Research On SOH Estimation And RUL Prediction Methods Of Lithium-ion Battery For Electric Vehicles

Along with the rapid progress of electric vehicle(EV), the EV battery’s producing and managing technologies attach increasingly general attention. Among the technologies, the methods of estimating battery’s State of Health(SOH) and predicting Remaining Useful Life(RUL) become the focus of investigation, which can obviously lead to improvement in endurance mileage and battery life. State of Charge(SOC) is a core parameter of battery, which also attracts broad interest among researchers. As a result, we accomplish the work in this paper for a better accuracy of the estimation and prediction.The influence of discharging current and cycle number to battery aging are analyzed based on electrochemical theories. First the battery cycling experiment under constant temperature is operated. Then employing the experiment results, we discuss the aging facts including cycle number and discharge rate. Finally we study on the electrochemical mechanism of Solid Electrolyte Interface(SEI), transformation of electrode structure and degradation of active material.Models for SOH estimation and RUL prediction are built. Thevenin model is chosen as the equivalent circuit, and then the kernel parameters are recognized, referring to curve fitting for Open Circuit Voltage(OCV) vs. SOC by rapid calibration method, and exponential fitting for resistance and capacitance value. Capacity variation model and attenuation model represent how the capacity varies and attenuates during aging respectively. Capacity variation model shares output equation with Thevenin model. And capacity attenuation model is exponentially fitted by the data of nominal capacity attenuation.The co-estimator of Lithium-ion battery SOC and SOH, named Multi-scale Extended Kalman Filter(MEKF), is proposed based on multi-scale theory. MEKF can be determined based on the scale-transformed Thevenin and capacity variation model and will benefit the co-estimation with reduced computational complexity and remained accuracy. Based on the simulation results of experiments adopting New European Driving Cycle and JC08, we compare MEKF and regular Dual Extended Kalman Filter(DEKF) to show the benefits.A novel particle filter for RUL prediction is proposed in this paper to solve the RUL uncertainties caused by battery’s unknown future operating condition. We first employ Monte Carlo method to present the probability of remaining cycling number, and then utilize particle filter to update the model parameters as states in attenuation model. Finally the probability distribution of battery RUL is estimated by the updated model. The expectation of RUL calculated by simulation results is consistent with measured data.This paper focuses on SOH estimation and RUL prediction of EV battery. We design cycling experiment to study on the influence of discharging current and cycle number to battery aging. Then we build equivalent circuit model, capacity variation model, and capacity attenuation model. Finally co-estimation of SOC and SOH adopting multi-scale theory and RUL prediction employing a novel particle filter are accomplished based on the three proposed models. This work reduces computational complexity of SOC and SOH co-estimation and improves accuracy of RUL prediction.