate differentiation between reversible and irreversible lithium plating. This effectively solves the problems of offline damage, poor anti-interference performance of single parameters, and inability to distinguish lithium types inherent in traditional methods. Furthermore, for reversible lithium plating, it dynamically adjusts the charging strategy to promote lithium dendrite re-intercalation, extending battery cycle life. For irreversible lithium plating, it promptly stops charging and issues warnings to prevent short circuits and fires. This solution requires no additional detection hardware, relying solely on data collected from existing BMS, significantly reducing detection costs. The model trained on massive cloud-based data possesses excellent generalization capabilities, allowing for batch adaptation to power batteries of different models, types, and aging levels. In addition, the continuously accumulated full-lifecycle data in the cloud provides data support for charging strategy iteration, battery health management, and next-generation battery technology development, comprehensively improving the safety, reliability, and economy of new energy vehicle power batteries.

[0056] like Figure 3 As shown, Figure 3 This is a schematic diagram of the composition structure of the electronic device 300 provided in the embodiments of this application. The electronic device 300 includes: The device includes a processor 301, a storage medium 302, and a bus 303. The storage medium 302 stores machine-readable instructions that can be executed by the processor 301. When the electronic device 300 is running, the processor 301 communicates with the storage medium 302 via the bus 303. The processor 301 executes the machine-readable instructions to perform the steps of the online detection method for lithium plating of power batteries based on cloud big data described in this application embodiment.

[0057] In practical applications, the various components in the electronic device 300 are coupled together via bus 303. It is understood that bus 303 is used to achieve communication between these components. In addition to a data bus, bus 303 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 3 The general designated all buses as Bus 303.

[0058] The aforementioned electronic devices integrate multi-dimensional charging data through a cloud-based big data platform, fusing high-sensitivity feature sets of capacity increment IC curves, electrical parameters, and temperature differences. Employing a two-step hybrid machine learning model combining SVM and random forest, they achieve in-situ real-time detection of lithium plating during charging and accurate differentiation between reversible and irreversible lithium plating. This effectively solves the problems of offline damage, poor anti-interference performance of single parameters, and inability to distinguish lithium types inherent in traditional methods. Furthermore, for reversible lithium plating, the charging strategy is dynamically adjusted to promote lithium dendrite re-intercalation, extending battery cycle life. For irreversible lithium plating, charging is stopped promptly with warnings to prevent short circuits and fires. This solution requires no additional detection hardware, relying solely on data collected from the existing BMS, significantly reducing detection costs. The model trained on massive cloud-based data possesses excellent generalization capabilities, allowing for batch adaptation to different vehicle models, types, and aging levels of power batteries. In addition, the continuously accumulated full-lifecycle data in the cloud provides data support for charging strategy iteration, battery health management, and next-generation battery technology development, comprehensively improving the safety, reliability, and economy of new energy vehicle power batteries.

[0059] This application also provides a computer-readable storage medium storing executable instructions. When the executable instructions are executed by at least one processor 301, the online detection method for lithium plating of power batteries based on cloud big data described in this application is implemented.

[0060] In some embodiments, the storage medium may be a magnetic random access memory (FRAM), a read-only memory (ROM), or a programmable read-only memory (PROM). Erasable Programmable Read-Only Memory (EPROM) Electrically Erasable Programmable Read-Only Memory (EEPROM) Read-only memory, flash memory, magnetic surface storage, optical disc, or CD-ROM ROM, Compact Disc Read It can be a memory such as a memory only; or it can be a device that includes one or any combination of the above-mentioned memories.

[0061] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0062] As an example, executable instructions may, but do not necessarily, correspond to files in the file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0063] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0064] The aforementioned computer-readable storage medium integrates multi-dimensional charging data through a cloud-based big data platform, fusing high-sensitivity feature sets of capacity increment IC curves, electrical parameters, and temperature differences. Employing a two-step hybrid machine learning model combining SVM and random forest, it achieves in-situ real-time detection of lithium plating during charging and accurate differentiation between reversible and irreversible lithium plating. This effectively solves the problems of offline damage, poor anti-interference performance of single parameters, and inability to distinguish lithium types inherent in traditional methods. Furthermore, for reversible lithium plating, it dynamically adjusts the charging strategy to promote lithium dendrite re-intercalation, extending battery cycle life; for irreversible lithium plating, it promptly stops charging and issues warnings to prevent short circuits and fires. Simultaneously, this solution requires no additional detection hardware, relying solely on data collected from existing BMS, significantly reducing detection costs. The model trained on massive cloud-based data possesses excellent generalization capabilities, allowing for batch adaptation to different vehicle models, types, and aging levels of power batteries. Moreover, the continuously accumulated full-lifecycle data in the cloud provides data support for charging strategy iteration, battery health management, and next-generation battery technology development, comprehensively improving the safety, reliability, and economy of new energy vehicle power batteries.

[0065] In the several embodiments provided in this application, it should be understood that the disclosed methods and electronic devices can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components may be combined, or integrated into another system, or some features may be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0066] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0067] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0068] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a platform server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0069] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A lithium precipitation online detection method for power batteries based on cloud big data, characterized in that, Includes the following steps: Collect multi-dimensional raw data during the charging process of new energy vehicle power batteries, and perform outlier removal, missing value imputation and data standardization in sequence to obtain standardized data; Extract a core feature set related to lithium plating from the standardized data. The core feature set includes capacity increment IC curve features, electrical parameter features, and temperature features. A hybrid machine learning model is constructed based on a lithium plating identification sub-model using Support Vector Machine (SVM) and a lithium plating type judgment sub-model using Random Forest. The hybrid model is then trained and optimized based on the core feature set. During the charging process of new energy vehicles, the BMS system transmits the multi-dimensional data collected in real time to the cloud. After processing, the data is input into the trained hybrid machine learning model. The lithium plating identification sub-model determines whether lithium plating has occurred. If lithium plating is determined to have occurred, the lithium plating type identification sub-model distinguishes between reversible and irreversible lithium plating. The judgment result is fed back to the vehicle-side BMS system, and the corresponding processing strategy is executed according to different judgment results. If there is weak lithium plating, normal charging is maintained. If reversible lithium plating occurs, the charging strategy is adjusted to promote lithium dendrite re-intercalation. If irreversible lithium plating occurs, charging is stopped and an early warning is issued.

2. The method according to claim 1, characterized in that, The multi-dimensional raw data is collected at a frequency of 10Hz, including: Electrical parameters: charging current I, charging voltage U, battery terminal voltage Ucell, SOC, coulombic efficiency η; Environmental and temperature parameters: cell temperature (Tcell), ambient temperature (Tenvironment); Basic battery parameters: battery model, production batch, number of cycles N, aging degree S, 0≤S≤1, S=1 is a brand new battery, S=0 is a scrapped battery; Charging parameters: charging rate C, charging mode.

3. The method according to claim 1, characterized in that, The outlier removal adopts the 3σ criterion. For any dimension of data X, if X exceeds the range of [μ-3σ,μ+3σ], it is determined to be an outlier and removed. μ is the data mean and σ is the standard deviation. The missing value imputation adopts the linear interpolation method. The data standardization adopts the Min-Max standardization method to map the data to the interval [0,1].

4. The method according to claim 1, characterized in that, The capacity increment IC curve features include five maximum peaks P1-P5, two minimum peaks P6-P7, and the battery terminal voltage value corresponding to the minimum peak P7. The electrical parameter characteristics include the charging current fluctuation coefficient KI, the voltage fluctuation coefficient KU, and the coulombic efficiency η; The temperature characteristic is the difference between the temperature of the battery cell and the ambient temperature, ΔT = Tcell - Tenvironment.

5. The method according to claim 1, characterized in that, The lithium plating identification sub-model is a support vector machine (SVM) model, which is used to achieve binary classification judgment between lithium plating-free and lithium plating-conducting samples. The lithium plating type judgment sub-model is a random forest model, which is used to classify samples judged to have lithium plating as either reversible or irreversible lithium plating. During model training, the dataset is divided into training set, validation set and test set in a ratio of 7:2:

1. The sample labels are divided into three categories: 0 - no lithium plating, 1 - reversible lithium plating, and 2 - irreversible lithium plating.

6. The method according to claim 5, characterized in that, The random forest model generates multiple decision trees through bootstrap sampling, with each decision tree having a maximum depth of 8. The Gini coefficient is used to measure feature importance and select core features. The penalty factor C of the SVM model is 10. The model was optimized using 5-fold cross-validation.

7. The method according to claim 1, characterized in that, Adjustments to the charging strategy for reversible lithium plating include: Adjust the current charging rate C to 50%-70% of the original rate; The temperature of individual battery cells should be controlled within the range of 25℃-35℃; Switch to constant voltage charging mode after SOC ≥ 80%; After adjustment, lithium plating is re-evaluated every first preset time. If no lithium plating is detected multiple times in a row, the normal charging strategy is restored. If reversible lithium plating is still detected after the second preset time, a maintenance reminder is triggered.

8. An online detection device for lithium plating in power batteries based on cloud-based big data, characterized in that, The device includes: The data acquisition module is used to collect multi-dimensional raw data during the charging process of new energy vehicle power batteries, and then perform outlier removal, missing value imputation and data standardization to obtain standardized data. The extraction module is used to extract a core feature set related to lithium plating from the standardized data. The core feature set includes capacity increment IC curve features, electrical parameter features, and temperature features. The module is used to build a hybrid machine learning model based on a lithium plating identification sub-model using support vector machine (SVM) and a lithium plating type judgment sub-model using random forest, and to train and optimize the hybrid model based on the core feature set. The judgment module is used during the charging process of new energy vehicles. It transmits multi-dimensional data collected in real time to the cloud through the BMS system. After processing, the data is input into the trained hybrid machine learning model. The lithium plating identification sub-model determines whether lithium plating has occurred. If lithium plating is determined to have occurred, the lithium plating type judgment sub-model distinguishes between reversible and irreversible lithium plating. The feedback module is used to send the judgment results back to the vehicle-side BMS system. Based on different judgment results, the corresponding processing strategy is executed. If there is weak lithium plating, normal charging is maintained. If reversible lithium plating occurs, the charging strategy is adjusted to promote lithium dendrite re-intercalation. If irreversible lithium plating occurs, charging is stopped and an early warning is issued.

9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the online detection method for lithium plating of power batteries based on cloud big data as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, performs the online detection method for lithium plating of power batteries based on cloud big data as described in any one of claims 1 to 7.

Images