Beruflich Dokumente
Kultur Dokumente
7, 2013
DOI: 10.1007/s11664-012-2460-4
Ó 2013 TMS
1.—Control Theory and Applications Centre, Faculty of Engineering and Computing, Coventry
University, Priory Street, Coventry CV1 5FB, UK. 2.—Jaguar Land Rover, Banbury Road, Gay-
don, Warwick CV35 0RG, UK. 3.—Ricardo UK Ltd, Southam Road, Radford Semele, Leamington
Spa, Warwickshire CV31 1FQ, UK. 4.—e-mail: navneesh.phillip@coventry.ac.uk
1900
Investigation of Maximum Power Point Tracking for Thermoelectric Generators 1901
extremum seeking control (ESC) algorithm; integrated within a full vehicle model for perfor-
although it is known that other MPPT algorithms mance analysis over drive cycle profiles.
exist, ESC is considered to be an appropriate choice
for the TEG application. For this reason the authors TEG MODEL DESCRIPTION
describe the development of a simplified transient
The HX subsystem has been modeled as a stack of
TEG model consisting of three subsystems, namely
several thermal layers as illustrated in Fig. 2. The
the HX, thermoelectric module/material (TEM), and
interactions between individual thermal layers are
PCU, which collectively serve as a surrogate of the
described using convection and conduction thermo-
real plant/system. The ESC algorithm has previ-
dynamic energy balance equations. When formu-
ously been used in photovoltaic applications,13,14
lating the energy balance equations it is assumed
where it has shown good tracking capability supe-
that the temperature distribution within the layers
rior to the P&O algorithm.15 Here the authors apply
is homogeneous,16 thereby simplifying the overall
the ESC algorithm to a simulated model of TEGs to
computation whilst retaining the main dynamic
investigate if similar tracking performance can be
features. Further detailed description of the model
achieved.
together with governing equations can be found in
Ref. 17, where 1-D heat transfer is considered. The
TEG MODEL INTEGRATION
HX is divided into several control volumes along
As aforementioned, the TEG system has been the flow of the exhaust gas stream. Heat flow from
modeled as an integration of three subsystems. the exhaust layer to the coolant layer, i.e., from hot
Figure 1 shows the input/output relations between to cold, is assumed to be uniform, a function of the
the individual subsystems, where the inputs for the mean temperature of the individual layer, and a
TEG model are exhaust and coolant inlet tempera- function of time. At this level of study, heat losses to
tures, Ta (K) and Tw (K), and corresponding mass adjacent control volumes and heat losses to ambient
_ a (kg/s) and m
flow rates, m _ w (kg/s). The gas-to-liquid are neglected given the uncertainty of environ-
HX takes these four inputs to simulate the hot- and mental effects within the system.
cold-side temperatures, Th (K) and Tc (K), for the Considering Eqs. 1 and 2, it is apparent that the
TEM. Given the hot- and cold-side temperatures, TEM is modeled using material data for n-type and
the TEM is able to simulate the open-circuit voltage, p-type legs of ncouple number of couples. A bismuth
Voc (V), and TEM internal resistance, Rm ðXÞ, using telluride (Bi2Te3) TEM has been validated in order
the thermoelectric (TE) equations as follows: to provide confidence in the modeling methodology.
The validated TEM model achieves an output with
Voc ¼ ncouple ðan þ ap Þ DT; (1) 94.6% accuracy at steady state. The model has been
developed such that the user can investigate a
variety of TEMs given the availability of their
ncouple ðqn þ qp Þ L material properties.
Rm ¼ ; (2) The transient nature of TEG operation, especially
A
for start–stop functionality auxiliary power units,
where a denotes the Seebeck coefficient (V/K), causes the power output of the TEG to vary con-
q denotes the electrical resistivity ðXmÞ, A denotes tinuously. For this reason it is essential to integrate
the cross-sectional area (m2), ncouple is the number of a PCU to regulate the output voltage. Depending on
couples, L denotes the leg length (m), and the sub- the choice and configuration of the TEMs, it is
scripts ‘‘n’’ and ‘‘p’’ refer to n-type and p-type legs. highly probable that the TEG voltage prior to con-
Voc and Rm are inputs to the PCU, where the voltage ditioning fluctuates higher and lower than the
and current is computed. The resulting power out to 14.4 V required to charge the battery. It is for this
the load is regulated at 14.4 V by a DC–DC con- reason that a step-up step-down buck–boost DC–DC
verter. The load in this case is considered to be a converter has been implemented within the PCU.
12-V car battery. In addition, the transient TEG Advantages of a buck–boost converter over a similar
model is suitably developed such that it can be step-up step-down single-ended primary inductor
capacitor (SEPIC) converter can be seen in Ref. 17, illustrates the waveform of current across the
where the authors observe that the SEPIC con- inductor under CCM mode.
verter is not as efficient as the buck–boost converter The DC inductor current ILB at the CCM/DCM
due to the utilization of two inductors rather than boundary is given as
the single inductor in the buck–boost topology.
An advancement made in this paper is to intro- Vo
ILB ¼ ð1 dmin Þ; (3)
duce a synchronous buck–boost converter. When 2fs Lmin
compared with a standard buck–boost converter, a
where fs denotes the switching frequency, dmin
synchronous buck–boost converter is capable of
denotes the minimum duty cycle, and Lmin denotes
converting voltage without reversing the terminals
the minimum inductance. For an ideal condition the
(inverting) across the load. In addition, a synchro-
output voltage of a synchronous buck–boost con-
nous DC–DC buck–boost converter provides higher
verter is given as
efficiency as compared with a standard DC–DC
buck–boost converter due to the fact that the diode d
utilized in a standard buck–boost is bypassed Vo ¼ Vin ; (4)
ð1 d Þ
(replaced) by a more efficient metal–oxide–semi-
conductor field-effect transistor (MOSFET). A basic where Vo denotes the output voltage, d denotes the
schematic diagram of a synchronous buck–boost duty cycle, and Vin denotes the input voltage. Since
converter is illustrated in Fig. 3. The converter the output is connected to a battery (constant volt-
comprises four MOSFETs, of which Q1 and Q4 must age), any change of d and Vin will not affect Vo .
be ON while the others, Q2 and Q3 , are in the OFF However, changes in d will cause changes in the
state. The converter provides a controllable and/or output current Io .
constant output DC voltage despite variation of the The output current is expressed as
input voltage. It also operates in two different
modes, namely continuous conduction mode (CCM) d
Io ¼ Iin : (5)
and discontinuous conduction mode (DCM). In ð1 dÞ
CCM, the inductor current flows continuously for
the entire period and never falls to zero, whereas in
DCM, the inductor current reduces to zero and
remains at zero level for the remainder of the period MAXIMUM POWER POINT TRACKING
before it begins to rise again. MPPT is an iterative approach which varies the
In this paper a synchronous buck–boost converter electrical output of a TEG to deliver the maximum
is considered to be operating in CCM mode. Figure 4 available power. Some of the commonly used MPPT
algorithms that have been used for photovoltaic
cells include P&O, constant voltage control (CVC),
constant current control (CCC), sophisticated ana-
log circuits, incremental conductance, and ESC.
Due to its perceived advantages, an ESC algo-
rithm is considered here for MPPT control applied
to a TEG. The performance of the ESC algorithm is
compared with the well-known P&O algorithm,
which serves as a benchmark. Maximum power is
essentially achieved by matching the internal
resistances of the TEG and the DC–DC converter.
determine the direction of change for maximizing the TEG temperature difference may give rise to
the output power. Figure 5 illustrates a flowchart of lack of tracking.
the P&O algorithm. Whilst the P&O approach The output power of the TEG oscillates around its
serves as a useful benchmark, it does have some maximum value; hence, even in the case of a
drawbacks as listed here: steady-state temperature difference, the output
power may not converge to its maximum value.
It is unable to adapt, hence any rapid variation of
1904 Phillip, Maganga, Burnham, Ellis, Robinson, Dunn, and Rouaud