Development of an intelligent Li-ion battery management system for electric vehicles
Date of Issue2016
School of Electrical and Electronic Engineering
To improve electric vehicle (EV)’s operation performance and reliability, an energy management system (EMS) should be designed to supervise and control the power sources including lithium-ion (Li-ion) batteries. In this thesis, some topics relevant to implementation of a new intelligent EMS for EVs and focusing on Li-ion battery’s operation and performance are researched. The intelligent EMS is based on some novel methods for modeling and estimation of Li-ion batteries, cell equalization in battery pack and power management strategy design for the power sources in EVs. The first part of this thesis focuses on Li-ion battery modeling. Various types of equivalent circuit model are established and compared. It is verified that the series circuit model with two resistor-capacitor (RC) networks has good performance. The model using fuzzy logic to describe the temperature effect based on experiments is proposed. Then, another new type of battery model trained by the extreme learning machine (ELM) algorithm is proposed in experimental condition with simple current patterns. The ELM model performs simpler modeling process and better accuracy comparing with existing radial basis function (RBF) neural network (NN) battery model. Based on existing and proposed models, Li-ion battery state of charge (SOC) estimation is researched and improved in this thesis. The Kalman filter (KF)-based methods including extended Kalman filter (EKF), unscented Kalman filter (UKF), adaptive EKF (AEKF) and adaptive UKF (AUKF) are applied on the ELM model. Experimental results and comparisons indicate that the AUKF algorithm achieves improved SOC estimation performance with better accuracy and faster convergence rate. The estimation results also verify that the ELM model is more suitable for SOC estimation than the conventional RBF NN model. Considering the battery model’s flexibility, accuracy and practical operation conditions, the particle filter (PF) methods are applied on an accurate nonlinear Li-ion battery equivalent circuit model. The model represents the circuit parameters’ variation according to SOC by nonlinear functions and achieves better accuracy than constant parameter circuit models. The algorithms of PF and unscented particle filter (UPF) for nonlinear systems are executed to estimate Li-ion battery SOC. The estimation results reveal that UPF has better accuracy and faster convergence rate than PF. However, the computational load for the PF methods is heavier, bringing limitations in EMS’s applications. Then, the accurate nonlinear equivalent circuit model is simplified to a constant circuit parameter model with system uncertainties to achieve simpler modeling and estimation process. The sliding mode observer with high accuracy and light computation is applied. The adaptive gain technique is used in the observer and SOC estimation with good performance is provided by this proposed adaptive observer. The adaptive observer based on sliding mode scheme is also applied to estimate Li-ion battery SOC and state of health (SOH) simultaneously. The estimation results verify that the proposed observer scheme has accurate and robust performance on battery SOC and SOH estimation. Another part in this thesis is about cell equalization in Li-ion battery packs. Since cell imbalance brings damage to Li-ion batteries in the pack, considering circuit size, system implementation and cost, the equalization method using switched capacitors is applied and improved for series battery strings in battery pack in this thesis. The proposed modularized cell equalization schematic using chain structure switched capacitors achieves fast equalizing speed and small voltage across the switches. Another modularized double switched capacitor equalization schematic considering battery SOC and SOH is also proposed to improve the equalizing efficiency and reliability. The trade-off between equalization speed and system simplicity should be considered to select the appropriate equalization schematic in applications. The last part of this thesis designs the power management strategy in EMS for EVs. Fuzzy logic control strategy is proposed and Li-ion battery’s aging levels represented by SOH is considered in the control system. By designing fuzzy rules, the fuzzy control strategy realizes the functions of EV power distribution and power sources operation supervision for various EV driving actions. Specifically, Li-ion battery life is extended by the designed fuzzy control strategy considering SOH. To summarize, the proposed power management strategy in EMS for EVs achieves good performance on EV power sources operation, supervision and Li-ion battery service life extension.