No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Battery Cross-Operation-Condition Lifetime Prediction via Interpretable Feature Engineering Assisted Adaptive Machine Learning
Abstract
We develop an adaptive machine-learning framework that addresses cross-operation-condition battery lifetime prediction, particularly under extreme conditions. This framework uses correlation alignment to correct feature divergence under fast-charging and extremely fast-charging conditions. We report a linear correlation between feature adaptability and prediction accuracy. Higher adaptability generally leads to better prediction accuracy, aiding efficient feature engineering. Our analysis shows that the first 120 cycles provide sufficient information for lifetime prediction, and extending data to the first 320 cycles only marginally improves prediction accuracy. An early prediction using only one feature at the 20th cycle produces a 93.3% accuracy, saving up to 99.4% computation time and repetitive tests. Our quantitative adaptability evaluation enhances prediction accuracy while reducing information redundancy via proper feature and cycle selections. The proposed framework is validated under another unseen complex operation condition with a 90.3% accuracy without prior knowledge. L ithium-ion batteries (LIBs) have been broadly deployed in consumer electronics, 1 electric vehicles, 2 battery energy storage systems, 3 and smart grid applications 4 due to their high energy density, 5 wide working temperature range, 6 and mature technology ecology. However, such batteries continuously degrade during cycling, leading to severe issues such as capacity drop, 7 temperature rise, 8 cell-to-cell inconsistency, 9,10 and shortened lifetime. For safety concerns, it is therefore essential for battery management systems to accurately predict the state of health (SOH) and the remaining useful life (RUL) of batteries. In addition, accurate knowledge of the SOH and RUL helps to evaluate the batteries for next-stage decision-making, such as repurposing in second life 11 and recycling routine selection 12 at the end of life. Therefore, prediction of the SOH and RUL is critically important throughout a battery's life, while it remains challenging due to the constantly changing operation conditions. Much previous research reported mechanism-driven and semiempirical prediction methods. For the mechanism-informed methods, a pseudo-two-dimensional model, 13 a single-particle model, 14 electrochemical impedance spectros-copy, 15,16 distribution of relaxation time, 17,18 an equivalent circuit model, 19 incremental capacity analysis, 20 and differential voltage analysis 21 are advantageous in accurately predicting microscopic degradation, such as lithium plating, 22 solid-electrolyte-interphase (SEI) formation, 23 loss of lithium inventory (LLI), 24 and loss of active materials (LAM). 25 However, the diverse operation conditions, such as dynamic charging and discharging protocols, 26 state of charge, 27 and ambient temperatures, 28 can cause significant divergence in the primary degradation mechanisms, leading to poor performance in practical use. In contrast to the mechanism-driven method, semiempirical methods are developed by assuming equivalent circuit model 29 and empirical battery degradation patterns by deliberately fitting the historical usage parameters into the
... The graphite comes into direct contact with the electrolyte, initiating reactions that consume both electrolyte and lithium-ions [37]. Consequently, the SEI film continues to grow and thicken, and the cathode-electrolyte interface (CEI) film also gradually grows with battery aging, causing gradually increased internal resistance [38,39]. Therefore, the battery resistance is extracted as one of the features. ...
... Elevated temperature exacerbates the aging-related side reactions within the temperature range above room temperature (25°C ), such as the growth of the SEI film and the decomposition of the cathode material [43]. On the other hand, low temperature can lead to irreversible consumption of active lithium-ions and hinder the intercalation of lithium-ions [39]. Therefore, the operating temperature of EVs has a significant impact on the capacity degradation of the battery pack. ...
... P. K. Jones et al., by integrating EIS with probabilistic machine learning, have proposed a new method for accurately predicting the performance of lithium-ion batteries under uneven usage conditions [16]. Researchers have also used feature-matching-based transfer learning methods [17] or interpretable feature engineeringassisted adaptive machine learning approaches [18] to achieve lithium-ion battery life prediction under similar uneven conditions, which can also be used for more accurately estimating the capacity of lithium-ion batteries. Some researchers have also introduced a federated learning architecture allowing multiple entities, such as battery recycling stations, manufacturers, or repair centers, to collaboratively train a global model without sharing their data, by combining EIS with other relevant parameters such as voltage, current, and temperature as features, the federated machine learning model can more accurately predict the battery's health status and remaining life, guiding effective sorting and recycling of retired batteries [19]. ...
Electrochemical Impedance Spectroscopy (EIS) plays a crucial role in characterizing the internal electrochemical states of lithium-ion batteries and proves to be effective for estimating battery states. Traditional EIS measurement, however, requires expensive electrochemical workstations with time-consuming signal injection, especially in low-frequency regions, thus limiting its practical applications. Here we show that applying our proposed pulse-like Binary Multi-Frequency Signals (BMFS) as the excitation signal in the EIS measurement, which simultaneously possesses numerous frequency components and maintains high energy at each frequency component, will significantly improve test speed while retaining accuracy. The applicability of the BMFS under various cathode material types, including nickel cobalt manganese (NCM), lithium cobalt oxide (LCO), and lithium iron phosphate (LFP) is demonstrated. The robustness of the signal is experimentally verified through varying C-rates and measurement window lengths. The BMFS, requiring only 30 s per test, can achieve test results with an amplitude error of 1% and a phase error of 1° as compared with those obtained from traditional EIS tests. Moreover, BMFS can also be applied in online EIS measurement scenarios, favorable for real-world applications. This work enables accurate and rapid acquisition of EIS results, which is currently expensive and time-consuming to obtain, ensuring a faster and more nuanced characterization of the internal states of many battery systems in an affordable and accessible manner, especially in data-driven and machine-learning approaches.
... Fan et al. performed an electrochemical impedance spectroscopy test at 5% SOC on the retired power batteries, thus saving long data acquisition time that was otherwise needed for higher SOC levels [25]. Another significant difference between SOH diagnosis for first-life and recycling scenarios is that the latter typically encounters considerable heterogeneity in terms of cathode chemistries, historical usages, cell-to-cell inconsistency, physical formats, capacity designs, and SOH distributions, which challenges diagnosis tasks [22,[26][27][28][29][30]39]. Therefore, Ran et al. first clustered similar retired batteries and then estimated remaining capacities for clustered batteries with a two-step learning method combining a pulse test [31] and reported a robust result when using an inter-cycle shift in voltage response as a health indicator in a wide SOC range [32]. ...
... Conventional screening technologies often focus on external characteristics that are easily observable but can miss critical internal processes that are crucial to battery health and safety. AI, with its advanced techniques like deep learning, transfer learning, and continual learning, enables a more in-depth analysis [266,267]. For instance, AI has been pivotal in identifying battery side reactions by characterizing their implicit correlations with battery thickness [269], voltage signals [269,270], and some aging parameters [271,272]. ...
Anticipating the imminent surge of retired lithium-ion batteries (R-LIBs) from electric vehicles, the need for safe, cost-effective and environmentally friendly disposal technologies has escalated. This paper seeks to offer a comprehensive overview of the entire disposal framework for R-LIBs, encompassing a broad spectrum of activities, including screening, repurposing and recycling. Firstly, we delve deeply into a thorough examination of current screening technologies, shifting the focus from a mere enumeration of screening methods to the exploration of the strategies for enhancing screening efficiency. Secondly, we outline battery repurposing with associated key factors, summarizing stationary applications and sizing methods for R-LIBs in their second life. A particular light is shed on available reconditioning solutions, demonstrating their great potential in facilitating battery safety and lifetime in repurposing scenarios and identifying their techno-economic issues. In the realm of battery recycling, we present an extensive survey of pre-treatment options and subsequent material recovery technologies. Particularly, we introduce several global leading recyclers to illustrate their industrial processes and technical intricacies. Furthermore, relevant challenges and evolving trends are investigated in pursuit of a sustainable end-of-life management and disposal framework. We hope that this study can serve as a valuable resource for researchers, industry professionals and policymakers in this field, ultimately facilitating the adoption of proper disposal practices.
... In the past few years, machine learning has emerged as a viable tool to tackle open questions in all battery fields. In other batteryrelated topics, machine learning has recently allowed us to automatically discover complex battery mechanisms [17][18][19] , predict remaining useful life [20][21][22][23][24] , evaluate the state of health 19,25,26 , optimize the cycling profile 27,28 , approximate the failure distribution 29 , even to guide the battery design 30,31 , and predict life-long performance immediately after manufacturing 32 . In the case of battery recycling, few works have investigated machine learning regarding cathode materials 33,34 , which blames the scarce battery data, especially for those cycled to the endof-life stage. ...
Unsorted retired batteries with varied cathode materials hinder the adoption of direct recycling due to their cathode-specific nature. The surge in retired batteries necessitates precise sorting for effective direct recycling, but challenges arise from varying operational histories, diverse manufacturers, and data privacy concerns of recycling collaborators (data owners). Here we show, from a unique dataset of 130 lithium-ion batteries spanning 5 cathode materials and 7 manufacturers, a federated machine learning approach can classify these retired batteries without relying on past operational data, safeguarding the data privacy of recycling collaborators. By utilizing the features extracted from the end-of-life charge-discharge cycle, our model exhibits 1% and 3% cathode sorting errors under homogeneous and heterogeneous battery recycling settings respectively, attributed to our innovative Wasserstein-distance voting strategy. Economically, the proposed method underscores the value of precise battery sorting for a prosperous and sustainable recycling industry. This study heralds a new paradigm of using privacy-sensitive data from diverse sources, facilitating collaborative and privacy-respecting decision making for distributed systems.
... Feature selection involves selecting a subset of all features that can be divided into filter, embedded, and wrapper methods [40,41]. Considering that the adaptability and prediction capability of features vary with different battery systems and operation conditions [42], cross-condition feature engineering is developed, which utilizes adaptive ML frameworks to address the features divergence [43]. The features divergence under different charging conditions can be corrected using correlation alignment-aided multilayer perceptron, together with effective removal of redundant features. ...
Metal-ion batteries (MIBs), including alkali metal-ion (Li+, Na+, and K+), multi-valent metal-ion (Zn2+, Mg2+, and Al3+), metal-air, and metal-sulfur batteries, play an indispensable role in electrochemical energy storage. However, the performance of
MIBs is significantly influenced by numerous variables, resulting in multi-dimensional and long-term challenges in the field of battery research and performance enhancement. Machine learning (ML), with its capability to solve intricate tasks and perform robust data processing, is now catalyzing a revolutionary transformation in the development of MIB materials and devices. In this review, we summarize the utilization of ML algorithms that have expedited research on MIBs over the past five years. We present an extensive overview of existing algorithms, elucidating their details, advantages, and limitations in various applications, which encompass electrode screening, material property prediction, electrolyte formulation design, electrode material characterization, manufacturing parameter optimization, and real-time battery status monitoring. Finally, we propose potential solutions and future directions for the application of ML in advancing MIB development.
... In our previous work, we have proved that correlation alignment (CORAL) has the ability to correct the feature divergence cross working conditions, and the CORAL-aided MLP outperforms the fine-tuning baseline model [47]. According to the results depicted in Section 2.2, we aim to further simplify the feature mapping relation under domain divergence by assuming a linear relation, and the principle of the FM-TL method is as follows. ...
Accurate capacity estimation is essential in the management of lithium-ion batteries, as it guarantees the safety and dependability of battery-powered systems. However, direct measurement of battery capacity is challenging due to the unpredictable working conditions and intricate electrochemical characteristics, which complicates the identification of battery degradation. In this work, through in-depth analysis of battery aging data, an incremental slope (IS) aided feature extraction method is proposed to obtain universal multidimensional features that adapt to different working conditions. With the extracted features, a simple multilayer perceptron (MLP) is used to achieve high-precision capacity estimation. Furthermore, a feature matching based transfer learning (FM-TL) method is proposed to automatically adapt the capacity estimation across different types of batteries that are cycled under various working conditions. 158 batteries covering five material types and 15 working conditions are used to validate the proposed method. Results suggest that the MLP model can provide an accurate capacity estimation, where the overall mean absolute percentage error (MAPE) and root mean square percentage error (RMSPE) are limited to 1.22% and 1.61%, respectively. Furthermore, compared with the traditional fine-tuning method, the overall MAPE and RMSPE under various transfer learning application scenarios respectively decrease by up to 78.23% and 75.31%, indicating that the FM-TL method is promising to construct a reliable transfer learning path, which improves the accuracy and reliability of capacity estimation when applied to various target domains.
Deep eutectic solvents (DESs) have been widely applied to recover spent Lithium-ion batteries (LIBs), however, developing effective and efficient systems for cathode leaching via traditional trial-and-error method requires substantial efforts....
The performance of electrochemical batteries is intricately tied to the physicochemical environments established by their employed electrolytes. Traditional battery designs utilizing a single electrolyte often impose identical anodic and cathodic redox conditions, limiting the ability to optimize redox environments for both anode and cathode materials. Consequently, advancements in electrolyte technologies are pivotal for addressing these challenges and fostering the development of next‐generation high‐performance electrochemical batteries. This review categorizes perspectives on electrolyte technology into three key areas: additives engineering, comprehensive component analysis encompassing solvents and solutes, and the effects of concentration. By summarizing significant studies, the efficacy of electrolyte engineering is highlighted, and the review advocates for further exploration of optimized component combinations. This review primarily focuses on liquid electrolyte technologies, briefly touching upon solid‐state electrolytes due to the former greater vulnerability to electrode and electrolyte interfacial effects. The ultimate goal is to generate increased awareness within the battery community regarding the holistic improvement of battery components through optimized combinations.
State of health (SOH) estimation of e-mobilities operated in real and dynamic conditions is essential and challenging. Most of existing estimations are based on a fixed constant current charging and discharging aging profiles, which overlooked the fact that the charging and discharging profiles are random and not complete in real application. This paper investigates the influence of feature engineering on the accuracy of different machine learning (ML)-based SOH estimations acting on different recharging sub-profiles where a realistic battery mission profile is considered. Fifteen features were extracted from the battery partial recharging profiles, considering different factors such as starting voltage values, charge amount, and charging sliding windows. Then, features were selected based on a feature selection pipeline consisting of filtering and supervised ML-based subset selection. Multiple linear regression (MLR), Gaussian process regression (GPR), and support vector regression (SVR) were applied to estimate SOH, and root mean square error (RMSE) was used to evaluate and compare the estimation performance. The results showed that the feature selection pipeline can improve SOH estimation accuracy by 55.05%, 2.57%, and 2.82% for MLR, GPR and SVR respectively. It was demonstrated that the estimation based on partial charging profiles with lower starting voltage, large charge, and large sliding window size is more likely to achieve higher accuracy. This work hopes to give some insights into the supervised ML-based feature engineering acting on random partial recharges on SOH estimation performance and tries to fill the gap of effective SOH estimation between theoretical study and real dynamic application.
Battery screening is the key segment in secondary applications. The benchmark for conventional methods is mainly based on the series connection and makes parameter difference as the screening index a gold standard. However, because of self-balancing current in parallel connection, the existence of a certain degree of parameter difference is allowed and parameter difference may not be the best option, which leads to lower screening efficiency due to the higher uniform of parameters. This work first identifies the boundary of parameter difference and provides the ideal working point (IWP), which is related to maximum capacity utilization efficiency, as a novel screening index for parallel connection derived from the current distribution. A modified shepherd model is employed to calculate the IWPs and is verified that the maximum dynamic error is below 1.1%. Therefore, it is quick to achieve screening by judging whether the IWP falls within the normal working range. The results show that the maximum capacity utilization efficiency always occurs at the load current close to the IWP, which verifies the validity of IWP. Compared with the conventional method, the proposed method is validated based on a case study to improve screening efficiency and provides different ideas for flexible grouping.
Accurate forecasting of lithium-ion battery performance is essential for easing consumer concerns about the safety and reliability of electric vehicles. Most research on battery health prognostics focuses on the research and development setting where cells are subjected to the same usage patterns. However, in practical operation, there is great variability in use across cells and cycles, thus making forecasting challenging. To address this challenge, here we propose a combination of electrochemical impedance spectroscopy measurements with probabilistic machine learning methods. Making use of a dataset of 88 commercial lithium-ion coin cells generated via multistage charging and discharging (with currents randomly changed between cycles), we show that future discharge capacities can be predicted with calibrated uncertainties, given the future cycling protocol and a single electrochemical impedance spectroscopy measurement made immediately before charging, and without any knowledge of usage history. The results are robust to cell manufacturer, the distribution of cycling protocols, and temperature. The research outcome also suggests that battery health is better quantified by a multidimensional vector rather than a scalar state of health.
Accurate capacity estimation is crucial for the reliable and safe operation of lithium-ion batteries. In particular, exploiting the relaxation voltage curve features could enable battery capacity estimation without additional cycling information. Here, we report the study of three datasets comprising 130 commercial lithium-ion cells cycled under various conditions to evaluate the capacity estimation approach. One dataset is collected for model building from batteries with LiNi0.86Co0.11Al0.03O2-based positive electrodes. The other two datasets, used for validation, are obtained from batteries with LiNi0.83Co0.11Mn0.07O2-based positive electrodes and batteries with the blend of Li(NiCoMn)O2 - Li(NiCoAl)O2 positive electrodes. Base models that use machine learning methods are employed to estimate the battery capacity using features derived from the relaxation voltage profiles. The best model achieves a root-mean-square error of 1.1% for the dataset used for the model building. A transfer learning model is then developed by adding a featured linear transformation to the base model. This extended model achieves a root-mean-square error of less than 1.7% on the datasets used for the model validation, indicating the successful applicability of the capacity estimation approach utilizing cell voltage relaxation. Accurate capacity estimation is crucial for lithium-ion batteries' reliable and safe operation. Here, the authors propose an approach exploiting features from the relaxation voltage curve for battery capacity estimation without requiring other previous cycling information.
The equivalent circuit model (ECM) is a battery model often used in the battery management system (BMS) to monitor and control lithium-ion batteries (LIBs). The accuracy and complexity of the ECM, hence, are very important. State of charge (SOC) and temperature are known to affect the parameters of the ECM and have been integrated into the model effectively. However, the effect of the state of health (SOH) on these parameters has not been widely investigated. Without a good understanding of the effect of SOH on ECM parameters, parameter identification would have to be done manually through calibration, which is inefficient. In this work, experiments were performed to investigate the effect of SOH on Thevenin ECM parameters, in addition to the effect of SOC and temperature. The results indicated that with decreasing SOH, the ohmic resistance and the polarization resistance increase while the polarization capacitance decreases. An empirical model was also proposed to represent the effect of SOH, SOC, and temperature on the ECM parameters. The model was then validated experimentally, yielding good results, and found to improve the accuracy of the Thevenin model significantly. With low complexity and high accuracy, this model can be easily integrated into real-world BMS applications.
The degradation of batteries is complex and dependent on several internal mechanisms. Variations arising from manufacturing uncertainties and real-world operating conditions make battery lifetime prediction challenging. Here, we introduce a deep learning-based battery health prognostics approach to predict the future degradation trajectory in one shot without iteration or feature extraction. We also predict the end-of-life point and the knee-point. The model correctly learns about intrinsic variability caused by manufacturing differences, and is able to make accurate cell-specific predictions from just 100 cycles of data, and the performance improves over time as more data become available. Validation in an embedded device is demonstrated with the best-case median prediction error over the lifetime being 1.1% with normal data and 1.3% with noisy data. Compared to state-of-the-art approaches, the one-shot approach shows an increase in accuracy as well as in computing speed by up to 15 times. This work further highlights the effectiveness of data-driven approaches in the domain of health prognostics.
Lithium-ion batteries are ubiquitous in applications ranging from portable electronics to electric vehicles. Irrespective of the application, reliable real-time estimation of battery state of health (SOH) by on-board computers is crucial to the safe operation of the battery, ultimately safeguarding asset integrity. In this Article, we design and evaluate a machine learning pipeline for estimation of battery capacity fade—a metric of battery health—on 179 cells cycled under various conditions. The pipeline estimates battery SOH with an associated confidence interval by using two parametric and two non-parametric algorithms. Using segments of charge voltage and current curves, the pipeline engineers 30 features, performs automatic feature selection and calibrates the algorithms. When deployed on cells operated under the fast-charging protocol, the best model achieves a root-mean-squared error of 0.45%. This work provides insights into the design of scalable data-driven models for battery SOH estimation, emphasizing the value of confidence bounds around the prediction. The pipeline methodology combines experimental data with machine learning modelling and could be applied to other critical components that require real-time estimation of SOH. Rechargeable lithium-ion batteries play a crucial role in many modern-day applications, including portable electronics and electric vehicles, but they degrade over time. To ensure safe operation, a battery’s ‘state of health’ should be monitored in real time, and this machine learning pipeline, tested on a variety of charging conditions, can provide such an online estimation of battery state of health.
Heat generation and therefore thermal transport plays a critical role in ensuring performance, ageing and safety for lithium-ion batteries (LIB). Increased battery temperature is the most important ageing accelerator. Understanding and managing temperature and ageing for batteries in operation is thus a multiscale challenge, ranging from the micro/nanoscale within the single material layers to large, integrated LIB packs. This paper includes an extended literature survey of experimental studies on commercial cells investigating the capacity and performance degradation of LIB. It compares the degradation behavior in terms of the influence of operating conditions for different chemistries and cell sizes. A simple thermal model for linking some of these parameters together is presented as well. While the temperature appears to have a large impact on ageing acceleration above room temperature during cycling for all studied cells, the effect of SOC and C rate appear to be rather cell dependent.Through the application of new simulations, it is shown that during cell testing, the actual cell temperature can deviate severely from the reported temperature depending on the thermal management during testing and C rate. It is shown, that the battery lifetime reduction at high C rates can be for large parts due to an increase in temperature especially for high energy cells and poor cooling during cycling studies. Measuring and reporting the actual battery (surface) temperature allow for a proper interpretation of results and transferring results from laboratory experiments to real applications.
The objective of this paper is development of a remaining useful life (RUL) prediction algorithm based on estimation of parameters of an enhanced single particle model (eSPM) that could be implemented using vehicle charging data. First, we use data from an aging study conducted on LMO-NMC-cathode graphite-anode battery cells to develop an eSPM that can predict the evolution of parameters associated with the aging of the battery. In particular, the parameters we estimate in this work as indicators of state of health (SOH) are number of moles of cyclable lithium and Ohmic resistance. A method is demonstrated for estimating these parameters from experimental data, and it is shown that they are correlated with battery SOH measured from the experimental aging study. Finally, a composite SOH metric derived from the estimated eSPM parameters is used to design a RUL predictor based on a particle filter (PF) that can predict the RUL utilizing the evolution of the SOH metric. The RUL estimation algorithm is validated using experimental data collected on several LMO-NMC battery cells, showing that it is possible to infer battery SOH and RUL from charging data readily available in plug-in battery-electric or hybrid vehicles.
Forecasting the state of health and remaining useful life of Li-ion batteries is an unsolved challenge that limits technologies such as consumer electronics and electric vehicles. Here, we build an accurate battery forecasting system by combining electrochemical impedance spectroscopy (EIS)—a real-time, non-invasive and information-rich measurement that is hitherto underused in battery diagnosis—with Gaussian process machine learning. Over 20,000 EIS spectra of commercial Li-ion batteries are collected at different states of health, states of charge and temperatures—the largest dataset to our knowledge of its kind. Our Gaussian process model takes the entire spectrum as input, without further feature engineering, and automatically determines which spectral features predict degradation. Our model accurately predicts the remaining useful life, even without complete knowledge of past operating conditions of the battery. Our results demonstrate the value of EIS signals in battery management systems. Forecasting the state of health and remaining useful life of batteries is a challenge that limits technologies such as electric vehicles. Here, the authors build an accurate battery performance forecasting system using machine learning.
Simultaneously optimizing many design parameters in time-consuming experiments causes bottlenecks in a broad range of scientific and engineering disciplines1,2. One such example is process and control optimization for lithium-ion batteries during materials selection, cell manufacturing and operation. A typical objective is to maximize battery lifetime; however, conducting even a single experiment to evaluate lifetime can take months to years3,4,5. Furthermore, both large parameter spaces and high sampling variability3,6,7 necessitate a large number of experiments. Hence, the key challenge is to reduce both the number and the duration of the experiments required. Here we develop and demonstrate a machine learning methodology to efficiently optimize a parameter space specifying the current and voltage profiles of six-step, ten-minute fast-charging protocols for maximizing battery cycle life, which can alleviate range anxiety for electric-vehicle users8,9. We combine two key elements to reduce the optimization cost: an early-prediction model⁵, which reduces the time per experiment by predicting the final cycle life using data from the first few cycles, and a Bayesian optimization algorithm10,11, which reduces the number of experiments by balancing exploration and exploitation to efficiently probe the parameter space of charging protocols. Using this methodology, we rapidly identify high-cycle-life charging protocols among 224 candidates in 16 days (compared with over 500 days using exhaustive search without early prediction), and subsequently validate the accuracy and efficiency of our optimization approach. Our closed-loop methodology automatically incorporates feedback from past experiments to inform future decisions and can be generalized to other applications in battery design and, more broadly, other scientific domains that involve time-intensive experiments and multi-dimensional design spaces.
Sustainable high performance batteries with desirable application-specific performance characteristics call for accurate early predictions of future capacity loss to advance the management of the battery. Li-ion cells exhibit a slow capacity degradation up to a knee-point, after which the degradation accelerates rapidly until the cell's end-of-life. Using capacity degradation data, we propose a robust method to identify the knee-point in capacity fade curves. In a new approach to knee research, we propose the concept of knee-onset, which marks the beginning of the nonlinear degradation, and provide a simple and robust identification mechanism for it. We link cycle life, knee-point and knee-onset so that predicting/identifying one promptly reveals the others. Our results, on data featuring high C-rates (1C-8C), show that, on average, the knee occurs at 95\% capacity with its onset at 97.1\% capacity, and that the knee and its onset are on average 108 cycles apart.
After the critical identification step, we employ a range of machine learning techniques for the early prediction of the knee-point and knee-onset. Our models predict the knee-point and onset quantitatively with 9.4\% error using only information on the first 50 cycles of the cells' life, and classify cells according to their knee-point into short, medium or long-range with 88~-~90\% accuracy using only information of the first 3 to 5 cycles. Our accuracy levels are on par with existing literature for end-of-life prediction, although we address here the more involved problem of knee prediction.
All estimations are enriched with confidence/credibility intervals. The uncertainty regarding the estimations produced by the machine learning models is quantified through prediction intervals. These provide a risk criteria that insurers and energy storage application manufacturers alike can use for cell/battery warranties. Lastly, our classification model provides a tool for cell manufacturers to speed up the validation of cell production techniques.
A fast and accurate capacity estimation method for lithium-ion batteries is developed. This method applies our developed semi-empirical model to a discharge curve of a lithium-ion battery for the determination of its maximum stored charge capacity after each discharge cycle. This model provides an accurate state-of-health (SoH) estimation with a difference of less than 2.22% when compared with the electrochemistry-based electrical (ECBE) SoH calculation. The model parameters derived from a lithium-ion battery can also be applied to other cells in the same pack with less than 2.5% difference from the complex ECBE model, showing the extendibility of the model. The parameters (k1, k2, and k3) calculated in the work can also be used to study the changes in battery internal structure, such as capacity losses at normal conditions, as well as cycling at high temperatures. The time for estimation after each discharge cycle is only 5 s, making it is suitable for on-line in-situ estimation.
Accurately predicting the lifetime of complex, nonlinear systems such as lithium-ion batteries is critical for accelerating technology development. However, diverse aging mechanisms, significant device variability and dynamic operating conditions have remained major challenges. We generate a comprehensive dataset consisting of 124 commercial lithium iron phosphate/graphite cells cycled under fast-charging conditions, with widely varying cycle lives ranging from 150 to 2,300 cycles. Using discharge voltage curves from early cycles yet to exhibit capacity degradation, we apply machine-learning tools to both predict and classify cells by cycle life. Our best models achieve 9.1% test error for quantitatively predicting cycle life using the first 100 cycles (exhibiting a median increase of 0.2% from initial capacity) and 4.9% test error using the first 5 cycles for classifying cycle life into two groups. This work highlights the promise of combining deliberate data generation with data-driven modelling to predict the behaviour of complex dynamical systems. Accurately predicting battery lifetime is difficult, and a prediction often cannot be made unless a battery has already degraded significantly. Here the authors report a machine-learning method to predict battery life before the onset of capacity degradation with high accuracy.
Batteries will play critical roles in modernizing energy grids, as they will allow a greater penetration of renewable energy and perform applications that better match supply with demand. Applying storage technology is a business decision that requires potential revenues to be accurately estimated to determine the economic viability, which requires models that consider market rules and prices, along with battery and application-specific constraints. Here we use models of storage connected to the California energy grid and show how the application-governed duty cycles (power profiles) of different applications affect different battery chemistries. We reveal critical trade-offs between battery chemistries and the applicability of energy content in the battery and show that accurate revenue measurement can only be achieved if a realistic battery operation in each application is considered. The findings in this work could call for a paradigm shift in how the true economic values of energy storage devices could be assessed. © 2018, The Author(s), under exclusive licence to Springer Nature Limited.
Lithium-ion batteries, with high energy density (up to 705 Wh/L) and power density (up to 10,000 W/L), exhibit high capacity and great working performance. As rechargeable batteries, lithium-ion batteries serve as power sources in various application systems. Temperature, as a critical factor, significantly impacts on the performance of lithium-ion batteries and also limits the application of lithium-ion batteries. Moreover, different temperature conditions result in different adverse effects. Accurate measurement of temperature inside lithium-ion batteries and understanding the temperature effects are important for the proper battery management. In this review, we discuss the effects of temperature to lithium-ion batteries at both low and high temperature ranges. The current approaches in monitoring the internal temperature of lithium-ion batteries via both contact and contactless processes are also discussed in the review.
The lifetime and safety of lithium-ion batteries are key requirements for successful market introduction of electro mobility as well as for the application in stationary use. Lithium plating, leading both to ageing as well as safety risks, is known to play a crucial role in system design of the application. To gain knowledge of different influence factors on lithium plating, low temperature ageing tests are performed in this work. Lithium-ion batteries of various types are tested under various operation conditions such as temperature, current, state of charge, charging strategy as well as state of health. To analyse the ageing behaviour capacity fade and resistance increase are tracked over lifetime. To further investigate the underlying degradation mechanisms, differential voltage curves and impedance spectra are analysed as well as a post-mortem analysis of anode degradation is performed for a selected technology. The results confirm the deposition of metallic lithium or lithium compounds in the porous structure and suggest a strongly inhomogeneous deposition over the electrode thickness with a dense deposition layer close to the separator for the considered cell. The plurality of the investigated technologies demonstrates large differences between different technologies concerning low temperature behaviour and gives insight to the impact of cell properties. For the investigated technologies no correlation is observed between the ageing behaviour and cell size or cathode material. However, cells rated to provide high power are found to be subject to faster degradation at low temperatures compared to high energy cells. For application this result shows that cells designed for high current rates are not necessarily providing a good low temperature performance. 2
Degradation in lithium ion (Li-ion) battery cells is the result of a complex interplay of a host of different physical and chemical mechanisms. The measurable, physical effects of these degradation mechanisms on the cell can be summarised in terms of three degradation modes, namely loss of lithium inventory, loss of active positive electrode material and loss of active negative electrode material. The different degradation modes are assumed to have unique and measurable effects on the open circuit voltage (OCV) of Li-ion cells and electrodes. The presumptive nature and extent of these effects has so far been based on logical arguments rather than experimental proof. This work presents, for the first time, experimental evidence supporting the widely reported degradation modes by means of tests conducted on coin cells, engineered to include different, known amounts of lithium inventory and active electrode material. Moreover, the general theory behind the effects of degradation modes on the OCV of cells and electrodes is refined and a diagnostic algorithm is devised, which allows the identification and quantification of the nature and extent of each degradation mode in Li-ion cells at any point in their service lives, by fitting the cells' OCV.
Batteries are central to modern society. They are no longer just a convenience but a critical enabler of the transition to a resilient, low-carbon economy. Battery development capabilities are provided by communities spanning materials discovery, battery chemistry and electrochemistry, cell and pack design, scale-up, manufacturing, and deployments. Despite their relative maturity, data-science practices among these diverse groups are far behind the state of the art in other fields, which have demonstrated an ability to significantly improve innovation and economic impact. The negative consequences of the present paradigm include incremental improvements but few breakthroughs, significant manufacturing uncertainties, and cascading investment risks that collectively slow deployments. The primary roadblock to a battery-data-science renaissance is the requirement for large amounts of high-quality data, which are not available in the current fragmented ecosystem. Here, we identify gaps and propose principles that enable the solution by building a robust community of data hubs with standardized practices and flexible sharing options that will seed advanced tools spanning innovation to deployment. Precedents are offered that demonstrate that both public good and immense economic gains will arise from sharing valuable battery data. The proposed Battery Data Genome looks to broadly transform innovations and revolutionize their translation from research to societal impact.
Lithium-ion batteries (LIBs) are widely used in the assembly of battery packs for electric vehicles and energy storage grids due to their high power density, low self-discharge rate and reasonable costs. Accurate estimation of state of health (SOH) and remaining useful life (RUL) are crucial challenges in developing battery management systems (BMS). In this paper, differential thermal voltammetry (DTV) signal analysis methods and recursive neural networks data-driven methods are combined to approach battery degradation tracking. Firstly, with the Savitzky-Golay (SG) method and Pearson correlation analysis, the DTV curve is smoothed, and three useful feature variables are extracted from different dimensions, bridging signaling characteristics and phase transition characteristics. Then four recursive neural networks are constructed and compared based on NASA databases. The Bayesian optimization method is applied to improve hyperparameter values and the Monte Carlo (MC) simulation is used to quantify uncertainties. The proposed data-driven method can predict the RUL and estimate the SOH of battery accurately. The root mean square error (RMSE) for prediction results could reach below 1% and the capacity rebound phenomenon could be captured as well. The proposed integrated degradation model can contribute to the real-time prediction and optimization of battery health conditions based on cloud computing platform, promoting the continuous development of cloud battery management systems in framework of Cyber Hierarchy and Interactional Network (CHAIN).
A comprehensive understanding of multiple Li kinetics in batteries is essential to break the limitations of mechanism study and materials design. Various kinetic processes with specific relaxation features can be clearly identified in timescales. Extracting and analyzing the timescale information in batteries will provide insights into investigating kinetic issues such as ionic conductions, charge transfer, diffusions, interfacial evolutions, and other unknown kinetic processes. In this regard, the timescale identification is an important method to combine with the non-destructive impedance characterizations in length scale for online battery monitoring. This perspective introduces and advocates the timescale characterization in the views of the basic timescale property in batteries, employing the concept of distribution of relaxation time (DRT) and presenting successful applications for battery diagnosis. In the future, we suggest that timescale characterizations will become powerful tools for data extraction and dataset building for various battery systems, which can realize data-driven machine learning modeling for practical application situations such as retired battery rapid sorting and battery status estimations.
Mechanochemical degradation processes such as the fracture of cathode particles play a major role in limiting the service life of advanced lithium-ion batteries (LIBs). In order to help alleviate the degradation of battery performance, it is necessary to measure the relationship between the degradation of the mechanical properties of cathodes and their concomitant degradation of electrochemical performance. In this review, measurements of the mechanical properties of LIB cathode materials are summarized from the literature, along with the range of experimental methods used in their determination. Dimensional changes that accompany charge and discharge are compared for active materials of olivine, spinel, and layered atomic structures. The sensitivity of indentation hardness, Young’s modulus and fracture strength to grain size, porosity, state of charge and charge/discharge history are critically reviewed and discussed with reference to the behavior of conventional, electrically inactive solids. This approach allows for the identification of microstructural properties that dictate the mechanical properties of LIB cathode materials.
Rechargeable batteries, such as LiFePO4/graphite cells, age differently by variability in manufacturing, charging (energy inflow) policy, temperature, discharging conditions, etc. Great economic and environmental value can be extracted if we can predict how a battery ages and ascertain its current state of health and residual useful life, based on just a few cycles of testing. Here, by developing novel-architecture deep neural networks with a special convolutional training strategy and taking advantage of recently published battery cycling data, we show that one can predict the residual life of a battery to a mean absolute percentage error of 6.46%, using only one cycle of testing. The cycle-by-cycle profiles, such as discharge voltage, capacity, and power curves of any given cycle, of used batteries with unknown age can also be accurately predicted for the first time. Moreover, our models can extract data-driven features from the data which were much more influential on the predicted properties than human-picked features. This work has shown that single cycle data contains a sufficient amount of information to predict essential battery properties with high accuracy. It is expected to provide tremendous economic and environmental benefits since reuse and recycling of batteries can be better planned and less lithium-ion batteries end up in landfills.
Advancing lithium-ion battery technology requires the optimization of cycling protocols. A new data-driven methodology is demonstrated for rapid, accurate prediction of the cycle life obtained by new cycling protocols using a single test lasting only 3 cycles, enabling rapid exploration of cycling protocol design spaces with orders of magnitude reduction in testing time. We achieve this by combining lifetime early prediction with a hierarchical Bayesian model (HBM) to rapidly predict performance distributions without the need for extensive repetitive testing. The methodology is applied to a comprehensive dataset of lithium-iron-phosphate/graphite comprising 29 different fast-charging protocols. HBM alone provides high protocol-lifetime prediction performance, with 6.5% of overall test average percent error, after cycling only one battery to failure. By combining HBM with a battery lifetime prediction model, we achieve a test error of 8.8% using a single 3-cycle test. In addition, the generalizability of the HBM approach is demonstrated for lithium-manganese-cobalt-oxide/graphite cells.
The electrification of mass transportation is hailed as a solution for reducing global greenhouse-gas emissions and dependence on unsustainable energy sources. The annual sales of electric vehicles (EVs) has continued to rise since 2011, with a global sale of EVs of 2.1 million in 2019. This increase in sales is mainly due to continued improvement in the cost and performance of commercial EVs, increased EV options available to consumers, and environmental awareness. However, despite the positive outlook, EVs still face major
challenges that hinder their rapid and widespread adoption: limited driving range, long charging times, and a lack of sufficient charging infrastructure. This review outlines the recent advances in EVs and related infrastructure, mainly from artificial intelligence (AI), which makes EVs a more attractive consumer option. The application of AI in improving EVs, facilitating EV charging stations, and EV integration with the smart grid is critically analyzed and reviewed. Finally, future trends and prospects in the area are discussed.
Accurate battery life prediction is a critical part of the business case for electric vehicles, stationary energy storage, and nascent applications such as electric aircraft. Existing methods are based on relatively small but well-designed lab datasets and controlled test conditions but incorporating field data is crucial to build a complete picture of how cells age in real-world situations. This comes with additional challenges because end-use applications have uncontrolled operating conditions, less accurate sensors, data collection and storage concerns, and infrequent access to validation checks. We explore a range of techniques for estimating lifetime from lab and field data and suggest that combining machine learning approaches with physical models is a promising method, enabling inference of battery life from noisy data, assessment of second-life condition, and extrapolation to future usage conditions. This work highlights the opportunity for insights gained from field data to reduce battery costs and improve designs.
Accurate degradation monitoring over battery life is indispensable for the safe and durable operation of battery-powered applications. In this work, we extend conventional capacity degradation estimation to the estimation of entire constant-current charging curves. A deep neural network (DNN) is developed to estimate complete charging curves by featuring small portions of the charging curves to form the input. We demonstrate that the charging curves can be accurately captured with an error of less than 16.9 mAh for 0.74 Ah batteries with 30 points collected in less than 10 min. Validation based on batteries working at different current rates and temperatures further demonstrates the effectiveness of the proposed method. This method also enjoys the advantage of transfer learning; that is, a DNN trained on one battery dataset can be used to improve the curve estimation of other batteries operating under different scenarios by using few training data.
For accelerating the technology development and facilitating the reliable operation of lithium-ion batteries, accurate prediction for battery cycle life and remaining useful life (RUL) are both critical. However, diverse aging mechanisms, significant device variability and random working conditions have remained challenges. A reasonable description and an effective prediction algorithm are indispensable for achieving accurate prediction results. In this paper, battery terminal voltage, current and temperature curves from several charge cycles and especially their difference between these cycles are first utilized for description of battery cycle life and RUL. Moreover, a hybrid convolutional neural network (CNN), which is based on a fusion of three-dimensional CNN and two-dimensional CNN, is designed for their predictions. The battery charge voltage, current and temperature and their curves are first fused for considering the strong relationships between them. And the features hidden in the curves are extracted and modelled automatically. Furthermore, a feature attention algorithm and a multi-scale cycle attention algorithm are proposed to estimate the relationships between different features and cycles respectively for further heightening the prediction performance. Experiments and comparisons are conducted. The results show that the proposed method is an accurate method for different applications. It achieved 1.1% test error for battery cycle life early prediction of different batteries under different charge policies, and 3.6% for RUL prediction.
A key challenge for lithium (Li)-ion batteries is the capability to manage battery performance and predict lifetime. Early detection of battery-aging phenomena and the implications for the performance are crucial for maintaining warranty and avoiding safety-related liabilities. We established a framework for early detection of loss of Li inventory, which is further separated into Li plating and normal solid electrolyte interphase (SEI) formation. Although SEI formation is inevitable, Li plating causes serious degradation and safety issues. Therefore, Li plating must be identified and avoided. Our framework differentiates Li plating from SEI-formation-dominated cells based on data from the first 25 aging cycles. This classification framework is based on machine learning (ML); multiple coherent and physically meaningful electrochemical signatures along the aging process are used. We also demonstrate that multiple electrochemical signatures must be combined to increase accuracy in the classification.
For the close relationship with safety, surface temperature has been a hot topic in the research of lithium-ion batteries. Studying surface temperature changes can ensure safety and provide information for degradation estimation. However, there are few studies on obtaining degradation information from surface temperature. In this paper, a new health indicator (HI) is proposed to predict the remaining useful life (RUL) of lithium-ion batteries from the discharge surface temperature, which is convenient for real-time measurement and online estimation. First, according to the actual trend of surface temperature, an exponential model is used to fit the surface temperature from 1000 s to the end of the discharge. Then, a new HI is extracted from the model parameters, which indicates the change rate of surface temperature. Pearson and Spearman correlation analysis verified that HI is related to battery capacity degradation. What's more, a comparison of the new HI and other types of HIs is performed. Finally, a method combining HI and relevance vector machine (RVM) is proposed for online RUL prediction. Predictions have been made for different algorithms and starting points, the results show that the new HI is effective for degradation modeling. And the RUL prediction error is less than 5 cycles for 5#, 6# and 7# batteries.
Enabling accurate prediction of battery failure will lead to safer battery systems, as well as accelerating cell design and manufacturing processes for increased consistency and reliability. Data-driven prediction methods have shown promise for accurately predicting cell behaviors with low computational cost, but they are expensive to train. Furthermore, given that the risk of battery failure is already very low, gathering enough relevant data to facilitate data-driven predictions is extremely challenging. Here, a perspective for designing experiments to facilitate a relatively low number of tests, handling the data, applying data-driven methods, and improving our understanding of behavior-dictating physics is outlined. This perspective starts with effective strategies for experimentally replicating rare failure scenarios and thus reducing the number of experiments, and proceeds to describe means to acquire high-quality datasets, apply data-driven prediction techniques, and to extract physical insights into the events that lead to failure by incorporating physics into data-driven approaches.
Machine learning is a specific application of artificial intelligence that allows computers to learn and improve from data and experience via sets of algorithms, without the need for reprogramming. In the field of energy storage, machine learning has recently emerged as a promising modelling approach to determine the state of charge, state of health and remaining useful life of batteries. First, we review the two most studied types of battery models in the literature for battery state prediction: the equivalent circuit and physics-based models. Based on the current limitations of these models, we showcase the promise of various machine learning techniques for fast and accurate battery state prediction. Finally, we highlight the major challenges involved, especially in accurate modelling over length and time, performing in situ calculations and high-throughput data generation. Overall, this work provides insights into real-time, explainable machine learning for battery production, management and optimization in the future.
Lithium-ion batteries have been widely used in many important applications. However, there are still many challenges facing lithium-ion batteries, one of them being degradation. Battery degradation is a complex problem, which involves many electrochemical side reactions in anode, electrolyte, and cathode. Operating conditions affect degradation significantly and therefore the battery lifetime. It is of extreme importance to achieve accurate predictions of the remaining battery lifetime under various operating conditions. This is essential for the battery management system to ensure reliable operation and timely maintenance and is also critical for battery second-life applications. After introducing the degradation mechanisms, this paper provides a timely and comprehensive review of the battery lifetime prognostic technologies with a focus on recent advances in model-based, data-driven, and hybrid approaches. The details, advantages, and limitations of these approaches are presented, analyzed, and compared. Future trends are presented, and key challenges and opportunities are discussed.
18650-type cells with 2.5 Ah capacity are cycled at both 25 °C and 0 °C separately, and at 25 °C two charging protocols (constant current, and constant current-constant voltage charge) are used. Differential voltage analysis (dV/dQ) and alternating current (AC) impedance are mainly used to investigate battery degradation mechanisms quantitatively. The dV/dQ suggests that active cathode loss and loss of lithium inventory (LLI) are the dominating degradation factors. Significant microcracks are observed in the fatigued cathode particles from the scanning electron microscopy (SEM) images. Crystal structure parameters of selected fatigued batteries at fully charged state are determined by in situ high-resolution neutron powder diffraction. Obvious increases of ohmic resistance and solid electrolyte interphase (SEI) resistance occur when the battery capacity fade falls beneath 20%. Continuous charge transfer resistance and Warburg impedance coefficient (W.eff) increase are observed in the course of cycling. Correlation analysis is performed to bridge the gap between material loss as well as LLI and impedance increase. The increase of the charge transfer resistance is related to both active cathode loss and LLI, and a functional relationship is revealed between LLI and W.eff regardless of the used cycling protocols.
The state of health (SOH) and remaining useful lifetime (RUL) estimation are important parameters for battery health forecasting as they reflect the health condition of battery and provide a basis for battery replacement. This study proposes a novel on-line synthesis method based on the fusion of partial incremental capacity and artificial neural network (ANN) to estimate SOH and RUL under constant current discharge. Firstly, the advanced filter methods are applied to smooth the initial incremental capacity curves. Then the strong correlation feature values are extracted from the partial incremental curves by using correlation analysis methods. Finally, two ANN models aiming at estimating SOH and RUL are established to estimate the SOH and RUL simultaneously. The training and verification results indicate that the proposed method has highly reliability and accuracy for SOH and RUL estimation.
Rapid growth in the market for electric vehicles is imperative, to meet global targets for reducing greenhouse gas emissions, to improve air quality in urban centres and to meet the needs of consumers, with whom electric vehicles are increasingly popular. However, growing numbers of electric vehicles present a serious waste-management challenge for recyclers at end-of-life. Nevertheless, spent batteries may also present an opportunity as manufacturers require access to strategic elements and critical materials for key components in electric-vehicle manufacture: recycled lithium-ion batteries from electric vehicles could provide a valuable secondary source of materials. Here we outline and evaluate the current range of approaches to electric-vehicle lithium-ion battery recycling and re-use, and highlight areas for future progress. Processes for dismantling and recycling lithium-ion battery packs from scrap electric vehicles are outlined.
A future electric transportation market will depend on battery innovation
Lithium-ion battery is widely used in various industrial applications including electric vehicles (EVs) and distributed grids due to its high energy density and long service life. As an essential performance indicator, state of charge (SOC) reflects the residual capacity of a battery. To ensure the safe operation of systems, it is vital to obtain battery SOC accurately. However, as a parameter which cannot be directly measured, the battery SOC are influenced not only by the measurement noise but also the cell temperature. Focusing on these challenging issues, this paper proposes a hybrid model to estimate the lithium-ion battery SOC under dynamic conditions. This method consists of deep belief network (DBN) and the Kalman filter (KF). The battery electric current, terminal voltage and temperature are used as the input of the proposed model of which output is the SOC. With the powerful nonlinear fitting capability of the DBN, the model can extract relationship between the measurable parameters and battery SOC. The KF algorithm is utilized to eliminate the effects from measurement noise and improve the estimation accuracy. Experiments under different operation conditions are carried out with commercial lithium-ion batteries. The biggest average estimation error is less than 2.2% which indicates that the proposed method is promising for battery SOC estimation especially for the complex operation conditions.
The lithium ion battery is widely used in electric vehicles (EV). The battery degradation is the key scientific problem in battery research. The battery aging limits its energy storage and power output capability, as well as the performance of the EV including the cost and life span. Therefore, a comprehensive review on the key issues of the battery degradation among the whole life cycle is provided in this paper. Firstly, the battery internal aging mechanisms are reviewed considering different anode and cathode materials for better understanding the battery fade characteristic. Then, to get better life performance, the influence factors affecting battery life are discussed in detail from the perspectives of design, production and application. Finally, considering the difference between the cell and system, the battery system degradation mechanism is discussed.
Lithium-ion battery is a critical part in various industrial applications. In practice, the performance of such batteries degrades over time. To maintain the battery performance and ensure their reliability, it is important to implement on-line life cycle health state assessment in a battery management system. However, two big challenges in on-line battery actual capacity estimation must be overcome. The first one is the on-line extraction of measurable degradation features. The other one is the on-line mapping from the degradation feature space to the battery capacity space. This paper proposes a self-adaptive life-cycle health state assessment method based on the on-line measurable parameters of lithium-ion battery. Ten different degradation features are extracted from the voltage, electric current and critical time during operation. These degradation features are fused to achieve a higher adaptability to complex operating conditions. The lithium-ion battery health state is assessed with a mapping model that links the feature space to the capacity space. The model is trained by the least squares support vector machine method for less computational complexity. The experimental results based on the real battery testing data show that the correlation between the degradation feature and the battery capacity is higher than 0.7 and the mean error of capacity estimation is less than 0.05. For the dynamic operation conditions, the mean error of capacity estimation is less than 11 mAh. This study illustrates the adaptability and applicability of the proposed on-line life-cycle health state assessment approach in various electric vehicle applications.
This paper presents a novel lithium-ion cell model, which simulates the current voltage characteristic as a function of state of charge (0%–100%) and temperature (0–30 °C). It predicts the cell voltage at each operating point by calculating the total overvoltage from the individual contributions of (i) the ohmic loss η0, (ii) the charge transfer loss of the cathode ηCT,C, (iii) the charge transfer loss and the solid electrolyte interface loss of the anode ηSEI/CT,A, and (iv) the solid state and electrolyte diffusion loss ηDiff,A/C/E. This approach is based on a physically meaningful equivalent circuit model, which is parametrized by electrochemical impedance spectroscopy and time domain measurements, covering a wide frequency range from MHz to μHz. The model is exemplarily parametrized to a commercial, high-power 350 mAh graphite/LiNiCoAlO2-LiCoO2 pouch cell and validated by continuous discharge and charge curves at varying temperature. For the first time, the physical background of the model allows the operator to draw conclusions about the performance-limiting factor at various operating conditions. Not only can the model help to choose application-optimized cell characteristics, but it can also support the battery management system when taking corrective actions during operation.
State of Health (SOH) estimation of lithium ion batteries is critical for Battery Management Systems (BMSs) in Electric Vehicles (EVs). Many estimation techniques utilize a battery model; however, the model must have high accuracy and high computational efficiency. Conventional electrochemical full-order models can accurately capture battery states, but they are too complex and computationally expensive to be used in a BMS. A Single Particle (SP) model is a good alternative that addresses this issue; however, existing SP models do not consider degradation physics. In this work, an SP-based degradation model is developed by including Solid Electrolyte Interface (SEI) layer formation, coupled with crack propagation due to the stress generated by the volume expansion of the particles in the active materials. A model of lithium ion loss from SEI layer formation is integrated with an advanced SP model that includes electrolytic physics. This low-order model quickly predicts capacity fade and voltage profile changes as a function of cycle number and temperature with high accuracy, allowing for the use of online estimation techniques. Lithium ion loss due to SEI layer formation, increase in battery resistance, and changes in the electrodes' open circuit potential operating windows are examined to account for capacity fade and power loss. In addition to the low-order implementation to facilitate on-line estimation, the model proposed in this paper provides quantitative information regarding SEI layer formation and crack propagation, as well as the resulting battery capacity fade and power dissipation, which are essential for SOH estimation in a BMS.
A physics-based Li-ion battery (LIB) aging model accounting for both lithium plating and solid electrolyte interphase (SEI) growth is presented, and is applied to study the aging behavior of a cell undergoing prolonged cycling at moderate operating conditions. Cell aging is found to be linear in the early stage of cycling but highly nonlinear in the end with rapid capacity drop and resistance rise. The linear aging stage is found to be dominated by SEI growth, while the transition from linear to nonlinear aging is attributed to the sharp rise of lithium plating rate. Lithium plating starts to occur in a narrow portion of the anode near the separator after a certain number of cycles. The onset of lithium plating is attributed to the drop of anode porosity associated with SEI growth, which aggravates the local electrolyte potential gradient in the anode. The presence of lithium metal accelerates the porosity reduction, further promoting lithium plating. This positive feedback leads to exponential increase of lithium plating rate in the late stage of cycling, as well as local pore clogging near the anode/separator interface which in turn leads to a sharp resistance rise.
The charging time-consuming and lifespan of lithium-ion batteries have always been the bottleneck for the tremendous application of electric vehicles. In this paper, cycle life tests are conducted to reveal the influence of different charging current rates and cut-off voltages on the aging mechanism of batteries. The long-term effects of charging current rates and cut-off voltages on capacity degradation and resistance increase are compared. The results show that there exists a critical charging current and a critical charging cut-off voltage. When the charging stress exceeds the critical value, battery degradation speed will be greatly accelerated. Furthermore, battery aging mechanisms at various charging currents and cut-off voltages are investigated using incremental capacity analysis. It is indicated that charging current and cut-off voltage should be reduced to retard battery degradation when the battery degrades to a certain extent. The time when the loss of electrode material accelerates is taken as the crisis to reduce charging current and the time when the loss of lithium inventory accelerates is taken as the crisis to reduce charging cut-off voltage. Moreover, an experiential model quantitatively describing the relationship between capacity degradation rate and charging stresses at different aging states is established.
The rate and shape of the charging current indubitably affect the charging time and the ageing rate of a battery. Depending on the application requirements, it is possible to use high-charging current in order to decrease the charging time. However, the influence of fast-charging current profiles should be investigated to identify their impact on battery functionality over time. In this article, static and dynamic fast-charging current profiles are applied to a high power 7 Ah LiFePO4-based cells (LFP), and the results of cycle-life and characterization tests are discussed. To select the proper fast-charging profile, the evaluation relies on some factors: discharge capacity retention, charging capacity, charging time, and cell temperature. After 1700 cycles, the results revealed that the dynamic fast-charging current profile has a prominent role in decreasing the degradation rate as well as the charging time of cells compared with the static fast-charging profile.
Over the last decade, many efforts have been deployed to develop models for the prediction, the control, the optimization and the parameter estimation of Lithium-ion (Li-ion) batteries. It appears that the most successful electrochemical-based model for Li-ion battery is the Pseudo-two-Dimensional model (P2D). Due to the fact that the governing equations are complex, this model cannot be used in real-time applications like Battery Management Systems (BMSs). To remedy the situation, several investigations have been carried out to simplify the P2D model. Mathematical and physical techniques are employed to reduce the order of magnitude of the P2D governing equations. The present paper is a review of the studies on the modeling of Li-ion batteries with simplified P2D models. The assumptions on which these models rest are stated, the calculation methods are examined, the advantages and the drawbacks of the models are discussed and their applications are presented. Suggestions for overcoming the shortcomings of the models are made. Challenges and future directions in the modeling of Li-ion batteries are also discussed.
Lithium-ion Battery Energy Storage Systems (BESS) are to be the next household electrical appliance in a smart grid environment. This is beside the growth of electrical vehicles with lithium-ion batteries. However, these batteries are yet to prove their performances in the household sector. This paper presents the performances of a small household scale battery energy storage system with a lithium-ion battery pack and a single-phase ac-dc inverter. Results of a list of tests conducted in a lab environment are presented explaining the test procedures and results. Test results show that the considered BESS is suitable for daily cycling applications but has low efficiencies at the rated power operations. Converter losses contribute to the low efficiency of the considered BESS. The BESS is more efficient in lower power (lower than rated) operations. However, the effects of the current harmonics need to be considered when it is operated with other appliances.
An in-depth historical and current review is presented on the science of lithium-ion battery (LIB) solid electrolyte interphase (SEI) formation on the graphite anode, including structure, morphology, composition, electrochemistry, and formation mechanism. During initial LIB operation, the SEI layer forms on the graphite surfaces, the most common anode material. The SEI is essential to the long-term performance of LIBs, and it also has an impact on its initial capacity loss, self-discharge characteristics, rate capability, and safety. While the presence of the anode SEI is vital, it is difficult to control its formation and growth, as they depend on several factors. These factors include the type of graphite, electrolyte composition, electrochemical conditions, and temperature. Thus, SEI formation and electrochemical stability over long-term operation should be a primary topic of future investigation in the LIB development. This article covers the progression of knowledge regarding the SEI, from its discovery in 1979 to the current state of understanding, and covers differences in the chemical and structural makeup when cell materials and components are varied. It also discusses the relationship of the SEI layer to the LIB formation step, involving both electrolyte wetting and subsequent slow charge-discharge cycles to grow the SEI.
Energy density is the main property of rechargeable batteries that has driven the entire technology forward in past decades. Lithium-ion batteries (LIBs) now surpass other, previously competitive battery types (for example, lead–acid and nickel metal hydride) but still require extensive further improvement to, in particular, extend the operation hours of mobile IT devices and the driving mileages of all-electric vehicles. In this Review, we present a critical overview of a wide range of post-LIB materials and systems that could have a pivotal role in meeting such demands. We divide battery systems into two categories: near-term and long-term technologies. To provide a realistic and balanced perspective, we describe the operating principles and remaining issues of each post-LIB technology, and also evaluate these materials under commercial cell configurations.