Larry W. Juang

System identification-based lead-acid battery online monitoring system for electric vehicles

A system identification-based model for the online monitoring of batteries for electric vehicles (EVs) is presented. This algorithm uses a combination of battery voltage and current measurements plus battery data sheet information to implement model-based estimation of the stored energy, also referred to as state-of-charge (SOC), and power capability, also referred to as state-of-function (SOF), for deep-cycle batteries. This online monitoring scheme has been implemented for a bank of deep-cycle lead-acid batteries and experimental laboratory tests using simulated driving cycles have yielded promising results. In addition, actual road data from an EV powered by these same batteries has been analyzed with the proposed model to demonstrate the systems usefulness in determining the battery state-of-health (SOH). Finally, the limitation of the use of a linear model for battery terminal voltage behavior is discussed.

Improved Nonlinear Model for Electrode Voltage–Current Relationship for More Consistent Online Battery System Identification

An improved nonlinear model for the electrode voltage-current relationship for online battery system identification is proposed. In contrast to the traditional linear-circuit model, the new approach employs a more accurate model of the battery electrode nonlinear steady-state voltage drop based on the Butler-Volmer (BV) equation. The new form uses an inverse hyperbolic sine approximation for the BV equation. Kalman-filter-based system identification is proposed for determining the model parameters based on the measured voltage and current. Both models have been implemented for lead-acid batteries and exercised using test data from a Corbin Sparrow electric vehicle. A comparison of predictions for the two models demonstrates the improvements that can be achieved using the new nonlinear model. The results include improved battery voltage predictions that provide the basis for more accurate state-of-function readings.

Implementation of online battery state-of-power and state-of-function estimation in electric vehicle applications

A method for estimating battery state-of-function (SOF) is presented with a mathematical probabilistic statement within the context of Kalman filter estimation. The traditional state-of-power (SOP) metric is replaced with an equivalent statistic that delivers the desired SOF estimate with defined variance characteristics. To reduce error in the recursive estimator, a model based on an offline test relating the open-circuit voltage (OCV) to its rate of change with battery charge is introduced that provides better temperature insensitivity than the SOC vs. OCV model typically used in literature. Experimental test results for a LiFePO4 battery with a vehicle drive cycle are used to build confidence in the estimator results. Additionally, results from the proposed estimator are compared with results from the hybrid pulse power characterization (HPPC) test and the important model assumptions are discussed.

Improved modeling of lithium-based batteries using temperature-dependent resistance and overpotential

The temperature-dependent behavior of the resistance and overpotential of a lithium-iron-phosphate (LiFePO4) battery cell is explored in this paper. Offline experimental results from hybrid pulse power characterization (HPPC) tests and electrochemical impedance spectroscopy (EIS) methods for resistance and overpotential are explained using Arrhenius equations. Using a nonlinear regression technique, simulated drive cycle data are used to confirm the experimental findings and construct a generic cell model that explicitly takes temperature effects on the resistance and overpotential into account. The significance of the work lies in its confirmation of the inadequacy of the baseline linear-circuit model for lithium batteries at low temperatures and its presentation of a modeling approach that provides much better agreement with measured battery characteristics.

Loss optimization and ultracapacitor pack sizing for vehicles with battery/ultracapacitor hybrid energy storage

Battery and ultracapacitor hybrid energy storage systems have long been proposed as alternatives to battery-only systems for electrified vehicles. Prior research has focused on system topologies, dc/dc converter design, control methods, and evaluating specific applications. This work utilizes a rule-based control method and dynamic programming combined with a vehicle model to evaluate the reduction in energy storage losses that can be achieved for various drive cycles and ultracapacitor pack sizes. Dynamic programming is used to determine the optimum power split between the battery and ultracapacitor pack for a specified drive cycle and calculates the best performance a causal control algorithm could achieve. The analysis shows that dynamic programming can significantly improve the loss reductions compared to the basic rule-based control, making it possible to reduce the size of the ultracapacitor pack needed to deliver meaningful energy savings. Experimental tests have been carried out on battery and ultracapacitor cells that raise confidence in the accuracy of the model predictions for drive cycle metrics.

Improved nonlinear model for electrode voltage-current relationship for more consistent online battery system identification

An improved nonlinear model for the electrode voltage-current relationship for online battery system identification is proposed. In contrast with the traditional linear-circuit model, the new approach employs a more accurate model of the battery electrode nonlinear steady-state voltage drop based on the Butler-Volmer equation. The new form uses an inverse hyperbolic sine approximation for the Butler-Volmer equation. Kalman filter-based system identification is proposed for determining the model parameters based on the measured voltage and current. Both models have been implemented for lead-acid batteries and exercised using test data from a Corbin Sparrow electric vehicle. A comparison of predictions for the two models demonstrates the improvements that can be achieved using the new nonlinear model. The results include improved battery voltage predictions that provide the basis for more accurate state-of-function (SOF) readings.

Modeling of second-life batteries for use in a CERTS microgrid

A second-life battery is an electric vehicle battery pack that has reached an end-of-life condition for its vehicular use, yet retains enough performance to be re-purposed for another application. One promising application of a second-life battery is stationary energy storage within a CERTS microgrid. This paper investigates the modeling of multiple paralleled traction battery packs within a CERTS microgrid, examining the impacts of elevated internal pack impedance on microgrid system operation. Impedance spectroscopy and hybrid-pulse power characterization are used to model vehicular Li-ion cells under a range of conditions that include second-life aging. The ac bus dynamics of the microgrid model are validated experimentally. Second-life battery models are incorporated into two CERTS microgrid architectures and system-level effects of changing the battery impedance are explored. Simulation results indicate that the modeled EV second-life batteries deliver promising performance characteristics in both CERTS microgrid architectures that were investigated.

Design of an electric powertrain for a Ford F150 crew cab truck utilizing a lithium battery pack and an interior PM synchronous machine drive

The design of an electric drive system for a Ford truck is presented. The electric drive system is designed to have similar performance to the stock truck, a 2002 model year Ford F150, with a 4.2L 150kW (peak) gasoline-powered engine. The power capabilities of lithium batteries from two manufacturers have been measured to aid in designing a battery pack that can supply sufficient voltage, power and range. A prototype interior permanent magnet (IPM) synchronous machine has been designed that can produce the desired output power with the chosen battery pack and a commercial motor controller. Dynamometer test results are provided for the machine that demonstrates a close match to the predicted performance. Additionally, the other vehicle system components that must be electrified as well as the custom data acquisition and battery management systems are described.

Evaluation of an electromechanical model for a Corbin Sparrow electric vehicle

An electromechanical model of a Corbin Sparrow, a small electric vehicle, is evaluated using a collection of experimental drive data. The vehicles stock electromechanical drive train and a custom-designed data logging system are presented, and the model equations and parameters are summarized. The experimental results demonstrate that the model can deliver promising accuracy when estimating the vehicle power versus time as a function of the vehicles velocity, acceleration, and vertical velocity. A single drive trip and a collection of drive trips are analyzed, and the model is shown to have a typical error range of approx. −5%/+20%. Statistics for the collected drive data are also provided, including the vehicles measured energy consumption per mile, motor efficiency, and motor controller efficiency. Finally, the Sparrows energy consumption is compared to that of electric vehicles tested by the US Department of Energy, and the Sparrow performance is predicted with installation of regenerative braking, a more efficient motor and controller, and a lithium-ion battery pack.

The impact of DC bias current on the modeling of lithium iron phosphate and lead-acid batteries observed using electrochemical impedance spectroscopy

This paper deploys electrochemical impedance spectroscopy (EIS) to investigate the impact of temperature and dc bias current on battery impedance characteristics. Measured test results are used to demonstrate that, under conditions where the nonlinear Butler-Volmer equation is necessary to model the electrode charge transfer characteristics, the semicircular trajectory that typically appears in the EIS results shrinks in diameter as the batterys dc bias current increases. For a lithium-based battery, the nonlinearity introduced by the Butler-Volmer relationship is more pronounced at low temperature, while lead-acid batteries typically exhibit this nonlinearity even at room temperature. The impact of dc bias current on the battery model and EIS characteristics are thoroughly investigated using a combination of experimental tests combined with theoretical justification based on the Arrhenius equation. The value of using this observed relationship to improve the accuracy of battery models and condition monitors (e.g., state-of charge, etc.) is discussed in the paper.