03-05-2012, 01:16 PM
State of charge estimation using an unscented filter for high power
lithium ion cells
2010 State of charge estimation using an unscented filter for Santhanagopalan White.pdf (Size: 270.02 KB / Downloads: 114)
INTRODUCTION
State of charge (SOC) is a measure of the amount of
deliverable capacity at any given instant when the
battery is used [1]. Estimation of SOC is an active area
of research and several approaches have been
presented in the literature [2–28] to monitor the
SOC of a cell. These include following the physical
property of the cell, such as the internal resistance or
the electrolyte density [3,4], measurement of the cell
impedance [5–8], optical methods [9], eddy current
methods [10], equivalent circuit analysis of the charge/
discharge and/or impedance curves [11–18,22], techniques
employing a Kalman-filter [19–21], techniques
ELECTROCHEMICAL MODEL FOR A LITHIUM ION CELL
A rigorous electrochemical model for a lithium ion
cell was proposed by Doyle et al. [35]. Several
attempts to simplify the model equations for the
battery are reported in the literature [29,30,32,33].
We presented a comparison of the different
physics-based models in an earlier work [32].
Here, we illustrate the concept of using an
unscented filter using a rigorous electrochemical
model.
THE UNSCENTED FILTER
The use of an EKF to estimate SOC of lithium ion
batteries is straightforward and has been recommended
due to the simplicity of the filtering
technique. We presented SOC estimation results
from a very simple electrochemical model employing
the EKF earlier. However, extension of these
results to batteries for electric vehicles that are
typically operated at very high currents ranging
from 5 to 10C rate discharges did not yield
predictions with sufficient accuracy. Under these
rigorous conditions, the performance of the cell is
limited by several factors such as the diffusion
limitations in the electrolyte, transport in the solid
phase and the energetics. The single particle model
presented before [41] does not account for solution
phase limitations.
RESULTS AND DISCUSSION
In the case of high power cells, the
instantaneous capacity during dynamic loads is
currently estimated by measuring the instantaneous
cell voltage. However, this approach poses a few
complications: the cell voltage measurements during
such pulse loads using on-board devices is often not
sufficiently accurate to provide good estimates of
the SOC and second, in case of systems employing a
chemistry with a flat voltage profile, the cell voltage
model shows poor convergence.