Search

Batteries are interconnected in series and/or parallel to meet wide-range power or energy demands in various industrial applications. To pursue the benefits of multiple connection structures in one system, reconfigurable battery systems (RBSs) have recently emerged for safe and efficient operation, extended energy storage, delivery, etc. Switches are the essential elements to enable the battery system reconfiguration, but selecting appropriate switches for RBS designs has not been systematically investigated. To bridge this gap, analytical expressions are derived in this article to estimate the maximum switch current and its upper limit to facilitate the selection of RBS switches. An RBS prototype based on H-bridges is set up and experimental results verify the effectiveness and advantage of the proposed estimation method. These analytical expressions, relying only on resistances of batteries and switches, are readily applicable to practical RBS design and much more efficient than conducting numerous circuit experiments, simulation tests, or circuit analyses, especially for large-scale systems. Moreover, the analysis framework and estimation method proposed for series–parallel mutual conversion can be adaptively extended to other complex system reconfigurations to facilitate various RBS designs. [ABSTRACT FROM AUTHOR]

Knowing the health state of the batteries would enhance the energy storage system’s reliability and safety, especially for fast charge applications. Here we propose a synergetic method with the help of the genetic algorithm (GA) and the support vector regression (SVR) for SOH estimation. Firstly, features for battery aging process description are selected from the multi-source data, including current, voltage, and temperature, in the battery charging process. The SVR is then employed to establish a battery aging model and estimate the SOH with the generated features. Afterward, the feature set which can optimize the pre-set objective, namely minimizing the SOH estimation error and the defined difficulty of feature acquisition, are selected by the GA via an iterative process. Experimental results indicate that the selected feature set generated from the charged capacity and temperature rise data may perform a better SOH estimation than the manually picked features and the optimized ones from a single source. Moreover, by collaborating with the chosen features, the SVR is found to have a similar SOH estimation accuracy to a more complex algorithm while using less computation power. Furthermore, it should be noted that the selected features are obtainable in about 95% of the charging operations according to the voltage distribution resulting from more than 40,000 actual charging bills. [ABSTRACT FROM AUTHOR]

With the continuous improvement of lithium-ion batteries in energy and power density, their safety and reliability concern is becoming increasingly urgent for energy storage systems. Thermal faults are one of the most critical faults in lithium-ion batteries that must be diagnosed in real time, because they can be potentially catastrophic. Motivated by this fact, a diagnostic algorithm is presented in this article to diagnose several thermal faults in cylindrical lithium-ion batteries, including heat generation fault and thermal parameter fault. This diagnostic algorithm is based on an electrothermal-coupled model describing both electrical and thermal dynamic behaviors of batteries. A Lyapunov-based battery internal resistance estimator is proposed to describe the dynamics of the surface and the core temperature of a battery cell, because the heat generation caused by the electrical loss is highly dependent on battery resistance. An observer-based fault detection framework is then established based on surface temperature measurements and resistance estimates, where stability and convergence are guaranteed by Lyapunov direct method. Furthermore, the adaptation law of fault evaluation threshold is designed to suppress modeling and measurement uncertainties. Simulation studies are presented to illustrate the effectiveness of the proposed model and diagnosis algorithm. [ABSTRACT FROM AUTHOR]

The state of charge (SOC) of lithium-ion batteries (LIBs) is an important evaluation index in battery management system of energy storage systems. However, it is always a challenging task to accurately estimate SOC of LIBs, because of the existence of nonlinear characteristics and significant temperature effects. To improve the SOC estimation accuracy and robustness, a model-based estimation approach for SOC and impedance of LIBs is proposed against uncertain loading profiles and ambient temperatures. First an auto regressive exogenous model for online model order determining and parameters identification is established to monitor parameters variations based on numerical subspace state space system identification method. Second, a sequential Monte Carlo filter is employed to overcome the nonlinear and non-Gaussian error distribution state estimation problems caused by complicated open-circuit voltage characteristics. Finally, evaluation of the adaptability and generality of the proposed method are verified by different LIBs under different operating conditions. Experimental results indicate that the proposed method shows great performance, whose estimation value converges to real SOC within an error of $\pm \text{3}\%$ , and the battery model can simulate battery dynamics robustly with high accuracy. [ABSTRACT FROM AUTHOR]

The state-of-charge estimation of lithium-ion batteries plays a key role in energy storage systems such as battery monitoring, fault detection, power and energy optimization control. However, it is technically challenging, in particular, for the simultaneous estimation of battery internal impedance and state-of-charge, which are two key state variables affecting battery performance. This paper reveals that the commonly used state-of-charge estimation schemes based on Bayesian filters are fundamentally flawed in taking state constraints into account. Constrained Bayesian dual filtering framework for parameter estimation and state-of-charge estimation are designed in this paper to improve the estimation accuracy and robustness. After a state-of-charge and open-circuit-voltage mapping is accurately identified, a dual-filtering framework is introduced to simultaneously estimate the state-of-charge and model parameters which gives rise to the dynamics. The inequality constraints of state variables in Bayesian dual-filtering framework are also taken into account. The state-of-charge and model parameter estimation results of the constrained dual-filtering are regarded as the mean of constraints. Extensive comparative experiments are conducted to validate that the proposed method is superior over existing methods in providing improved accuracy and robustness. [ABSTRACT FROM AUTHOR]

Accurate remaining useful life (RUL) prediction and state-of-health (SOH) diagnosis are of extreme importance for safety, durability, and cost of energy storage systems based on lithium-ion batteries. It is also a crucial challenge for energy storage systems to predict RUL and diagnose SOH of batteries due to the complicated aging mechanism. In this paper, a novel method for battery RUL prediction and SOH estimation is proposed. First, a novel support vector regression-based battery SOH state-space model is established to simulate the battery aging mechanism, which takes the capacity as the state variable and takes the representative features during a constant-current and constant-voltage protocol as the input variables. The estimated impedance variables are taken as the output due to the correlation between battery capacity and the sum of charge transfer resistance and electrolyte resistance. Second, in order to suppress the measurement noises of current and voltage, a particle filter is employed to estimate the impedance degradation parameters. Furthermore, experiments are conducted to validate the proposed method. The results show that the proposed SOH estimation method can provide an accurate and robustness result. The proposed RUL prediction framework can also ensure an accurate RUL prediction result. [ABSTRACT FROM PUBLISHER]

Cell variations caused by the inevitable inconsistency during manufacture and use of battery cells have significant impacts on battery capacity, security and durability for battery energy storage systems. Thus, the battery equalization systems are essentially required to reduce variations of in-pack cells and increase battery pack capability. In order to protect all in-pack cells from damaging, estimate battery state and reduce variations, a system state estimation and energy optimal control framework for multicell lithium-ion battery system is proposed. The state-of-charge (SOC) values of all in-pack cells are firstly estimated using a dual-scale extended Kalman filtering (EKF) to improve estimation accuracy and reduce computation simultaneously. These estimated SOC values provide specific details of battery system, which cannot only be used to protect cells from over-charging/over-discharging, but also be employed to design state-feedback controller for battery equalization system. A two-stage hybrid state-feedback and output-feedback equalization algorithm is proposed. The state-feedback controller is firstly employed for coarse-grained adjustment to reduce equalization time cost with large current. However, due to the inevitable SOC estimation errors, the output-feedback controller is then used for fine-grained adjustment with trickle current. Experimental results show that the proposed framework can provide an effectively estimation and energy control for multicell battery systems. Finally, the implementation of the proposed method is further discussed for the real applications. [ABSTRACT FROM AUTHOR]

Due to lithium-ion batteries' complex behaviors, accurate estimation of state-of-health is still a critical challenge in battery systems' prognosis and health management. Most existing efforts in battery health prognosis focus on feature engineering using low-frequency sampled time-domain response. These efforts may not completely reflect the battery health status in automotive applications due to information missing in the high or medium frequency range. This paper proposes a data-driven state-of-health estimation method using high and medium frequency range impedance spectroscopy data. First, battery health indicators are extracted from electrochemical impedance spectroscopy data. It is found that the Nyquist diagram shows semicircle characteristics at high and medium frequency ranges. The center and radius of this circle show high dependence on battery health. Then, a recurrent Gaussian process regression with a one-step delay feedback loop is designed to provide a smooth and accurate battery state-of-health estimate. Finally, the proposed health indicators and state-of-health estimators are validated using experimental data on different cells. The results demonstrate the high accuracy and robustness of the proposed health indicators and state-of-health estimator, suggesting a 1.12% estimation error. This study shows the prospect of health prognosis using robust geometric impedance spectrum indicators in energy storage systems. • Geometric impedance spectrum features are extracted for battery health estimation. • A recurrent Gaussian process regression is employed for smooth health estimation. • Health indicators are evaluated by Spearman correlation and Pearson coefficients. • The proposed methods are validated by extensive experimental datasets. [ABSTRACT FROM AUTHOR]

The state-of-energy of lithium-ion batteries is an important evaluation index for energy storage systems in electric vehicles and smart grids. To improve the battery state-of-energy estimation accuracy and reliability, an online model-based estimation approach is proposed against uncertain dynamic load currents and environment temperatures. Firstly, a three-dimensional response surface open-circuit-voltage model is built up to improve the battery state-of-energy estimation accuracy, taking various temperatures into account. Secondly, a total-available-energy-capacity model that involves temperatures and discharge rates is reconstructed to improve the accuracy of the battery model. An extended-Kalman-filter and particle-filter based dual filters algorithm is then developed to establish an online model-based estimator for the battery state-of-energy. The extended-Kalman-filter is employed to update parameters of the battery model using real-time battery current and voltage at each sampling interval, while the particle-filter is applied to estimate the battery state-of-energy. Finally, the proposed approach is verified by experiments conducted on a LiFePO 4 lithium-ion battery under different operating currents and temperatures. Experimental results indicate that the battery model simulates battery dynamics robustly with high accuracy, and the estimates of the dual filters converge to the real state-of-energy within an error of ±4%. [ABSTRACT FROM AUTHOR]

• A voltage fault detection method for lithium-ion battery pack is proposed. • The proposed method is based on system identification and outlier detection. • The recursive least squares method is employed for parameter identification. • The local outlier factor and Grubbs rule are utilized to detect the fault. The lithium-ion battery is the critical component in the microgrid energy storage systems. Affected by factors such as abuse operation and aging, voltage fault including over-voltage and under-voltage may occur to battery, which implies more serious faults including short-circuit, thermal runaway and so on. Detecting the voltage fault accurately is critical for enhancing the safety of battery pack. Therefore, this paper presents a voltage fault detection method for lithium-ion battery pack using local outlier factor (LOF). The proposed method systematically incorporates a model-based system identification algorithm into an outlier detection algorithm. Firstly, the model parameters are identified to represent the battery dynamics, which indicate the fault condition of the battery. By doing this, the fault detection problem is converted into detecting abnormal parameter in the model parameter set. Then, the local outlier factor is utilized to describe the degree of abnormality of parameter by evaluating the local deviation of the observed data with respect to its neighbors. Finally, an outlier filter based on Grubbs criterion is utilized to detect the fault battery cell using calculated local outlier factors. The simulations and experimental results show that the proposed method can detect the fault accurately, which verify the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]