Machine Learning Techniques for Battery State of Health Prediction: A Comparative Review

Estimating battery state of health using electrochemical impedance spectroscopy and the relaxation effect

With the wide application of lithium-ion battery in various fields, the State of Health (SOH) estimation has become a research hotspot for advanced battery management system (BMS). Accurate SOH estimation is helpful to ensure the safe operation of equipment or system in practical applications. Among various lithium-ion battery health diagnosis methods, particle filter and its variants are the mainstream with the significant advantages in non-linear and non-Gaussian system modeling. But in practical applications, the BMS always suffers from the limited power supplication and finite computing resources. Therefore, this paper implemented a comparative study on particle filter (PF) and typical variants, including extended Kalman particle filter (EPF), unscented particle filter (UPF), regularized particle filter (RPF). Through the NASA’s battery degradation model, the performance of the above particle filter algorithms is compared and analyzed. The experimental results show that UPF has the highest estimation accuracy, and it is more suitable for the situation with higher prediction accuracy requirements. PF has the least time consumption and is more suitable for on-line health assessment.

To prolong battery lifetime and maximize performance, battery management systems (BMSs) need to estimate internal variables that are related to battery aging. For that, accurate battery models that can represent internal variables that are not directly measurable are required. However, most present BMSs use equivalent-circuit models (ECMs or EQMs) for state of charge (SOC) and state of health (SOH) estimation. These models are not able to predict these aging-related variables, and therefore, they cannot be used to limit battery degradation. In this paper, we propose a method for electrode-level SOC (eSOC) and electrode-level SOH (eSOH) estimation using an electrode-level ECM (eECM), which can describe both electrodes’ potentials, giving valuable information for aging prevention. The method produces state of lithiation (SOL) and potential estimates of both electrodes, and updates the eSOH parameters to maintain estimation accuracy through the lifetime of the battery. Furthermore, the eSOH parameter estimates are used to obtain degradation mode information, which could be used to improve state estimation, health diagnosis and prognosis.

The state of health (SOH) is a significant index for the safe and reliable performance of lithium-ion batteries (LIBs). Although electrochemical impedance spectroscopy (EIS) is a promising nondestructive technique for SOH estimation, existing investigations rarely apply EIS to establish SOH estimation mathematical models. Considering the SOH estimation model based on EIS is imposed by the nonlinear and coupled constraints including temperature and state of charge (SOC), in this paper, the charge transfer resistance is extracted first as the aging feature through EIS analysis under different conditions of SOH, SOC, and temperature. Then, a multi-state SOH model is developed based on the charge transfer resistance. The model quantitatively elucidates the interdependencies among charge transfer resistance, SOH, SOC, and temperature. A two-stage SOH estimation strategy is subsequently proposed based on the proposed model to address the inferior of accuracy from the traditional probabilistic model for SOH estimation due to inaccurate model parameters. Specifically, in the first stage, the SOH is initialized. In the second stage, SOH is re-estimated based on optimal point parameters selected by the SOH obtained in the first stage. Finally, the proposed estimation strategy for SOH is verified through experiments, where the estimation maximum RMSE is less than 0.04. Compared to the previous methods, the proposed strategy demonstrates more accuracy in SOH estimation.

Accelerated state of health estimation of second life lithium-ion batteries via electrochemical impedance spectroscopy tests and machine learning techniques

The state-of-energy (SOE) and state-of-health (SOH) are two crucial quotas in the battery management systems, whose accurate estimation is facing challenges by electric vehicles’ (EVs) complexity and changeable external environment. Although the machine learning algorithm can significantly improve the accuracy of battery estimation, it cannot be performed on the vehicle control unit as it requires a large amount of data and computing power. This paper proposes a joint SOE and SOH prediction algorithm, which combines long short-term memory (LSTM), Bi-directional LSTM (Bi-LSTM), and convolutional neural networks (CNNs) for EVs based on vehicle-cloud collaboration. Firstly, the indicator of battery performance degradation is extracted for SOH prediction according to the historical data; the Bayesian optimization approach is applied to the SOH prediction combined with Bi-LSTM. Then, the CNN-LSTM is implemented to provide direct and nonlinear mapping models for SOE. These direct mapping models avoid parameter identification and updating, which are applicable in cases with complex operating conditions. Finally, the SOH correction in SOE estimation achieves the joint estimation with different time scales. With the validation of the National Aeronautics and Space Administration battery data set, as well as the established battery platform, the error of the proposed method is kept within 3%. The proposed vehicle-cloud approach performs high-precision joint estimation of battery SOE and SOH. It can not only use the battery historical data of the cloud platform to predict the SOH but also correct the SOE according to the predicted value of the SOH. The feasibility of vehicle-cloud collaboration is promising in future battery management systems.

The state of health (SOH) estimation of lithium-ion batteries is essential for the safe operation and intelligent maintenance of the battery energy storage system in electric vehicles. Up to now, researches have been done a lot on the SOH estimation of the batteries under standard charging rate, but seldom touched the SOH estimation for fast-charging scenario. This paper proposes a convolutional neural network (CNN) based SOH estimation method, which can extract health features from only 3-minute partial fast-charging segment and then accurately estimate the SOH of the batteries under fast-charging. The proposed method is validated by an open-source dataset, where the root mean square percentage error and mean absolute percentage error are less than 1% and the maximum percentage error is less than 2%. The proposed method may have a great potential in real application for the advantage of high accuracy and small amount of charging data needed.

<p>Accurate water level (WL) forecasting is important for water resources management and planning purposes in the Great Lakes. The objectives of this research are two-fold.  The first objective is to apply machine learning (ML) (i.e., random forest (RF) and support vector regression (SVR)) and hybrid convolutional neural network(CNN)-long-short term memory (LSTM) deep learning (DL) models for multi-step (i.e., one-, two- and three-monthly step ahead) WL forecasting in the Great Lakes (Michigan and Ontario). The second objective is to integrate the boundary corrected (BC) maximal overlap discrete wavelet transform (MODWT) with SVR, RF, and CNN-LSTM models to improve the performance of the individual models. By employing a BC-wavelet decomposition method, the ‘future data’ issue (i.e., data from the future that is not available), often overlooked in the literature and a major barrier to achieving realistic forecasting performance is overcome. </p><p>For Lakes Michigan and Ontario, 1212 monthly WL (m) records (spanning Jan 1918–Dec 2018) were used to develop the models. For the non-wavelet-based models (SVR, RF, and CNN-LSTM), candidate model inputs included the WL recorded over the previous 12 months.  For the BC-MODWT-based models (BC-MODWT-SVR, BC-MODWT-RF, and BC-MODWT-CNN-LSTM), the lagged input time series were decomposed into BC-wavelet and scaling coefficients by using different mother wavelets (Haar, Daubechies, Symlets, Fejer-Korovkin and Coiflets), filter lengths (from two up to 12) and decomposition levels (from one up to seven).  For each method (SVR, RF, and CNN-LSTM), mother wavelet, and decomposition level a model was generated.  For both wavelet- and non-wavelet-based models, the particle swarm optimization (PSO) method was used to select the most appropriate inputs to include in the proposed multi-step WL forecasting models.</p><p>The datasets were partitioned into calibration and validation subsets. After calibrating the models, various performance evaluation metrics, e.g., coefficient of determination (R<sup>2</sup>), root mean square error (RMSE), mean absolute error (MAE), root mean square percentage error (RMSPE), mean absolute percentage error (MAPE) and the Nash-Sutcliffe efficiency coefficient (NSC) were used to assess model accuracy.</p><p>Of the ML models, the SVR outperformed RF while the DL models outperformed the ML models for each forecast lead time (one-, two-, and three-step(s) ahead). Results from this case study indicate that not all wavelet families and decomposition levels perform equally and in some cases, the wavelet-based models do not improve performance over the non-wavelet-based models. However, the BC-MODWT-CNN-LSTM using suitable mother wavelets (e.g., Haar) outperforms the individual ML and BC-MODWT-ML-based models. More accurate forecasts were obtained for Lake Michigan although the performance in both Great Lakes was accurate. The outcomes of this research indicate that the BC-MODWT-CNN-LSTM model is a promising tool for generating accurate WL forecasts.</p>

With the wide application of lithium batteries (LIBs) in electrified transportation and smart grids, especially in the pure electric vehicle industry, the accurate health maintenance monitoring of LIBs has emerged as critical to safe battery operation. Although many data-driven methods with state of health (SOH) estimation for LIBs have been proposed, the problems of industrial application and computational cost still need to be improved further. In contrast, this article carried out a low-complexity SOH estimation method for LIBs. Specifically, the seven health indicators are extracted firstly to characterize battery health status from voltage, current, temperature, and other data that can be obtained online. Then, the optimized Gaussian process regression (GPR) algorithm is proposed with proper computational cost. Ultimately, by combining a multi-indirect features extraction and optimized GPR algorithm, the online SOH estimation for LIBs was established and verified with NASA experiment data. The experimental results show that the maximum MAPE of SOH estimation from the proposed method is 1.4496 and the minimum MAPE only reaches 0.5635. More importantly, the optimized GPR for SOH estimation can achieve a maximum 65.37% improvement under multiple evaluation criteria compared to traditional GPR. The method proposed in this article is helpful for realizing online SOH estimation in battery management systems.

Joint estimation of State of Charge (SOC) and State of Health (SOH) for lithium ion batteries using Support Vector Machine (SVM), Convolutional Neural Network (CNN) and Long Sort Term Memory Network (LSTM) models

This paper considers the issue of Li-Ion batteries State of Health (SoH) and State of Charge (SoC) accurate and robust estimation for electric vehicle applications. SoC and SoH are two monitoring indicators of primary importance that are used by the Battery Management System (BMS) , amongst other benefits, to manage and equalize the battery cells. Improving the estimation precision and reliability of the SoC and the SoH indicators is highly beneficial during operation and maintenance of the vehicle. We propose in this paper a new scheme of SoC and SoH estimation using an hybridization of Kalman filtering, Recursive Least Squares approach and Support Vector Machines learning. The battery SoC and SoH indicators are estimated using an adaptive-Sigma Point Kalman Filter. The battery cell impedance equivalent filter is obtained in real-time by a Recursive Least Square. Furthermore, the cell capacity evolution tracking is achieved by using a Support Vector Machine (SVM) method. Finally, the battery cell capacity and impedance equivalent filter are provided to the SoC estimator in order to update its state and observation models. This architecture yields to a complete SoC and SoH algorithmic solution exhibiting a high level of accuracy and robustness. The SVM method which requires the highest computational load in the architecture is designed to be used only for estimating the variable with the lowest evolution dynamics.

The state of health (SOH) of lithium-ion batteries (LIBs) needs to be accurately estimated to ensure the safety and stability of electric vehicles (EVs) while in operation. In this paper, we proposed a SOH estimation method based on Improved Aquila Optimizer (IAO) and Support Vector Regression (SVR) to achieve an accurate estimation of SOH. During the charging and discharging phases of the battery, we analyzed the trends in current, voltage, and energy, then extracted four features. We used the Kendall coefficient and gray relational grade to prove that features and SOH were highly correlated. On the other hand, IAO was used to optimize the penalty factor and kernel function parameters of the SVR to further improve the generalization and mapping ability. The proposed method was verified under different operating conditions using the CACLE battery data set; the results show that high accuracy can be achieved in SOH estimation via IAO–SVR, and the estimation error of mean MAE is remaining within 2%.

Lithium-ion batteries health prognosis via differential thermal capacity with simulated annealing and support vector regression

ABSTRACTAccurate State of Health (SOH) estimation is critical for battery management systems (BMS) in electric vehicles (EVs). However, the absence of a universal aging model for power batteries presents significant challenges. This study leverages the open‐source battery cell data set from the University of Maryland and focuses on private battery packs to address the aging model SOH estimation. Two aging features indicative of capacity degradation are extracted from constant current charging data using incremental capacity analysis (ICA). To handle nonlinearity and feature coupling, a flexible data‐driven aging model is proposed, employing dual Gaussian process regressions (GPRs) and transfer learning to enhance model efficiency and accuracy. Adaptive filtering via the Particle filter (PF) further refines the model by integrating aging features and output capacity, resulting in a closed‐loop data fusion approach for precise SOH estimation. Battery pack aging experiments validate the proposed method, demonstrating that transfer learning effectively improves estimation accuracy. The proposed method achieves closed‐loop SOH estimation with a mean root mean square error (RMSE) of 0.87, underscoring its reliability and precision.

This paper proposes an innovative state of health (SoH) estimation of the battery pack for the reused Li-Ion cells. Due to the elevated significance and interest of the overall system performance, reliability, safety and long lifetime operation, it is necessary an accurate estimation of the lithiumion batteries health status. In case of a battery pack, the degradation of each cell, will affect and result in the increase of the deterioration process of conjoint cell, and furthermore to the total degradation of the battery pack and malfunction of the normal operation. Therefore, to resolve such issue, a novel method is developed in this article, based on multi series-parallel ANN structure which will concludes with accurate and fast SoH estimation of the battery with used Li-Ion cells. Its accuracy and calculation time are improved and accelerated using an online modeling approach (during operation) with optimal generalization. By the voltage and current as the main inputs and with the help of secondary information as temperature and cycle usage, it can be possible to describe the actual quantity of energy, which is a key factor in applications. The structure of the ANN is based on multi-series parallel configuration algorithm, combining the long-short term memory (LSTM) NN features with the Convolutional neural network (CNN) abilities which enhance the ability of stable and fast estimation during all the testing period of the pack, by further reducing the computational power and increasing the generalization of the proposed paper. In this paper, the SoH estimation time is reduced to 46%, peak error and average error are reduced by 36.3% and 23.4%, comparing with the previous work of the authors. Also, it is confirmed on this study that the proposed method can be applicable not only to a battery pack, but also to single cell or multi modules, increasing its range of applications. In simple words, this study proposes an alternative and cost-effective SoH estimation and diagnosis approach for the deteriorated battery, comparing to high-cost industrial devices, focused on the low input data scenario, taking into account the inter-degradation between cells.