Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
1

A Linear Extrapolation Based MPPT Algorithm


for Thermoelectric Generators under
Dynamically Varying Temperature Conditions
B. Bijukumar, Student Member, IEEE, A. G. Kaushik Raam, Saravana Ilango Ganesan, Senior
Member, IEEE and Chilakapati Nagamani, Senior Member, IEEE

Most of the above applications are susceptible to frequent


Abstract-- This paper presents a new Maximum Power Point temperature variations. Investigators have analyzed the impact
Tracking (MPPT) technique for extracting the maximum power in of thermal imbalance on the optimal electrical operating point
Thermoelectric Generator (TEG) systems. Considering the linear of the TEG [9]. It is noted that the thermal imbalance moves the
nature of I-V characteristics of a TEG source, the proposed
electrical operating point away from the optimal point. Optimal
technique is based on the linear extrapolation principle. The I-V
co-ordinates of two random operating points are used in operating point is the point at which the TEG can deliver
computing the co-ordinates of new MPP. As anticipated, the maximum power to the load, also called Maximum Power Point
proposed method involves only three sampling periods to reach the (MPP). Therefore, an efficient Maximum Power Point Tracking
MPP under any dynamic conditions. The Linear Extrapolation (MPPT) method is essential to maximize the power transfer
based MPPT (LEMPPT) technique has several advantages such as capability of TEG.
simplicity of analysis, fixed and minimal convergence time, ease of
Among the various MPPT algorithms, Perturb and Observe
implementation and a high tracking efficiency due to the absence
of steady-state oscillations. Further, there is no need to interrupt (P&O), Incremental Conductance (INC), Fractional Open
the circuit for measuring the open-circuit voltage of TEG. Also, Circuit Voltage (FOCV) and Fractional Short Circuit Current
there is no requirement for additional switches or components. (FSCC) are widely used [11], [12]. P&O is a classical algorithm
The simulation and test results show that the proposed algorithm which moves the operating point to the peak by perturbing the
has a superior performance as compared to that of the Perturb voltage [13], [14]. Since no stopping criterion exists, the
and Observe (P&O) algorithm.
operating point oscillates around the MPP. These steady-state
oscillations can be reduced by adopting smaller step size. But
Index Terms— DC-DC boost converter, linear extrapolation
based MPPT (LEMPPT), maximum power point (MPP), perturb this, in turn, affects the speed of operation and increases the
and observe (P&O) method, thermoelectric generator (TEG). time taken to reach MPP. Thus, a fixed step size in perturbation
gives rise to poor performance in steady-state and dynamic
I. INTRODUCTION response. INC algorithm adopts an error margin and an
operating point within this range is considered as MPP. For
T HE rising concerns about pollution caused by conventional
energy sources such as coal, petroleum etc. have propelled
the need for an energy source with minimum environmental
better steady-state operation, the step size should be smaller
than the error margin [12], [15]. Therefore, both P&O and INC
impact. Thermoelectric Generator (TEG) is one such source algorithms are in need of variable step size to eliminate this
which employs waste heat to generate electricity. In the Internal drawback [16], [17].
Combustion (IC) engine of automobiles, about 40% of the input Some research works [18]-[22] have focused on
energy gets wasted in the form of exhaust heat [1]. In implementing MPPT algorithms based on measurement of
manufacturing process such as boilers, furnaces, reactors etc. open-circuit voltage (OCV)/short-circuit current (SCC). In
around 20 to 50% of the energy consumed is lost in the form of [12], the authors have reported that FOCV/FSCC control
hot exhaust gases [2]. Researchers have reported the potential algorithms render better steady-state performance compared to
application of thermoelectric waste heat recovery in automobile P&O and INC. However, the major problem is loss of power
industry [3], boilers [4], steam-based power plants [5], due to disconnection of TEG from the converter during the
combustion systems [6], and furnaces etc. [7]. TEG is an period of measurement. In [19], [20], the authors have proposed
environment-friendly source of electric power which has no a MPPT method by measuring the OCV with minimal loss of
moving parts, silent in operation and reliable [8]. But, this power. The disconnection of input capacitor causes overvoltage
energy harvesting method encounters major challenges such as transients which reduces the speed of the program and requires
low energy efficiency, mismatch of temperature and subsequent more number of measurement cycles [19]. These overvoltage
changes in internal resistance [9], [10]. transients are reduced in [20] using a snubber circuit which

B. Bijukumar, A.G. Kaushik Raam, Saravana Ilango Ganesan and Chilakapati Nagamani are with the Department of Electrical and Electronics Engineering,
National Institute of Technology, Tiruchirappalli- 620015, India. (email: bkbijuzion@gmail.com, kaushik.raam96@gmail.com, gsilango@nitt.edu,
cnmani@nitt.edu)

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
2

requires extra components. This method is not suitable for resistor ‘Rint’ [28], [29]. The electrical equivalent circuit of a
continuous source current dc-dc converter topologies. TEG module is shown in Fig.1.
A simple MPPT control method is proposed in [23], which From the equivalent circuit, the current flowing through the
calculates the duty cycle at MPP without measuring the OCV. load can be determined as follows,
This method is applicable only when the load is constant. Also, V OC
the authors have approximated that the relationship between I TEG = (2)
OCV and duty ratio is linear in a small operating range. Hence,
(R int + RL )
this method cannot track MPP for the remaining operating where VOC is the open-circuit voltage across the TEG terminals,
range and requires prior knowledge of converter parameters. An Rint is the internal resistance of the TEG module and RL is the
indirect OCV and SCC detection based MPPT algorithm is load resistance.
proposed in [24]. This method requires a robust current
controller for calculating the optimum current value and
employs P&O algorithm near MPP in which small oscillations
are bound to exist. In [25], the OCV is calculated using curve
fitting method and the P&O method is used to track the MPP.
Hence, steady-state oscillations persist at MPP thereby
necessitating a robust PI controller to attain the operating
voltage corresponding to each reference voltage. A current-
sensorless MPPT algorithm is proposed in [26] which uses a
hysteresis controller and a reference voltage generator based on Fig. 1. Equivalent circuit of a thermoelectric generator
P&O. Even though no steady-state oscillations exist at MPP, The power delivered to the load can be determined as
the tracking speed depends on the step size of the change in follows,
reference voltage. Most of the aforementioned algorithms VOC 2
involve random, although guided, tracking using a closed loop P= × RL (3)
controller as they work based on trial and error principle. Also, (Rint + R L )2
the operating point perpetually oscillates around MPP in From (2), it can be deciphered that load current varies
steady-state. Unlike most other guided tracking methods based linearly with the voltage across TEG whereas from (3), it is seen
on trial and error principle, the proposed method involves the that the variation of power w.r.to current or voltage is parabolic
direct computation of the optimal duty ratio to reach the MPP. in nature. By maximum power transfer theorem, maximum
This paper presents a new MPPT algorithm which excludes power is transferred from source to load when Rint = RL. The
the use of closed loop controller and steady-state oscillations, condition for maximum power in a TEG can be expressed as
and takes a fixed time of three sampling periods to track the follows,
MPP. The idea underlying the algorithm is to calculate the OCV
without using trial and error control. In this method, the V MPP = 0.5 × VOC and I MPP = 0.5 × I SC (4)
magnitude of OCV and SCC are calculated from the measured
where VMPP and IMPP are the voltage and current at MPP and ISC
voltage and current at two different values of the duty cycle.
is the short-circuit current.
The duty ratio pertaining to the MPP is in turn calculated from
Since the I-V characteristic of the TEG is linear, it is possible
the OCV and SCC thus obtained. The theory behind the to obtain the OCV and the SCC through linear extrapolation by
proposed algorithm is explained in section III and its working measuring the I-V coordinates of two random operating points.
is described in section IV. The results of simulations and The coordinates of maximum power point then correspond to
experiments under steady-state and dynamic conditions are 50% of ISC and 50% of VOC. Therefore, one can easily find out
presented in section V. The paper is finally concluded in section the MPP by measuring or calculating VOC and ISC.
VI.
III. PROPOSED LINEAR EXTRAPOLATION BASED MPPT
II. TEG SYSTEM MODELING (LEMPPT) METHOD
TEG is a solid semiconductor device which works on the The theoretical concepts of the LEMPPT scheme are
principle of Seebeck Effect. If a temperature difference is described in this section. The basic principle of this method is
maintained between the hot and cold junction of the device, an linear extrapolation in which the co-ordinates of the MPP are
electromotive force will be induced across the couple [27] and computed based on the extrapolation of two measurable
is given by, operating points. The LEMPPT method mainly involves two
computation process- (i) Calculation of open-circuit voltage
V OC = α np × (T h − T c ) = α np × ∆ T (1) and short-circuit current through extrapolation and (ii)
Calculation of duty ratio at MPP (DMPP). Thus, the VOC can be
where αnp is the Seebeck coefficient of the material, Th and Tc indirectly calculated without opening the TEG terminals just by
are the hot and cold side temperatures and ∆T is the temperature measuring the source voltage and current. The linear I-V
difference across the junctions. relationship of TEG and impedance matching property of boost
Under steady-state, a TEG module can be electrically converter are employed in this method to calculate the DMPP.
modeled as a constant voltage source ‘VOC’ in series with a

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
3

A. TEG with Boost Converter From (11) and (12), it can be mathematically proven that the
The optimal operating point of the TEG module is the point points A (V1, I1) and B (V2, I2) are located on the straight line.
at which the load resistance as seen by the TEG module matches For any value of duty cycle, the corresponding operating point
that of the TEG source. A boost converter acts as an impedance is located on the straight line. Hence, the co-ordinates of points
matching device which shifts the operating point to MPP by C and D can be extrapolated using two operating points A and
B. The y-intercept thus obtained is the same as ISC of the TEG
varying the duty ratio as determined by the MPPT controller.
module. The gradient of the line can be calculated using the
The input voltage and input current of the boost converter are
fundamental line theory and is given as follows,
given in (5) and (6) as follows,
I 2 − I1
V TEG = V O × (1 − D ) (5) m1 = (13)
V2 − V1
IO
I TEG = (6)
(1 − D ) At D =D1, (8) can be rewritten to obtain (14) as follows,

where VO is the output voltage of the boost converter and IO is I 1 = m1 × V1 + I SC (14)


the output current of the converter.
From (14), the short-circuit current can be written as follows,
The equivalent input resistance of the converter as seen by
the TEG module can be obtained from (5) and (6) as follows, I SC = I 1 − m1 × V1 (15)

Rin = R L × (1 − D )2 (7) Under open-circuit condition, ITEG =0 and (8) can be


rewritten to obtain (16) as follows,
From (7), the equivalent input resistance seen by the TEG at
MPP condition can be written as RMPP = RL (1-DMPP) 2 which is m1 × VOC + I SC = 0 (16)
equal to the internal resistance Rint of the device, where DMPP is
From (16), OCV can be calculated as follows,
the optimal duty ratio generated by the MPPT controller.
I SC
B. Calculation of Open-Circuit Voltage (VOC) and Short- VOC = (17)
Circuit Current (ISC) through Extrapolation − m1
This section illustrates the determination of open-circuit Hence the VOC and ISC of the TEG module can be obtained
voltage and short-circuit current of TEG through extrapolation. using the linear characteristics of the device. Then the MPP
From (2), the current flowing through the device is given by, conditions can be derived using (4). In Fig. 2, line 1 and line 2
represent the reciprocal of resistance corresponding to duty
 −1  V
I TEG =   ×VTEG + OC
 (8) ratios D1 and D2 respectively, while the MPP line corresponds
 Rint  Rint to that of DMPP. Now, in order to get point E, DMPP needs to be
evaluated which is explained in the following section.
Equation (8) is a linear equation, which is in the slope-
intercept form y = m1x + b1, where the slope m1 = -1/Rint and
the y-intercept b1=VOC/Rint.
Similarly, the voltage across the TEG terminals is given by,
VTEG = (− Rint )× I TEG + VOC (9)

Equation (9) is also of the slope-intercept form y = m2x + b2,


where the slope m2 = -Rint and y-intercept b2=VOC. Fig.2 shows
the determination of open-circuit voltage and short-circuit
current from the linear I-V characteristics of TEG. Consider the
operating points A and B shown in Fig.2 which correspond to
duty ratio, voltage, current- D1, V1, I1 and D2, V2, I2
respectively. These two operating points are located on the I-V
line. The voltage across the input resistance can be written as
follows, Fig. 2 Determination of open-circuit voltage and short-circuit current from the
I-V characteristics of TEG
VTEG = R L × (1 − D )2 × I TEG (10) C. Calculation of Duty Ratio at MPP (DMPP)
From (10), the relationship between VTEG and ITEG for After finding the VOC and ISC, the DMPP can be derived from
different values of duty cycle D1 and D2 are given as follows, V1, I1, D1, VMPP and IMPP as follows.
At D =DMPP, the relationship between VTEG and ITEG can be
V1 = R L × (1 − D1 )2 × I1 (11) written as follows,
V 2 = R L × (1 − D 2 )2 × I 2
V MPP = R L × (1 − D MPP )2 × I MPP
(12)
(18)

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
4

On dividing (11) by (18), (19) can be obtained as follows, and current (I) values are noted and the power is calculated (step
7).
V1 (1 − D )2 × I1 If ∆V is not within the desired range ε1 about VMPP (step 8),
= (19)
V MPP (1 − D MPP )2 × I MPP the values of V1, I1 and D1 are updated to the values VMPP, IMPP
and DMPP respectively (step 9) and then the process is reiterated
On rearranging (19), DMPP can be calculated as follows, starting from step 5. If ∆V is within the desirable range ε1, then
MPP is reached. In case of a temperature variation, i.e. when |P-
V MPP × I 1 × (1 − D1 ) 2 Pprev| > ε2 (step 10), the entire process is repeated starting from
D MPP = 1 − (20) step 1. On the other hand, when |P-Pprev| ≤ ε2, the algorithm
I MPP × V1
returns to step 7.
In the proposed algorithm, the ISC and VOC of the TEG
system are calculated by using (15) and (17). Then, (20) is used
to calculate the optimal duty ratio whenever there is variation
in temperature gradient or load resistance. Equation (20) is easy
to solve and is independent of the load and circuit parameters.
Only input voltage and current information are required in this
proposed MPPT algorithm.
D. Operation of LEMPPT method under dynamically varying
temperature conditions
Fig. 3 illustrates the operation of LEMPPT algorithm under
a sudden change in temperature gradient from 140oC to 205oC.
The TEG array initially operates in load line 1 with a
temperature gradient of 140oC, at point A (MPP for ∆T =140oC)
as shown in Fig. 3. When the temperature gradient is suddenly
increased from 140oC to 205oC, the operating point moves from
A to B and the duty ratio remains unchanged for a while. Then,
the LEMPPT algorithm is re-initiated. Points C (V1, I1) and D
(V2,I2) are the operating points for the initial duty ratios D1 and
D2. Then, the algorithm computes the DMPP using (20). Finally,
the operating point of the TEG array converges to point E (MPP
for ∆T =205oC) swiftly, and the converging time is significantly
reduced.

Fig. 4 Flowchart for the LEMPPT algorithm


From Fig.5 (a) and (b), it can be observed that the P&O
algorithm takes a large number of steps to move to the vicinity
of MPP depending on the step size whereas the LEMPPT
algorithm takes just three to four steps. Hence, it can be
concluded that the proposed algorithm takes less time to attain
MPP when compared to P&O (and even other methods).
Fig. 3 Operation of LEMPPT algorithm during dynamic variation in Moreover, in P&O, the operating point continues to oscillate
temperature gradient from 140oC to 205oC about the MPP (between H &G) whereas, in the proposed
scheme, the operating point is fixed when it is within the
IV. PROPOSED MPPT ALGORITHM specified voltage range about MPP.
The flow chart of the LEMPPT algorithm is shown in Fig.4.
The proposed algorithm begins with two arbitrary duty cycle V. SIMULATIONS AND EXPERIMENTAL VALIDATION
values D1 and D2 and the corresponding values of voltage (V1
& V2) and current (I1 & I2) are noted (step 1 & 2). Using these A. System Description
values, ISC and VOC are calculated from (15) and (17) (step 3). In order to validate the tracking performance of the
VMPP and IMPP are in turn calculated using (4) (step 4). Then the LEMPPT algorithm, simulations and experiments are carried
optimal duty ratio (DMPP) is computed from (20) (step 5). By out in a series connected TEG array system consisting of seven
applying the obtained DMPP (step 6), corresponding voltage (V) modules as shown in Fig.6. The specifications of the modules
at Th=300oC and Tc=30oC are given in Table I [30].

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
5

The initial duty ratio for P&O is selected as Dinitial=9% and


the proposed method is initiated with values of D1=9% and D2
=10%.
C. Hardware Setup
The schematic diagram and the experimental setup of the
TEG system to validate the LEMPPT method are shown in
Fig.7 and Fig.8 respectively. The proposed system uses an
MSP430G2553 microcontroller and an Agilent E4360A
programmable dc power source to emulate the characteristics
of TEG-12708T237 module. The LEMPPT algorithm is tested
(a) (b) for series connected TEG array system under different
Fig.5 Movement of operating point (a) P &O (b) LEMPPT temperature gradient profiles using a programmable dc source.

Fig.6 Series connected TEG array

TABLE I
TEG SPECIFICATIONS

Parameters Symbol Value

TEG power PMPP 21.6 W


open-circuit voltage VOC 14.4 V
short-circuit current ISC 6.0 A Fig.7 Schematic diagram of the TEG system with MPPT controller
voltage at MPP VMPP 7.2 V
current at MPP IMPP 3.0 A
Seebeck coefficient αnp 0.0533 V/K
internal resistance Rint 2.39 Ω
B. Simulation Setup
A TEG array has been modeled [28] and connected to a DC-
DC boost converter whose specifications are L = 470µH, Cin =
Cout =100µF, fs = 20 kHz and RL = 25Ω. The sampling time (Ta)
for both P&O and proposed MPPT algorithm are taken as 0.02s
and the perturbation step size (∆D) for P&O is selected as 1% Fig.8 Snapshot of experimental setup
[31], [32].

(a)
(b)

(c) (d)

Fig.9 Simulated start-up tracking waveforms for ΔT= 205oC (a) Duty Cycle (b) Power (c) Voltage (d) Current

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
6

D. Results =25oC). From the simulated waveforms Fig. 11 (b), it can be


seen that the P&O algorithm takes 0.199s and the LEMPPT
1) Tracking on Start-up
method takes 0.059s to track the MPP at an efficiency of
To elucidate the tracking performance of the LEMPPT 99.98% and 100% respectively.
algorithm under start-up condition, the temperature gradient of
all the TEG modules is set at a value of 205oC. Fig.9 (a), (b),
(c) and (d) show the simulated waveforms of duty cycle, power,
voltage and current with respect to time respectively during
start-up condition.
The LEMPPT method calculates DMPP to be 21.59% and
tracks the MPP at 0.059s. In case of P&O, shown in Fig.9 (b),
the tracking time is found to be 0.2936s and the optimal duty
ratio oscillates between 20% and 22%. The overall tracking
efficiency of the LEMPPT and P&O algorithm is found to be
100% and 99.98% respectively.
The experimental start-up tracking curves for a temperature
gradient of 205oC are shown in Fig.10 (a) and (b). The P&O
algorithm tracks the MPP in 2s and has a ripple of 0.75% of (a)
average power. The LEMPPT method attains MPP in a time
span of 0.28s and it is clear that no oscillations exist
thenceforth. The voltage, current, power and tracking efficiency
at MPP are 36.39V, 2.422A, 88.14W and 99.85% respectively.

(b)

(a)

(c)

(b)
Fig.10 Experimental start-up tracking waveforms for a temperature gradient of
205oC (Power scale=38W/div) (a) P&O (Time scale=2s/div) (b) LEMPPT
(Time scale=140ms/div)

2) Dynamic Temperature Variation


To interpret the tracking performance of the LEMPPT
algorithm, two sequences with a sudden change in temperature
are considered. In either of the sequences, a step change in
temperature is introduced at 0.6s in the simulations.
a) Sequence-1(Increase in Temperature) (d)
Fig.11 Tracking waveforms for sequence-1 (a) P-V curves (b) Simulated
Fig. 11 (a) shows the P-V curves of TEG array for ∆T = waveform (c) Experimental waveforms for P&O (Time scale=2s/div)
140oC (Th =240oC, Tc =100oC) and ∆T = 205oC (Th =230oC, Tc (d) Experimental waveforms for LEMPPT (Time scale=140ms/div) (Power
scale=38W/div)

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
7

From the experimental waveforms shown in Fig.11 (c), it


can be observed that P&O takes 2.4s to track the MPP after a
dynamic change in temperature. From Fig. 11 (d), it can be
observed that the LEMPPT method initially tracks the MPP
wherein the voltage, current and power values are 25.27V,
1.37A and 34.6W respectively. Following a finite change in
power (∆P˃ε2), the algorithm is re-iterated and the new MPP
(35.104V, 2.5082A, 88.048W) is tracked within 0.28s. The
tracking efficiency of the LEMPPT method for a ∆T of 140oC
is 99.86% and that for 205oC is 99.75%.
(b) Sequence-2 (Decrease in Temperature)
In sequence-2, the temperature gradient is decreased from
175oC (Th=200oC, Tc =25oC) to 125oC (Th=225oC, Tc=100oC). (a)
The simulated and experimental tracking waveforms for
sequence-2 are shown in Fig.12 (a), (b), (c) and (d) respectively.
Fig.12 (a) shows the P-V characteristics of the TEG array for
∆T = 175oC and ∆T = 125oC. From Fig.12 (b), it can be seen
that the P&O method initially takes 0.333s and the LEMPPT
method takes 0.053s to track the MPP at an efficiency of
99.98% and 100% respectively for ∆T = 175oC. After t=0.6s
P&O slowly moves towards the new MPP for a duration of
0.213s. On the other hand, the LEMPPT algorithm takes just
0.054s. From Fig.12 (c), it can be seen that P&O tracks the MPP
after a sudden change in temperature gradient in about 4.25s
and oscillations can be observed in the steady-state. In Fig.12
(d), the LEMPPT method initially tracks the MPP for ∆T =
175oC with an efficiency of 99.90%. When it detects a sudden (b)
change in temperature, it shifts the operating point to a new
MPP (22.41V, 1.254A and 28.1W) with an efficiency of
99.86% in 0.262s.
3) Sudden Change in Load Resistance
To validate the tracking performance under a sudden change
in the load resistance, a change in load from 25Ω to 100Ω is
introduced at t=0.6s. During this period, the temperature
gradient is kept constant at 205oC. From the simulation results
shown in Fig.13 (a), P&O algorithm takes 0.294s while the
LEMPPT algorithm takes 0.054s to reach the MPP for the load
resistance of 25Ω. Following change in load, the P&O (c)
algorithm slowly pushes the operating point back to the MPP in
about 0.774s whereas the proposed scheme reiterates the search
and attains the MPP swiftly within 0.056s. The test
oscillographs for P&O and LEMPPT method are shown in
Fig.13 (b) and (c) respectively. When the load resistance is
suddenly changed to 100Ω, the P&O algorithm slowly climbs
to the MPP taking a time of 9s whereas the proposed method
takes just 0.357s. From the above results, it is evident that the
LEMPPT algorithm converges much faster than P&O scheme
under both start up and dynamic conditions. The steady-state
performance of the proposed method is observed to be superior
to P&O with zero oscillations. Also, the efficiency is higher
(d)
than the conventional P&O method. Hence, it can be concluded
that the tracking performance of the LEMPPT method is Fig.12 Tracking waveforms for sequence-2 (a) P-V curves (b) Simulated
waveform (c) Experimental waveforms for P&O (Time scale=2s/div) (d)
superior to P&O method considering all major aspects. Experimental waveforms for LEMPPT (Time scale=140ms/div) (Power
E. Comparison of Tracking Speed scale=38W/div)

The minimum and maximum internal resistances of a TEG- Correspondingly, the minimum and maximum duty ratios
12708T237 module are obtained as 1.67Ω and 2.94Ω are 0.09 and 0.32 respectively.
respectively [30].

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
8

quicker in the remaining region (B). Hence, it can be inferred


that the LEMPPT algorithm is faster than P&O in the majority
of the operating region. Further, it can be shown that the duty
cycle and power for MPP from the LEMPPT algorithm is
immune to measurement errors in voltage and current.

(a)

Fig. 14 Comparison of tracking speed

VI. CONCLUSION
This paper presents a new MPPT algorithm to extract
maximum power from a thermoelectric generator under any
dynamic condition. The idea underlying the LEMPPT
(b) algorithm is to calculate the open-circuit voltage based on linear
extrapolation of two measurable operating points without using
a trial and error control. Also, there is no need to interrupt the
power during the period of measurement of open-circuit
voltage. Thus, the optimal duty ratio is directly computed with
just a few tracking steps. The inherently linear nature of TEG
characteristics is the basis for LEMPPT algorithm which is
notably simple and easy to implement. In addition, the
LEMPPT method is free from steady-state oscillations and
hence, power loss is significantly reduced. Moreover, the
LEMPPT method does not require additional switches for
opening the TEG terminals in order to find the open-circuit
(c) voltage and can be easily implemented using a low-cost
Fig.13 Transient responses for sudden change in load resistance (a) Simulated
microcontroller. The theoretical concept of the proposed
waveform (b) Experimental waveforms for P&O (Time scale=5s/div)
(c) Experimental waveforms for LEMPPT (Time scale=140ms/div) (Power algorithm is explained and the tracking performance is
scale=50W/div) validated through simulation and test results. The simulation
The total time taken by the P&O algorithm to reach MPP can and test results substantiate that the tracking time of LEMPPT
be expressed as follows, method is only three sampling periods and is independent of the
changes in temperature and load conditions.
(D MPP − Dinitial )
(Ttrack ) P&O = × Ta (21)
∆D VII. REFERENCES
[1] Chuang Yu and K. T. Chau, “Thermoelectric automotive waste heat
The total time taken by the LEMPPT algorithm to reach energy recovery using maximum power point tracking,” Ener. Conv. and
MPP can be expressed as follows, Manag., vol. 50, no.6, pp. 1506–1512, Jun. 2009.
[2] I. Johnson and W.T. Choate, “Waste heat recovery: technology and
(Ttrack ) LEMPPT = 3T a (22) opportunities in U.S industry,” BCS, March, 2008 [Online].
Available:https://www1.eere.energy.gov/manufacturing/intensiveprocess
The variation of tracking time with respect to internal es/pdfs/waste_heat_recovery.pdf.
resistance for both algorithms is shown in Fig. 14. From (21) [3] J. Yang, “Potential applications of thermoelectric waste heat recovery in
the automotive industry,” in Proc. IEEE ICT, USA, 2005, pp. 170-174.
and (22), the value of DMPP for which both algorithms take equal [4] M. Brazdil and J. Pospisil, “Thermoelectric Power Generation Utilizing
time is 0.12 and the corresponding internal resistance is 19.36Ω the Waste Heat from a Biomass Boiler,” Journal of Electronic Materials,
(Point C). It can be seen that the LEMPPT algorithm takes 0.06s vol. 42, no. 7, pp. 2198-2202, Apr. 2013.
to track the MPP irrespective of the value of internal resistance, [5] T. Kyono, R. O. Suzuki, and K. Ono, “Conversion of unused heat energy
to electricity by means of thermoelectric generation in condenser,” IEEE
whereas with the P&O algorithm, the tracking time is Trans. on Energy Conv., vol. 18, no. 2, pp. 330-334, Jun. 2003.
proportional to the difference between DMPP and Dinitial. It can [6] G. Min and D. M. Rowe, “Conversion efficiency of thermoelectric
be verified that the LEMPPT algorithm is faster than P&O for combustion systems,” IEEE Trans. on Energy Conv., vol. 22, no. 2, pp.
86.28% (A) of the total operating region while P&O is a little 528-534, June 2007.

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2018.2830796, IEEE
Transactions on Energy Conversion
9

[7] K. Yazawa and A. hakouri, “Thermal optimization of embedded http://www.customthermoelectric.com/powergen/pdf/2411G-7L31-


thermoelectric generators in refractory furnaces,” in Proc. IEEE 15CX1_20140508_spec_sht.pdf.
ITHERM, Las Vegas NV, USA, 2016, pp. 1489-1493. [31] N. Femia, G. Petrone, G. Spagnuolo and M. Vitelli, “Optimization of
[8] D. M. Rowe, “Thermoelectrics an environmentally friendly source of perturb and observe maximum power point tracking method,”IEEE
electrical power,” Ren. Ener, vol. 16, no. 14, pp. 1251-1256, Jan. 1999. Trans. on Pow. Electr., vol. 20, no. 4, pp. 963-973, Jul. 2005.
[9] A. Montecucco, J. Siviter and A. R. Knox, “The effect of temperature [32] M. Killi and S. Samanta, “Modified perturb and observe MPPT algorithm
mismatch on thermoelectric generators electrically connected in series for drift avoidance in photovoltaic systems,” IEEE Trans. on Ind. Electr.,
and parallel,” Applied Energy, vol. 123, no. 15, pp. 47-54, Jun. 2014. vol. 62, no. 9, pp.5549-5559, Sep. 2015.
[10] K. M. Saqr and M. N. Musa, “Critical Review of thermoelectrics in
modern power generation applications,” Thermal Science, vol. 13, no. 3,
pp. 165-174, Jan. 2009.
VIII. BIOGRAPHIES
[11] T. Esram and P. L. Chapman, “Comparison of Photovoltaic Array B. Bijukumar (S’17) received the Diploma in
Maximum Power Point Tracking Techniques,” IEEE Trans. on Energy electrical engineering from Central Polytechnic,
Convers., vol. 22, no. 2, pp.439-449, June 2007. Thiruvananthapuram, India, in 2005, the B.Tech
[12] I. Laird, H. Lovatt, N. Savvides, D. Lu and V. G. Agelidis, “Comparative degree in electrical and electronics engineering from
study of maximum power point tracking algorithms for thermoelectric Government Engineering College, Idukki, India, in
generators,” in Proc. IEEE AUPEC, Sydney, Australia, 2008, pp. 1-8. 2008 and the Master’s degree in power electronics
[13] R. Y. Kim and J. S. Lai, “Seamless mode transfer maximum power point and drives from Anna University, Chennai, India, in
tracking controller for thermoelectric generator applications, IEEE Trans. 2012.
on Pow. Electr., vol. 23, no. 5, pp. 2310-2318, Sep. 2008. He is currently working toward the Ph.D. degree
[14] R. C. N. Pilawa-Podgurski, N. A. Pallo, W. R. Chan, D. J. Perreault and at National Institute of Technology, Tiruchirappalli,
I. L. Celanovic, “Low-power maximum power point tracker with digital India. His research interests are in the analysis, modeling and control of power
control for thermophotovoltaic generators” in Proc. IEEE APEC, Palm electronic circuits in renewable energy systems.
Springs, CA, USA, 2010, pp. 961-967.
[15] K. Bunthern, B. Long, G. Christophe, D. Bruno, M. Pascal, “Modeling
A. G. Kaushik Raam is currently working toward the
and tuning of MPPT controllers for a thermoelectric generator,” in Proc.
B.Tech degree in electrical and electronics
IEEE ICGE, Sfax Tunisia, 2014, pp. 220-226.
engineering at the National Institute of Technology,
[16] S. K. Kollimalla and M. K. Mishra, “A Novel adaptive P&O MPPT
Tiruchirappalli, India.
algorithm considering sudden changes in the irradiance,” IEEE Trans. on
His research interests include efficient design of
Energy Conv., vol. 29, no. 3, pp. 602-610, Sep. 2014.
power converters and application of power
[17] A. Pandey, N. Dasgupta, and A. K. Mukerjee, “High-performance
electronics in renewable energy systems.
algorithms for drift avoidance and fast tracking in solar MPPT system,”
IEEE Trans. on Energy Conv., vol. 23, no. 2, pp. 681-689, Jun. 2008.
[18] I. Laird and D. Dah-Chuan Lu, “High step-up dc/dc topology and MPPT
algorithm for use with a thermoelectric generator,” IEEE Trans. on Pow.
Electr., vol. 28, no. 7, pp. 3147-3157, Jul. 2013.
[19] A. Montecucco, J. Siviter and A. R. Knox, “Simple, fast and accurate G. Saravana Ilango (SM’17) received the Graduate
maximum power point tracking converter for thermoelectric generators,” degree electrical and electronics engineering from the
in Proc. IEEE ICCE, Raleigh, NC, USA, 2012, pp. 2777-2783. University of Madras, Chennai, India, in 2000, the
[20] A. Montecucco and A. R. Knox, “Maximum Power Point Tracking Master’s degree in power electronics and drives from
Converter Based on the Open-Circuit Voltage Method for Thermoelectric Bharathidasan University, Tiruchirappalli, India, in
Generators,” IEEE Trans. on Pow. Electr., vol. 30, no. 2, pp. 828-839, 2001, and the Ph.D. degree in electrical engineering
Feb. 2015. from the National Institute of Technology (NIT),
[21] J. Kim, and C. Kim, “DC–DC boost converter with variation-tolerant Tiruchirappalli, India, in 2009.
MPPT technique and efficient ZCS circuit for thermoelectric energy From 2001 to 2004, he was a Lecturer with the
harvesting applications,” IEEE Trans. on Pow. Electr., vol. 28, no. 8, pp. Noorul Islam College of Engineering, Kumaracoil,
3827-3833, Aug. 2013. India. In 2006, he joined the Department of Electrical and Electronics
[22] S. Carreon-Bautista, A. Eladawy, A. N. Mohieldin, and E. Sánchez- Engineering, National Institute of Technology Tiruchirappalli, India, where he
Sinencio “Boost converter with dynamic input impedance matching for is currently an Associate Professor. His research interests include FACTS
energy harvesting with multi-array thermoelectric generators,” IEEE controllers, digital controllers, and renewable energy systems.
Trans. on Ind. Electr., vol. 61, no. 10, pp.5345-5353, Oct. 2014.
[23] S. Nakayama, K. Kimura, Y. Kushino and H. Koizumi, “A simple MPPT
control method for thermoelectric energy harvesting,” in Proc. IEEE Nagamani Chilakapati (M’10–SM’16) received the
ECCE, Montreal, QC, Canada, 2015, pp. 6455-6460. M.Tech. degree in power electronics from the Indian
[24] Z. M. Dalala and Z. U. Zahid, “New MPPT algorithm based on indirect Institute of Technology, Kanpur, India, in 1984, and
open circuit voltage and short circuit current detection for thermoelectric the Ph.D. degree in electrical engineering from the
generators” in Proc. IEEE ECCE, Canada ,2015, pp. 1062-1067. University of Technology, Sydney, Australia, in 2001.
[25] Y. H. Liu, Y. H. Chiu, J. W. Huang and S.C. Wang, “A novel maximum From 1985 to 1991, she was with the Central
power point tracker for thermoelectric generation system,” Renewable Power Research Institute, Bangaluru, India.
Energy, vol. 97, pp. 306-318, June 2016. Subsequently, she joined the Department of Electrical
[26] M. Bond, and J. D. Park, “Current-sensorless power estimation and MPPT and Electronics Engineering, National Institute of
implementation for thermoelectric generators,” IEEE Trans. on Ind. Technology, Tiruchirappalli, India, where she is
Electronics, vol. 62, no. 9, pp. 5539-5548, Sep. 2015. currently a Professor. Her research interests include power electronics and
[27] A. G. Heaton, “Thermo-electrical engineering,” The Inst. of Elect. Eng., drives, renewable energy systems, and FACTS controllers.
vol. 109, no. 45, pp. 223-232, May 1962.
[28] S. Lineykin and S. Ben-Yaakov, “Modeling and Analysis of
Thermoelectric Modules,” IEEE Trans. on Ind. Appl., vol. 43, no. 2, pp.
505-512, Mar. 2007.
[29] M. Chen, L. A. Rosendahl, T. J. Condra, and J. K. Pedersen, “Numerical
modeling of thermoelectric generators with varing material properties in
a circuit simulator,” IEEE Trans. on Energy Conv., vol. 24, no. 1, pp. 112-
124, Mar. 2009.
[30] Custom Thermoelectric (2014). Datasheet of 2411G-7L31-
15CX1Thermoelectric Generator Module [Online]. Available:

0885-8969 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

You might also like