25-08-2012, 05:06 PM
Identification of Dynamic Model Parameters for Lithium-Ion Batteries used in Hybrid Electric Vehicles
1Identification of Dynamic Model.pdf (Size: 97.94 KB / Downloads: 56)
Abstract:
This paper presents an electrical equivalent circuit model for lithium-ion batteries used
for hybrid electric vehicles (HEV). The model has two RC networks characterizing battery
activation and concentration polarization process. The parameters of the model are identified using
combined experimental and Extended Kalman Filter (EKF) recursive methods. The open-circuit
voltage and ohmic resistance of the battery are directly measured and calculated from
experimental measurements, respectively. The rest of the coupled dynamic parameters, i.e. the RC
network parameters, are estimated using the EKF method. Experimental and simulation results are
presented to demonstrate the efficacy of the proposed circuit model and parameter identification
techniques for simulating battery dynamics.
Introduction
Lithium-ion batteries are increasingly used in portable electronics, automotive and aerospace
applications, as well as in back-up power applications due to their high voltage, high energy
density, none memory effect, and low self-discharge during storage. Recently there has been a fast
growth in demand for large lithium-ion batteries for direct power supply of electric vehicles (EV)
and hybrid electric vehicles (HEVs). Accurate modeling of battery dynamics is important for
accurate simulation and optimization, and real time energy management of EVs and HEVs.
An accurate dynamic model of battery usually involves the relations of the terminal voltage to
current, power, temperature, state of charge (SOC), the effects of self-discharge, and the effects of
aging. Currently, there are three main types of model used to describe the relationship between
input and output of a battery system. These are electrochemical model [1, 2], artificial neural
network model [3, 4] and electrical equivalent circuit model [5-7]. The electrochemical battery model
describes the dynamic process of chemical reactions occurring on the electrodes based on
mathematical method, which can integrally reflect dynamic characteristics of the battery. However,
this model requires battery chemical parameters and detailed knowledge of the battery
construction and material properties which is not normally available to designers of vehicles.
Model formulation
The power assist unit in the hybrid electric vehicle described in this paper, is composed of 144
lithium-ion cells. Each battery module (battery box) consists of 16 lithium-ion cells. Ideally, tests
and model identification need to be carried out on each cell with a large quantity of computation.
But this will be expensive and time consuming. Alternatively, tests and modelling could be based
on one cell, and then multiply by144 to obtain a model of the whole battery pack. But this may not
be accurate due to tolerance variations between the battery cells. As a compromise the battery
module of 16 cells is regarded as the object for modelling, multiplying the amount of battery
modules as the total battery pack model.
The electrical circuit model is used to describe the relationship between the currents and voltages
measured at the terminal of the battery. The model used for the lithium-ion battery comprises three
parts, as shown in Fig. 1: 1) open-circuit battery voltageVoc , which is composed of an equilibrium
potentialVe and a hysteresis voltageVh , 2) internal resistance Ri contains the ohmic resistance Ro
and the polarization resistance.
Parameters estimation
Open-circuit voltage
The equilibrium potential is the open circuit voltage measured when the forward and reverse
reaction rates are equal in an electrolytic solution, thereby establishing the potential of an
electrode. The equilibrium potential of the battery, which is determined by Nernst equation,
depends on the temperature and the amount of active material left in the electrolyte. Fig. 2
illustrates the open-circuit voltage (OCV) as a function of SOC after charge and discharge at room
temperature. In this experiment, the battery was first discharged at constant current of 30 A from
fully charged state till 10 % of the nominal capacity (100 A h) was consumed. It was subsequently
left in open-circuit condition, while the open-circuit voltage was observed. After one hour, the
measured voltage was considered as equilibrium voltage since the rate of the increase of the open
circuit voltage was negligible and hence the battery was assumed to be got to a steady state. The
battery was subsequently discharged by a further 10 % of the nominal at the same current and the
equilibrium voltage measured after waiting for one hour, and the procedure was repeated to obtain
the remaining data points on the discharge curve in Fig. 2. The battery was then recharged at the
specified current, the equilibrium voltage after charge could be obtained every 0.1 SOC.
Conclusions
This paper presented an equivalent circuit model with two RC networks characterising battery
activation and concentration polarization process. The extended Kalman filter was used to
estimate the coupled parameters reflecting battery polarization characteristics. The parameters
characterising battery equilibrium potential and ohmic resistance were determined experimentally.
Simulation results using proposed model with the identified parameters, were found to agree
satisfactorily with the experimental results. Future work will focus on using EKF to estimate
battery states (for example SOC and state of health) and for online model parameters
identification.