Design and experimental investigation of a low-voltage thermoelectric energy harvesting system for wireless sensor nodes

Energy Conversion and Management 138 (2017) 30–37

Contents lists available at ScienceDirect

Energy Conversion and Management journal homepage: www.elsevier.com/locate/enconman

Design and experimental investigation of a low-voltage thermoelectric energy harvesting system for wireless sensor nodes Mingjie Guan a, Kunpeng Wang a, Dazheng Xu a, Wei-Hsin Liao b,⇑ a b

School of Aerospace Engineering, Xiamen University, Xiamen, Fujian, China Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong, China

a r t i c l e

i n f o

Article history: Received 12 November 2016 Received in revised form 21 December 2016 Accepted 19 January 2017

Keywords: Thermoelectric energy conversion Low-voltage Energy harvesting Wireless sensor nodes

a b s t r a c t A thermoelectric energy harvesting system designed to harvest tens of microwatts to several milliwatts from low-voltage thermoelectric generators is presented in this paper. The proposed system is based-on a two-stage boost scheme with self-startup ability. A maximum power point tracking technique based on the open-circuit voltage is adopted in the boost converter for high efficiency. Experimental results indicate that the proposed system can harvest thermoelectric energy and run a microcontroller unit and a wireless sensor node under low input voltage and power with high efficiency. The harvest system and wireless sensor node can be self-powered with minimum thermoelectric open-circuit voltage as 62 mV and input power of 84 lW. With a self-startup scheme, the proposed system can self-start with a 20 mV input voltage. Low power designs are applied in the system to reduce the quiescent dissipation power. It results in better performance considering the conversion efficiency and self-startup ability compared to commercial boost systems used for thermal energy harvesting. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Thermoelectric generators (TEGs) are usually applied to convert thermal energy into electrical energy, provided that a temperature difference existed between the hot side and the cold side of the TEGs, exploiting the Seebeck effect. Applications of the thermoelectric generators have been explored on vehicle exhaust heat recovery, geothermal, power stations, and woodstoves [1–4]. Kempf and Zhang [1] investigated a high-temperature thermoelectric generator that converted engine exhaust waste heat into electricity and found that both the optimal TEG design and the fuel efficiency increase were highly dependent on the thermal conductivity of the heat exchanger material. Stevens [2] investigated ground-air thermoelectric generators operating between the air and ground temperatures. It was shown that a finned prototype produced power at an average rate of 1046 lW. Yazawa et al. [3] showed that thermoelectric generators could add 4–6% to the overall system efficiency for advanced supercritical steam turbines. Sornek et al. [4] exhibited the high potential of using thermoelectric generators to provide self-sufficient operation of stove fireplaces. Different designs of heat exchanger configurations and TEG modules were tested and compared. The maximum obtained

⇑ Corresponding author. E-mail address: [email protected] (W.-H. Liao). http://dx.doi.org/10.1016/j.enconman.2017.01.049 0196-8904/Ó 2017 Elsevier Ltd. All rights reserved.

power can reach 6 W. However, the relatively low efficiency of thermoelectric generator compared with typical conversion systems restricts their applications in large-scale [5]. Promising applications of the thermoelectric generators would like to be in small-scale applications for example the ‘‘self-powered” wireless sensor networks (WSNs). WSNs have found more and more applications in recent years. However, one issue in the applications of WSNs is the power supply for wireless sensor nodes. Due to the limited battery capacity, to replace a large number of batteries for WSN nodes is inconvenient and sometimes impractical. Energy harvesting technologies with ambient energy offer a solution to solve this problem. Various kinds of ambient energy sources have been considered for energy harvesting [6,7]. Challenges and potentials of the renewable power sources including piezoelectric, solar, thermoelectric, wind and RF energy were evaluated and it was demonstrated that renewable power sources were able to generate sufficient power for remote sensors [6]. Electromagnetic, kinetic, thermoelectric and airflowbased energy sources were identified as potential energy sources within buildings and the available energy was measured [7]. Within these ambient energy sources, thermoelectric energy is an attractive solution [8,9]. A thermoelectric energy harvester powered wireless sensor networks designed for building energy management applications was built and tested [8]. Their tests demonstrated that the proposed thermoelectric generator could

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37

effectively power WSN module when the prototype was placed on a typical wall-mount heater. Applications of thermoelectric generators for wearable wireless sensors were also shown [9]. Their thermoelectric harvesting systems provided power about 280 lW to operate wearable wireless sensors. The magnitude of the TEG’s open-circuit voltage is directly proportional to the temperature difference. It should be noted that, in natural convection environment, the temperature gradient cannot be kept at a significant level due to the homogenization process between the hot and cold junction. Therefore, in these applications of TEGs, the temperature difference is usually small and the output voltage can be in range of 20–400 mV. Therefore, boost converters are usually used to boost the voltage for use. Moreover, TEGs are often employed in environments with time-varying temperature differences. Therefore, it is necessary to control the converters with a maximum power point tracking (MPPT) algorithm to enhance the converter efficiency. Another important issue of the converter system is the ability to self-start. The system should be able to selfstart when the thermal gradient becomes high enough for harvesting energy. Designs of energy harvesting circuits from low-voltage TEGs have been investigated intensively in recent years [10–17]. Efficiency and self-startup ability are two main considered issues. Carlson et al. [10] presented a DC-DC converter that can boost a small input voltage of 20 mV to an output voltage of 1 V with high efficiency. However, a switched capacitor circuit should be used to generate a 600 mV voltage to start up the converter. Paraskevas and Koutroulis [11] proposed a MPPT method for TEG elements by controlling a power converter with the use of low-cost and off-the-shelf microelectronic components for high efficiency. However, the power consumption of the control unit they used was as high as 5.13 mW. Guan et al. [12] designed and investigated a boost converter for low voltage thermoelectric generator with MPPT scheme; however, the power consumption of the converter and microcontroller was still higher than 200 lW and the converter did not have the self-startup ability. Ramadass and Chandrakasan [13] tried to solve the start-up problem of the boost converter by applying a mechanically assisted startup circuit, which enabled operation of the boost system from an input voltage as low as 35 mV. However, the start-up mechanism will not work when there is no mechanical energy input. Im et al. [14] used a dual-mode converter to solve the start-up problem. The converter runs in transformer oscillation mode at low input voltage and in normal boost converter mode when input voltage exceeds a threshold value. However, the converter efficiencies are low with a peak efficiency of 61% and the maximum input voltage is constrained by 300 mV. Teh and Mok [15] also addressed the start-up problem in their boost converter. A self-start oscillation mechanism depending on the mutual coupling between the two identical transformer coils was applied. However, the efficiencies of their converters are low at low input voltages under 200 mV. There are some commercially available boost converters, which were designed for low-power energy harvesting systems. Texas Instruments released a low-power boost converter chip BQ25504 with battery management for energy harvesting [16]. However, it needs a minimum input voltage of 80 mV, which is relatively high for low-voltage TEGs. And the self-startup voltage of 330 mV is too high. LTC3108 from Linear Technology [17] can operate at very low input voltage (as low as 20 mV). However, it does not possess an MPPT technique and its efficiency is relatively low as no more than 40%. In order to power the wireless sensor nodes, there are some more considerations in the design of power management circuits. The operating voltage of the wireless sensor nodes (usually higher than 1.8 V) requires high conversion ratio boost systems from low voltage. Moreover, during transmit/receive mode, wireless

31

transmitter will consume a high current (more than 15 mA). The high conversion ratio and current requirements both pose challenges for the power management system. To improve efficiency at high conversion ratio, usually a multistage boost system is applied [8,18]. To supply a high current during transmitting, usually a supercapacitor or a rechargeable battery is applied as an energy storage element. In this paper, a low-voltage thermoelectric energy harvesting and management system for powering wireless sensor nodes is presented. The proposed system adopts a two-stage boost scheme and is able to self-start at a very low input voltage. The efficiency of the system is investigated experimentally. This paper is organized as follows. In Section 2, the overview of the proposed system and low-power design techniques employed for the proposed system are described. The experimental results are given in Section 3. Finally, conclusions and discussions are presented in Section 4.

2. Energy harvesting and management system The design and operation of the system including the converter, self-startup system and MCU are described in this section. 2.1. System diagram A block diagram of the proposed TEG generator system is shown in Fig. 1. The thermoelectric generator can be modeled as a voltage source VT in series with its internal resistance RT. The internal resistance RT is slightly changed depending on the temperature difference level [19]. There are basically two working modes in this system. One is the two-stage booster working mode which is the normal working mode. The other is the self-startup working mode. 2.2. Design and operation of the two-stage converter In the two-stage booster working mode, the energy from input capacitor Ci is transferred to a temporary storage capacitor Ctem with a boosted voltage. The second-stage boost converter is used to boost the voltage on the capacitor Ctem to a final energy storage device, usually a supercapacitor or a rechargeable battery. A linear regulator is used to regulate the voltage on the energy storage device to a working voltage for the MCU and WSN node. As the voltage on the energy storage device may be much higher than the needed working voltage for the MCU and WSN node, the linear regulator can reduce the unnecessary power consumption due to the higher voltage. As the BQ25504 chip can boost a low voltage (higher than 80 mV) to a battery voltage and possess the battery management functions, it is very suitable for working as a second stage converter. A high efficiency first-stage boost converter is needed to boost the ultra-low voltages from TEGs to a workable voltage Vtem for BQ25504, which is set as around 1 V in the proposed system. The topology of the first-stage boost converter is similar as in [12] and is shown in Fig. 2. Differently, a DCM working mode is applied to the first-stage boost converter to lower down the power consumption of the converter. The first-stage booster mainly includes controlled on-off switches G1, G2 and S2, and an inductor L. Switches G1, G2 and S2 are all based on n-channel metal–oxide–semiconductor fieldeffect (NMOS) switches. To implement the impedance matching technique, the converter is periodically disconnected from the TEG element by a switch S2 and the open-circuit voltage is measured during the disconnection period by the controller. The impedance matching scheme is implemented by keeping the voltage on the input capacitor Ci optimal.

32

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37

S1

RT

S2

1st-stage Converter

+ Vi

VT

2nd-stage Converter

+ Vtem

Ci

Ctem Linear Regulator I

MCU

TEG

Voltage Measurement

Wireless Sensor Node

Low-Voltage Starter

Threshold Control

Cls

+ Vls

+ Vesd

Linear Regulator II

Fig. 1. Energy harvesting system topology with WSN node.

vg2 ii RT

iL S2 + Vi

VT

RL

G2 L

Ci

+

ig2

vg1

ig1

item

Ctem

Vtem

2nd-stage converter

G1

Fig. 2. First-stage DC-DC converter in the system.

The operating waveforms of the first-stage converter in DCM mode are shown in Fig. 3. As shown in Fig. 3, in the internal t1, when switch G1 is turned on and switch G2 is turned off, the current increases with time approximately linearly, there will be

DiL V i  V RL  V on1 iL max ¼ ¼ t1 L t1

ð1Þ

where L is the inductance of the inductor; iLmax is the maximum current reached; Von1 is the voltage drop of the switch G1 when it is ON; VRL is the voltage drop from resistance of the inductor. In order to keep the voltage Vi across the capacitor Ci at an optimal value, the average current iL should be equal to the average current ii from the TEG, where

vg1

ii ¼ ðV T  V i Þ=RT 0

t

vg2

The voltage Vi is set as one half of the open-circuit voltage VT, as applied in [10–14]. Assuming

t3 ¼ ðb  1Þðt 1 þ t 2 Þ; 0

t

ig1

t

ig2 0

t iLm

0

2bLii V i  V RL  V on1

t1

t2

t3

Fig. 3. Waveforms of the voltages and currents in the converter.

ð5Þ

From (3) and (5), there will be

t2 ¼

t

ð4Þ

In the interval t2, the current iL will decrease approximately linearly from iLmax to zero. There will be

DiL V i  V RL  ðV on2 þ V tem Þ iL max ¼ ¼ L t2 t2

iL

ð3Þ

there will be

t1 ¼

0

ð2Þ

2bLii V tem þ V on2 þ V RL  V i

ð6Þ

In a high conversion ratio boost converter, t2 is much smaller than t1. From Eqs. (3)–(6), by sensing the open-circuit voltage VT, the controller is able to compute optimal switch ON-time and OFF-time of the switches G1 and G2 for impedance matching scheme.

33

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37

In the proposed system, the WSN node used is the ez430RF2500 board from Texas Instruments. The MCU applied on the board is MSP430. The minimum working voltage of the MCU and the WSN node is 1.8 V. Considering the high consumed current during transmit/receive mode of WSN node, a higher voltage of 2.3 V is used as the working voltage Vcc to ensure safe working. In the proposed system, a rechargeable battery of 3.6 V (Model LIR2032, OAHE Co., 40 mAH) is used to store the output energy from the second-stage converter. A linear regulator TPS78223 is applied to regulate the voltage on the rechargeable battery to a voltage of 2.3 V. The circuits of the second-stage DC-DC converter and linear regulator I are shown in Fig. 4

Threshold Control From LowVoltage Starter

Linear Regulator II G3

TPS78223

C6

R0

OUT

C7

Cls G4

G5

R1

R2 G6 R15

2.3. Design and operation of the self-startup system As shown in Fig. 1, in the self-startup working mode, a lowvoltage starter is used to accumulate the input energy on the capacitor Cls. As mentioned in Section 1, LTC3108 from Linear Technology can boost a low input voltage to a high voltage of 4.5 V. In the proposed system, a ultra-low voltage boost circuit based on LTC3108 chip is applied as the low-voltage starter [17], using an external 1:100 ratio transformer (Modeled LPR6235-752SML). Although the efficiency of LTC3108 is not high, it only works during the self-startup process and will not affect the efficiency of the whole system because it is very short compared with normal working mode. The low-power design is very important for the whole system. Design choices are made to reduce the quiescent power dissipation. The design of the threshold control circuit and linear regulator II in self-startup process is shown in Fig. 5. The threshold control is designed to disconnect the low-voltage starter from the other circuit part when the voltage on the capacitor Cls is lower than the threshold voltage. When the voltage on the capacitor Cls reaches the threshold, the switches G3, G4 and G6 will be turned on and the linear regulator II (based on TPS78223 chip) will supply a voltage to start the MCU.

P4.6 of MCU

Vcc

Fig. 5. Threshold control circuit and linear regulator II.

The design of the switch S1 and its controlling circuit is shown in Fig. 6. A depletion mode NMOS is used as the switch S1. The switch is connected to a negative voltage generating circuit controlled by MCU. Before self-startup, the MCU is out of work and the I/O port P4.2 will be at zero voltage and the input current from TEG will flow to the low-voltage starter. After self-startup, the I/O port P4.2 will output a PWM voltage to generate a negative voltage for S1, and switch S1 will switch to the two-stage boost converter. When the input power from the TEG is harvested by the twostage boost converter, the energy is first stored on a capacitor Cstor of the second-stage converter. After a certain period of time, the voltage on the capacitor Cstor is high enough to support the normal working of the MCU. Then MCU will send out a high voltage to I/O Port P4.6 to enable linear regulator I and disable linear regulator II. The working voltage Vcc will be supplied by the linear regulator I

Linear Regulator I L1

Vstor +

Cstor

C13

Vbat + C14

C15

R12

R13

+ Vtem

R11

R7

R10

R8

Ctem

R9

C11

R14 R3

R4

R5 R6

Fig. 4. Second-stage DC-DC converter and linear regulator I.

34

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37

To Low-Voltage Starter VT From TEG

S1 To Two-Stage Converter D1

P4.2 of MCU

enters the LPM3 mode. The LPM3 mode is interrupted every certain period and the MCU works in active mode with an 8 MHz internal clock frequency for a short time. During this time, the controller will measure the external open-circuit voltage, modify the settings t1, t2 and t3 and send signal through the wireless transmitter. The low power designs result in the average MCU power consumption of tens of microwatts during the operation. 3. Experiments and results

C10 D2

C16

Fig. 6. Switch S1 and the controlling circuit.

then. And switch G6 in Fig. 5 will be turned off to decrease the power loss. The diagram of the related voltage signals in self-startup procedure and switching procedure are shown in Fig. 7. In the diagram, Vls is the voltage stored on Cls; linear regulator I and linear regulator II are noted as LDO1 and LDO2, respectively; Vout-LDO2 is the output voltage of linear regulator II (LDO2); Vcc is the working voltage of MCU; and Vgs9 is the controlled voltage of switch S1. 2.4. Low-power designs in MCU Power consumption of system is mainly due to two parts: the power consumption of the converter and the MCU. The low power consumption in MCU is very important for the whole system. When the MCU works in active mode, it will consume a large current (about 2.7 mA). Therefore, the MCU should work in low power mode for most of the time. However, timers inside the MCU need to be working to generate PWM signals for the controlled switches and to interrupt the MCU from the low power mode. In order to lower down the average power consumption of the MCU, a low clock frequency for the timer is preferred. In the proposed system, the MCU works in Low-power-mode 3 (LPM3), LFXT1 (consumed current 900 nA) for sleep mode. Under this low power mode, the main clock (MCLK) and sub-main clock (SMCLK) are both disabled to reduce the power consumption of the MCU. Only the auxiliary clock (ACLK) is active for the timers inside the MCU, supplied by an external oscillator with a frequency of 32.768 kHz. Since temperature difference of the TEGs usually changes slowly, the MPPT scheme is performed periodically. The converter maintains settings t1, t2 and t3 (as shown in Fig. 3) when MCU

To verify the feasibility and evaluate the performance of the proposed energy harvesting and management system, a prototype system is built and experiments with the thermoelectric generator and a rechargeable battery are carried out. Experimental results are presented in this section. 3.1. Prototyping and experimental setup Based on the theoretical analysis and design considerations described in the previous section, components with appropriate parameters are chosen to build the whole system. The prototype system and experimental setup is shown in Fig. 8. The main components and circuit parameters for the proposed system are listed in Table 1. For the thermoelectric element, a commercial TEG modeled F40550 from Xinghe Electronics Co., China is used, as shown in Fig. 8, which has an internal resistance of 6.8 X and a Seebeck coefficient of about 25 mV/°C. The hot side of the TEG element is mounted on the heater (IKA C-MAG HP 7). The cold side of the TEG is clamped with a heat sink cooled by an open-loop cooling water system. To generate different open-circuit voltages of the TEG, the temperature difference between the hot and cold sides of the TEG can be varied by the heater and cooling water system. Two thermocouples are tightly attached to the hot and cold sides of the TEG by thermally conductive paste to measure temperatures. Hot side temperature, cold side temperature and the generated voltage are recorded by a data acquisition unit (Keysight, 34972A). During the experiments, the hot side temperature of the TEG ranges from 27.4–46.2 °C and the cold side temperature ranges

Heater

Heat sink

Thermocouple

Data acquisition unit

Temperatures Threshold voltage Supply 2.3V for MCU

TEG Voltage

Vout_LDO 2

In p V ut/O ol u ta tp ge ut s

TEG

Vls

WSN gate

LDO 2 Disabled Supplied by LDO 2

Wireless Transmitting

Vcc Supplied by LDO 1

Vgs9 of S1 Switch

Switch S1 off, TEG connects with self-startup scheme

S1

on

Switch S1 on, TEG connects with two-stage harvesting scheme

Fig. 7. Waveforms of related voltage signals in self-startup and switching process.

Energy harvesting system

MCU & WSN node WSN Monitoring

Fig. 8. Energy harvesting system prototype and experimental setup.

35

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37 Table 1 Main components and circuit parameters. Part number/parameter

G1, G2, S2 G3 G4, G5, G6 G7, G8, S1 D1, D2 L Rl Ci Cls Ctem Cstor Vtem

WNM4153 SI2301 BSS138 BSP149 1N4148 2.2 mH 1.9 X 100 lF 2000 lF 22 lF 100 lF 1.255 V

Vls

Voltages (V)

Component

4

0 3

Vout-LDO2 0 2

Vcc 0 0

Vgs9

-2 0

50

100

150

200 222

Time (s)

3.2. Self-startup procedure

(a) 4

Vls

Voltages (V)

from 24.8–31.1 °C. For example, when the hot side temperature is 46.2 °C and the cold side temperature is 31.1 °C (DT = 15.1 °C), the TEG element gives out an open-circuit voltage of 400 mV. On the other hand, when the hot side temperature is 27.4 °C and the cold side temperature is 24.8 °C (DT = 2.6 °C), the TEG element gives out an open-circuit voltage of 62 mV. The TEG output voltage is sent to the energy harvesting and management system. The energy harvesting system, the MCU and the WSN node are self-powered by the harvested energy. The WSN node has an embedded temperature sensor on the board. It senses the temperature and sends the temperature data to the WSN gate wirelessly. The temperature data from the WSN node are monitored on a laptop by a monitoring software.

0 3 LDO2 disabled Vout-LDO2

0 2

Switch to 2 stage harvesting Vgs9

-2 195

According to the self-startup strategy described in Section 2. When the voltage on the capacitor Cls reaches a preset threshold of 3.28 V, the linear regulator II output a 2.3 V voltage to start the MCU. In the experiment, when the input voltage is 62 mV, the related voltages during the self-startup procedure are recorded. The experimental voltages vls, vout-LDO2, Vcc and vgs9 are captured as shown in Fig. 9(a). The close-up view of waveforms during the switching procedure is shown in Fig. 9(b) for details. From Fig. 9, it is found that the voltage on the capacitor Cls takes about 195 s to rise from zero to the preset threshold voltage. Then, the linear regulator II (LDO2) outputs a 2.3 V for the working voltage of MCU Vcc. After a certain time period (about 0.3 s), the MCU outputs a PWM voltage on I/O port P4.2 to generate a negative voltage (about 1.6 V) for the controlled voltage Vgs9 of switch S1. Then the system is switched to the two-stage converter working mode and the input power is stored on the energy storage element device. About 0.75 s later, MCU sends out a high voltage to I/O Port P4.6 to enable linear regulator I and disable linear regulator II. Then Vcc is supplied by linear regulator I. 3.3. First-stage converter with MPPT technique The first-stage converter is working under DCM mode with MPPT technique. The open-circuit voltage is measured and optimal on–off control of the switches is made according to the opencircuit voltage. For every 25 millivolts of TEG open-circuit voltage from 50 to 500 mV, optimal parameters of t1, t2, and t3 are set based on Eqs. (3), (4), and (6) and stored in MCU to fulfill zero current switching (ZCS) in DCM mode. When the MCU is interrupted from the LPM3 mode, it will perform the MPPT scheme by adjusting the parameters of t1, t2, and t3 based on the measured opencircuit voltage. For example, when the open-circuit voltage of TEG is 108 mV, the input voltage of the first-stage converter is controlled to be 55 mV. The control voltages of switches in the

Supplied by LDO1

Supplied by LDO2

Vcc 0 0

196

197

Time (s)

(b) Fig. 9. Waveforms during the self-startup and switching procedure.

first-stage converter vg1 and vg2, the voltage vl2 on the terminal of inductor L connected to G2, are recorded and shown in Fig. 10(a). The close-up view of the waveforms is shown in Fig. 10(b) for details. From Fig. 10, it is found that the first-stage converter is working under DCM mode with t1 = 1404 ls, t2 = 30.5 ls, t3 = 855.5 ls, respectively. The switching period is 2.29 ms and the switching frequency is 436.7 Hz. From the vl2 waveform in Fig. 10(b), it is found that the inductor L is working under zero current switching mode. Zero current switching is beneficial to high system efficiency. If the parameters of t1, t2, and t3 are not set correctly, the inductor L may not work under zero current switching mode. For example, if the switch G2 is switched off too quick (t2 is shorter than the optimal value), shown as the third curve in Fig. 11, labeled as ‘‘Vl2- early off”, there will be a suddenly high voltage of vl2 due to the parasitic diode in switch G2, marked with the dashed circle. On the other hand, if the switch G2 is turned off too late, shown as the last curve in Fig. 11, labeled as ‘‘Vl2- late off”, the voltage vl2 will go under zero for a short time due to the parasitic diode in switch G1, marked with the dashed circle. The positive directions of the parasitic diodes in switches G1 and G2 are both from source to drain. 3.4. Converter efficiency and system efficiency The efficiency of the first-stage converter can be calculated by the output power divided by the input power. The experimental efficiency of the first-stage converter of TEG open-circuit voltage from 62 to 400 mV is shown as the triangle-marked curve in

36

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37

100 2

t1

Vg1

80

Efficiency (%)

0

Votages (V)

t3 2

Vg2 0 2

60 40

First-stage converter

20

Vl2 0

BQ25504 system 0 5

6

7

8

9

Time (ms)

Vg1

Voltage (V)

0 2

t2

Vg2 0 2

Vl2 0

6.86

6.88

6.9

6.92

6.94

6.96

Time (ms)

(b) Fig. 10. Waveforms of voltages

vg1, vg2, vl2 in the first stage converter.

2

Vg2 Voltages (V)

0 2

Vl2-ZCS 0 2

Vl2-early off 0 2

Vl2-late off 0 6.84

6.86

6.88

6.9

6.92

Time (ms) Fig. 11. Waveforms of voltages others.

vg2, vl2 to

0.2

0.3

0.4

Open-Circuit Voltage (V) Fig. 12. Efficiency versus open-circuit voltage.

(a) 2

0.1

voltage for the system with WSN node to be self-powered is 62 mV, and lowest input power to ‘‘self-power” the system with WSN node is 84 lW. When the open-circuit voltage is higher than 62 mV, the excess energy is harvested and stored. The system efficiency can be measured by the output power to the energy storage element divided by the input power from the TEG. The whole system efficiency with running WSN node at the set transmission cycle is measured and plotted in Fig. 12 as the starmarked curve. From the experimental results, it is shown that the system efficiency ranges from 44.2% to 75.4% in the 84– 400 mV open-circuit voltage range. It may be worth comparing the proposed system with the commercially available system. A converter system based on the BQ25504 chip is built for comparison. The experimental system efficiency based on the BQ25504 converter system towards the same open-circuit voltage range is measured, as shown in Fig. 12 in plus-marked curve. It is shown that the BQ25504 converter system can only harvest power from 108 mV open-circuit voltage. It is also shown that the proposed two-stage energy harvesting system efficiency is higher than the BQ25504 converter system in the considered voltage range. Considering the self-startup ability, the BQ25504 converter system can only self-start from input voltage higher than 330 mV, which is much higher than the lowest selfstartup voltage of the proposed system. The better efficiency performance of the developed system than BQ25504 converter in the considered voltage range may mainly come from two factors: First, for the considered low voltage working range (TEG open-circuit voltage of 62–400 mV), a higher value inductor and lower switching frequency in the first-stage converter is beneficial to higher efficiency as the decreased switching power loss due to the lower switching frequency is greater than the increased conductive loss due to the higher parasitic resistance of the higher value inductor. Second, the zero current switching technique by optimally setting the parameters of t1, t2, and t3 is implemented in the converter.

show the differences between ZCS and

Fig. 12. The efficiency ranges from 76.1% to 81% in the considering range. The two-stage energy harvesting system efficiency is related with the first-stage converter efficiency and second-stage converter efficiency. At the same time, the WSN node will consume the harvested energy. The consumed power relates with the transmission cycle of the WSN node. In the experiment, when the transmission cycle of WSN node is set as 3 min, the lowest open-circuit

4. Conclusions and discussions A low-voltage thermoelectric energy harvesting and management system with self-startup ability for wireless sensor nodes was presented in this paper. A high-efficiency two-stage energy harvesting scheme working under normal working mode and a self-startup scheme working during self-startup were combined in the energy harvesting system. DCM mode and MPPT technique were applied in the low-voltage converter for high efficiency. The MPPT technique intelligently adjusts the on/off times of the switches according to the open-circuit voltage of the TEGs. Expressions of the optimal switching on/off times of the first-stage con-

M. Guan et al. / Energy Conversion and Management 138 (2017) 30–37

verter were derived. Low-power designs in switching mechanism between two energy harvesting schemes were applied to reduce the quiescent power dissipation. Low power designs were applied to the MCU to reduce the power consumption of the MCU and enhance the whole system efficiency. Experimental results showed that the first-stage converter can achieve a high efficiency of 72.1–86.8% for TEG with open-circuit voltage range of 62–400 mV to 1.255 V. With a low-voltage starter, the energy harvesting system can self-start from a low input voltage as 20 mV. Experimental results showed that with a 6.8 X TEG and an input voltage of 62 mV, the self-startup scheme will take 196.05 s to switch to two-stage harvesting scheme. From the experimental results, when the WSN node sends signals every 3 min, the lowest open-circuit voltage for the energy harvesting system and WSN node to be self-powered is 62 mV, which is much lower than the BQ25504 converter. The whole system efficiency is much higher than the BQ25504 converter. Future work includes further improvement of the system efficiency and development of the energy harvesting system in a monolithic IC. An integrated system based on monolithic IC will be more efficient than a system built by discrete components. Acknowledgments The work described in this paper was supported by a grant from the China Science Foundation (Project No. 11202176) and a grant from the Innovation and Technology Commission of Hong Kong Special Administrative Region, China (Project No. ITS/248/14FP). References [1] Kempf N, Zhang Y. Design and optimization of automotive thermoelectric generators for maximum fuel efficiency improvement. Energy Convers Manage 2016;121:224–31. [2] Stevens JW. Performance factors for ground-air thermoelectric power generators. Energy Convers Manage 2013;68:114–23.

37

[3] Yazawa K, Hao M, Wu B, Silaen AK, Zhou CQ, Fisher TS, et al. Thermoelectric topping cycles for power plants to eliminate cooling water consumption. Energy Convers Manage 2014;84:244–52. [4] Sornek K, Filipowicz M, Rzepka K. The development of a thermoelectric power generator dedicated to stove-fireplaces with heat accumulation systems. Energy Convers Manage 2016;125:185–93. [5] Vining CB. An inconvenient truth about thermoelectric. Nat Mater 2009;8:83–5. [6] Dewan A, Ay SU, Karim MN, Beyenal H. Alternative power sources for remote sensors: a review. J Power Sources 2014;245:129–43. [7] Matiko JW, Grabham NJ, Beeby SP, Tudor MJ. Review of the application of energy harvesting in buildings. Meas Sci Technol 2014;25(1):012002. [8] Wang W, Cionca V, Wang N, Hayes M, O’Flynn B, O’Mathuna C. Thermoelectric energy harvesting for building energy management wireless sensor networks. Int J Distrib Sens Netw 2013;2013:761–4. [9] Thielen M, Sigrist L, Magno M, Hierold C, Benini L. Human body heat for powering wearable devices: from thermal energy to application. Energy Convers Manage 2017;131:44–54. [10] Carlson EJ, Strunz K, Otis BP. A 20 mV input boost converter with efficient digital control for thermoelectric energy harvesting. IEEE J Solid-State Circuits 2010;45(4):741–50. [11] Paraskevas A, Koutroulis E. A simple maximum power point tracker for thermoelectric generators. Energy Convers Manage 2016;108:355–65. [12] Guan M, Wang K, Zhu Q, Liao WH. A high efficiency boost converter with MPPT scheme for low voltage thermoelectric energy harvesting. J Electron Mater 2016;45(11):5514–20. [13] Ramadass YK, Chandrakasan AP. A battery-less thermoelectric energy harvesting interface circuit with 35 mV startup voltage. IEEE J Solid-State Circuits 2011;46(1):333–41. [14] Im JP, Wang SW, Ryu ST, Cho GH. A 40 mV transformer-reuse self-startup boost converter with MPPT control for thermoelectric energy harvesting. IEEE J Solid-State Circuits 2012;47(12):3055–67. [15] Teh YK, Mok KT. Design of transformer-based boost converter for high internal resistance energy harvesting sources with 21 mV self-startup voltage and 74% power efficiency. IEEE J Solid-State Circuits 2014;49(11):2694–704. [16] Texas Instruments, BQ25504 datasheet, 2012. . [17] Linear Technology, LTC 3108 datasheet, 2010. . [18] Ahmed KZ, Mukhopadhyay S. A 190 nA bias current 10 mV input multistage boost regulator with intermediate-node control to supply RF blocks in selfpowered wireless sensors. IEEE Trans Power Electron 2016;31(2):1322–33. [19] Mamur H, Ahiska R. Application of a DC–DC boost converter with maximum power point tracking for low power thermoelectric generators. Energy Convers Manage 2015;97:265–72.