THESIS
2020
xx, 152 pages : illustrations (some color) ; 30 cm
Abstract
With the advantages of accurate and elaborate, the application of the electrochemical
models has become a potential trend for states estimation and performance evaluation
of Li-ion batteries in an advanced battery management system (BMS). This thesis
investigates an original approach for joint estimation of state-of-charge (SoC) and
state-of-health (SoH) based on the pseudo-two-dimensional (P2D) electrochemical
model. First, the standard P2D model is established and reformulated into a nonlinear
state-space form with guaranteed observability. Next, the particle filter (PF) algorithm
is employed to estimate the average lithium concentration in real-time for SoC
calculation. In addition, the battery SoH is calibrated based on the prediction of the
average lithium concentration at...[
Read more ]
With the advantages of accurate and elaborate, the application of the electrochemical
models has become a potential trend for states estimation and performance evaluation
of Li-ion batteries in an advanced battery management system (BMS). This thesis
investigates an original approach for joint estimation of state-of-charge (SoC) and
state-of-health (SoH) based on the pseudo-two-dimensional (P2D) electrochemical
model. First, the standard P2D model is established and reformulated into a nonlinear
state-space form with guaranteed observability. Next, the particle filter (PF) algorithm
is employed to estimate the average lithium concentration in real-time for SoC
calculation. In addition, the battery SoH is calibrated based on the prediction of the
average lithium concentration at the cut-off voltages of battery charging and
discharging processes, which, in turn, improves the accuracy of online SoC estimation
for aged batteries. For validation purposes, fifteen batteries with the aging state from
100 % to 70 % of initial capacity are tested under dynamic current profiles. The
results show that the maximum SoH estimation error can be limited to 2.8%, and the
SoC estimation error is bounded by 2% for new and aged batteries with the calibrated
SoH.
The trustworthy state estimation of the electrochemical model based method is dependent on the accuracy of the model parameters being used, which need to be
determined for individual batteries under dynamic operating conditions. In this thesis,
critical model parameters are investigated and a novel online parameter identification
method for the electrochemical model is established. In the method, the transfer
function of the P2D model output equation is simplified to an auto regressive
exogenous (ARX) form and the recursive least square (RLS) algorithm is employed
for online parameter identification. The effectiveness of the proposed method is
validated and compared with the conventional offline particle swarm optimization
(PSO) method. The simulated battery terminal voltage using the identified model
parameters are compared with the experimental measurements under 0°C, 25°C and
45°C. The results show that the proposed method generates more accurate model
output than the conventional method with a much lower computational cost. The
design scheme of a BMS platform specifically developed for electric vehicle (EV)
application is also briefly presented in this thesis, which can also be used for verifying
the applicability of the proposed state estimation and parameter identification
algorithms.
Post a Comment