Deep Learning-Based SoC Estimation of Lithium-Ion Batteries in EVs Using 2D CNN

An optimized long short-term memory-weighted fading extended Kalman filtering model with wide temperature adaptation for the state of charge estimation of lithium-ion batteries

Dynamic ultrasonic response modeling and accurate state of charge estimation for lithium ion batteries under various load profiles and temperatures

Highly accurate state of charge (SOC) estimation of lithium-ion batteries is one of the key technologies of battery management systems in electric vehicles. The performance of SOC estimation directly influences the driving range and safety of these vehicles. Due to external disturbances, temperature variation and electromagnetic interference, accurate SOC estimation becomes difficult. To accurately estimate the SOC of lithium-ion batteries, this article presents a novel machine-learning method to address the risk of gradient explosion and gradient decent using the dynamic nonlinear auto-regressive models with exogenous input neural network (NARX) with long short-term memories (LSTM).The proposed hybrid NARX model embeds LSTM memory, which provides jump-ahead connections in the time-unfolded model. These jump-ahead connections provide a shorter path for the propagation of gradient information, therefore reducing long-term dependence on the recurrent neural network. Experimental results show that the estimation performance root mean square error (RMSE) of the proposed model is less than 1%, and this model has better multitime prediction performance. Finally, the hybrid NARX and LSTM model is compared with the standard back propagation neural network based on particle swarm optimization (BPNN-PSO), the least-squares support vector machine (LS-SVM) and LSTM existing models under urban dynamometer driving schedule (UDDS) and dynamic stress test (DST) conditions. The proposed hybrid NARX-LSTM model yield relative to other methods and can estimate the battery SOC with high accuracy. The RMSE of proposed model is improved by approximately 60% compared with the standard LSTM.

In order to improve the accuracy of the battery state of charge (SOC) estimation, in this paper we take a lithium-ion battery as an example to study the adaptive Kalman filter based SOC estimation algorithm. Firstly, the second-order battery system model is introduced. Meanwhile, the temperature and charge rate are introduced into the model. Then, the temperature and the charge rate are adopted to estimate the battery SOC, with the help of the parameters of an adaptive Kalman filter based estimation algorithm model. Afterwards, it is verified by the numerical simulation that in the ideal case, the accuracy of SOC estimation can be enhanced by adding two elements, namely, the temperature and charge rate. Finally, the actual road conditions are simulated with ADVISOR, and the simulation results show that the proposed method improves the accuracy of battery SOC estimation under actual road conditions. Thus, its application scope in engineering is greatly expanded.

Battery is one of the main parts in Electrical vehicles as it serves as energy source. In order to prolong battery lifetime and optimally operate the battery, battery performance needs to be monitored and controlled using Battery Management System (BMS). One among important BMS functions is to determine the amount of remaining energy capacity available in a battery, represented by State of Charge (SOC), through estimation. The SOC estimation accuracy is important since it is used as indicator to prevent battery from failures due to overcharging and over discharging. One of the SOC estimation methods is Coulomb counting (CC) which is easy to implement. However, a false initial value of the SOC in this method will degrade the accuracy of the SOC estimation. In this paper an SOC estimation method is proposed by using CC method combined with recursive least square (RLS) algorithm and Proportional Integral (PI) controller to correct the error of SOC estimation due to false initial value of SOC. The result indicates that the proposed method is able to correct the error of SOC estimation due to false initial value of SOC with average of SOC estimation error equals 0.498%.

An accurate state-of-charge estimation of lithium-ion batteries based on improved particle swarm optimization-adaptive square root cubature kalman filter

Accurate State of Charge (SOC) estimation is vital for ensuring the efficiency and reliability of lithium-ion battery systems. Accurately estimating SOC under complex environmental conditions remains a significant challenge. Electric forklifts offer advantages such as low operating costs and convenient charging, presenting broad application prospects. However, during year-round operation, they face issues like large temperature variations and high complexity in battery data acquisition and processing. This paper first analyzes the operational characteristics of forklift batteries. Pearson correlation coefficient analysis is employed for feature extraction from the data. These features are directly derived from the actual charge-discharge operational data of the batteries, forming the core inputs for subsequent SOC estimation. To enhance SOC estimation accuracy, a temperature correction method is applied to adjust the maximum available capacity at the current temperature. Subsequently, a neural network model combining Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Attention mechanism is utilized. Different optimization algorithms are applied to tune the model's hyperparameters. Validation is performed by comparing various deep learning frameworks and optimization methods across different datasets.Results demonstrate that the SOC estimation accuracy based on the CNN-BiLSTM-Attention neural network algorithm is high. After incorporating the temperature correction method and the Grey Wolf Optimizer (GWO), the Root Mean Square Error (RMSE) is reduced to 1.1%, and the overall relative error is decreased by 52%.

The electrical equivalent circuit model (ECM) is widely employed for state of charge (SOC) estimation in lithium-ion batteries. Among ECM-based approaches, the Thevenin equivalent circuit model (TECM) is particularly favored due to its computational efficiency and ease of parameter identification. TECM can be implemented in various configurations, with 1RC, 2RC, and 3RC structures being the most common. In these configurations, each RC unit consists of a resistor (R) and a capacitor (C) connected in parallel and incorporated into the circuit branch in series. As the number of RC branches increases, the computational burden of SOC estimation also rises. However, the improvement in estimation accuracy does not scale proportionally with the increased model complexity. A critical factor influencing the accuracy of SOC estimation is the precise determination of ECM parameters. A widely accepted principle suggests that when a parameter can be directly computed from the available data, it is preferable to use the calculated value rather than an estimated one. In line with this principle, this study directly calculates the internal resistance parameter (R0), which is connected in series with the RC branches in TECM, using test data while estimating the remaining RC parameters. SOC estimation is conducted for 1RC, 2RC, and 3RC configurations, and results are compared with those obtained from ECMs where all parameters, including R0, are estimated. To ensure a rigorous comparison, all estimations are performed using the nonlinear least squares method (LSM). The study employs test data from the US06 and UDDS driving cycles, and performance evaluation is conducted based on error distributions (box plots), root mean square error (RMSE), mean absolute error (MAE), and computational cost. The results showed that there is no significant difference between the error values of the predictions made using the model in which R0 is directly calculated and those made using the model in which R0 is estimated. Specifically, for the 1RC model structure, the lowest MAE values are 6.2 and 6.1 millivolts, respectively. However, these values indicate that the nonlinear LSM can be effectively used for estimating the parameters of the battery electrical circuit model.

Lithium-ion batteries have been extensively utilized as a high-power, rechargeable, and dischargeable energy storage medium. Accurate estimation of the battery state of charge (SOC) in the battery management system (BMS) is imperative for ensuring the safe and stable operation of electric vehicles. This paper proposes an SOC estimation method based on the equivalent circuit model as well as the ampere-time integration method with a physical informed neural network. The network enhances the estimation of SOC by introducing two mechanistic information sources: the equivalent circuit model (ECM) and the ampere-time integration method (Ah-I method). These are utilized as a priori knowledge to constrain the estimation of SOC. Initially, the Rint model is selected as the physical analysis model of the lithium-ion battery, and subsequently, the Ah-I method is chosen as the auxiliary model for SOC output estimation. A deep learning network is then employed to establish the mapping between the battery input parameters and the SOC output. Finally, the SOC is estimated by fusing the physical model and the data-driven model. The results demonstrate the efficacy of the method in accurately estimating the state of charge of lithium batteries, with a root mean square error within 1%. The validity of the research methodology was further validated through comparison with other approaches.

The SOC (State Of Charge) estimation is a core aspect for the associated BMS (Battery Management System) equipment of the lithium battery pack, in which the KF (Kalman Filter) -based methods have been extensively used but suffers from the linearization accuracy drawbacks. A novel UKF (Unscented Kalman Filter) estimation method battery model is proposed for the SOC estimation of the lithium battery pack, in which the linearization treatment is not required and fewer Sigma data points are used, reducing the computational requirement of the SOC estimation. The UKF method improves the SOC covariance properties for the lithium battery pack, the estimation performance of which has been validated by the experimental results. The proposed SOC estimation method has a RMSE (Root Mean Square Error) value of 1.42%, playing an important role in the popularization and application of the lithium battery pack.

Accurate estimation of State of Charge (SOC) and State of Health (SOH) in lithium-ion batteries is critical for effective Battery Management Systems (BMS). A key factor in enhancing estimation accuracy is the appropriate selection of the Open-Circuit Voltage (OCV-SOC) model. This study investigates two prevalent methods for OCV modeling: the Low-Current (LC) method and the Incremental Current (IC) method, and evaluates their impact on the performance of the Unscented Kalman Filter (UKF) for SOC estimation and instantaneous internal resistance (R0) tracking. In this work, SOH is primarily assessed through R0​, which is adopted as the main degradation indicator under dynamic operating conditions. The dynamic model employed is an Equivalent Circuit Model (ECM) of a lithium-ion battery, with estimations conducted using real-world driving data (Federal Urban Driving Schedule, FUDS). The results demonstrate that the IC method, due to its higher local resolution in the OCV modeling, achieves faster convergence and higher estimation accuracy for SOC, particularly when the initial SOC is subject to error. In contrast, the LC method, despite its experimental simplicity, exhibits significant errors in the early stages of simulation due to excessive smoothing of the OCV curve. Furthermore, a comparison of R0 trends reveals that the IC method provides a more consistent and stable estimation, whereas the LC method shows greater short-term fluctuations and reduced stability. This difference is attributed to the higher accuracy of the OCV curve in the IC method. Ultimately, the choice of the OCV extraction method directly influences the accuracy and stability of SOC estimation and SOH analysis based on R0 tracking.

A well-known Luenburger observer and two sliding mode observers (SMO) for battery state of charge (SOC) estimation based on an improved battery equivalent circuit are presented. The comparison of their SOC estimation results is discussed. The testing data under different discharge current profiles generated by DUALFOIL battery simulation program are used to extract the circuit model parameters and verify the effectiveness of the observers for the SOC estimation. It shows that two sliding mode based SOC observers have robust tracking performance and provide more accurate SOC estimation in electric vehicle driving conditions.

In an electric vehicle (EV), the battery management system (BMS) is crucial for managing the health and safety of the battery. The accurate estimation of battery state of charge (SOC) offers critical information about the battery’s remaining capacity. The SOC of the battery mainly depends on its non-linear internal parameters, battery chemistry, ambient temperature, aging factor etc. So, accurate SOC estimation is still a significant challenge. Many researchers have developed several model-based methods that are more complex to develop. Another approach is a data-driven based SOC estimation algorithm, which is less complex but requires more data and it may be inaccurate. In this context, this paper presents a robust and accurate SOC estimation algorithm for a Lithium-ion battery using a deep learning feed-forward neural network (DLFFNN) approach. The proposed algorithm accurately characterizes the battery’s non-linear behavior. To develop a robust SOC estimation algorithm, data is collected at different temperatures with 5% error in data (4 mV-voltage, 110 mA-current, 5∘C temperature) is added to battery datasets. The obtained results demonstrated that the performance of the proposed DLFNN is robust and accurate on different drive cycles with 1.14% Root mean squared error (RMSE), 0.66% mean absolute error (MAE), and 6.65% maximum error (MAX).

To accurately estimate the state of charge (SOC) of lithium-ion power batteries in the event of errors in the battery model or unknown external noise, an SOC estimation method based on the H-infinity filter (HIF) algorithm is proposed in this paper. Firstly, a fractional-order battery model based on a dual polarization equivalent circuit model is established. Then, the parameters of the fractional-order battery model are identified by the hybrid particle swarm optimization (HPSO) algorithm, based on a genetic crossover factor. Finally, the accuracy of the SOC estimation results of the lithium-ion batteries, using the HIF algorithm and extended Kalman filter (EKF) algorithm, are verified and compared under three conditions: uncertain measurement accuracy, uncertain SOC initial value, and uncertain application conditions. The simulation results show that the SOC estimation method based on HIF can ensure that the SOC estimation error value fluctuates within ±0.02 in any case, and is slightly affected by environmental and other factors. It provides a way to improve the accuracy of SOC estimation in a battery management system.

Accurate state of charge (SOC) estimation is key for the efficient management of lithium–ion (Li-ion) batteries, yet is often compromised by noise levels in measurement data. This study introduces a new approach that uses wavelet denoising with a machine learning regression model to enhance SOC prediction accuracy. The application of wavelet transform in data pre-processing is investigated to assess the impact of denoising on SOC estimation accuracy. The efficacy of the proposed technique has been evaluated using various polynomial and ensemble regression models. For empirical validation, this study employs four Li-ion battery datasets from NASA’s prognostics center, implementing a holdout method wherein one cell is reserved for testing to ensure robustness. The results, optimized through wavelet-denoised data using polynomial regression models, demonstrate improved SOC estimation with RMSE values of 0.09, 0.25, 0.28, and 0.19 for the respective battery datasets. In particular, significant improvements (p-value < 0.05) with variations of 0.18, 0.20, 0.16, and 0.14 were observed between the original and wavelet-denoised SOC estimates. This study proves the effectiveness of wavelet-denoised input in minimizing prediction errors and establishes a new standard for reliable SOC estimation methods.