Thermoelectrical Energy Harvesting System: Modelling, Simulation and Implementation

Mihail Octavian Cernaianu, Cosmin Cirstea, Aurel Gontean

Applied Electronics Department “Politehnica” University of Timisoara

Timisoara, Romania mihail.cernaianu; cosmin.cirstea; aurel.gontean@etc.upt.ro

Abstract—This paper presents a thermoelectrical energy harvesting system SPICE simulation that is using thermoelectrical generators (TEGs). The authors propose an improved TEG thermoelectrical model that takes into account the parameter variations. A specialized boost DC-DC converter is used to store the harvested energy into a Li-Ion rechargeable battery. An autonomous maximum power point tracking (MPPT) system that is able to harvest the maximum energy delivered by the TEG is proposed and simulated.

Keywords - thermoelectric generator, energy harvesting, thermoelectrical model, MPPT, SPICE simulation.

I. INTRODUCTION In 1821 the German physicist Thomas Seebeck discovered

the thermoelectric effect that stands on the basis of nowadays TEGs, devices that convert heat to electricity.

Thermoelectric technology has been used since 1930’s as an alternative to batteries in devices such as gas operated radios and home generating stations [1]. Later on, radio receivers were powered by TEGs that were heated from kerosene lamps. As technology evolved, different implementations were made where TEGs were used to recover the heat generated by the exhaust pipes of trucks and cars [2].

TEGs are also used in devices called radioisotope thermoelectric generators (RTEGs) that obtain their power by converting the heat released by the radioactive decay into electricity through the Seebeck effect. These devices have been used to power satellites and space probes over time. One of the latest applications is powering the “Curiosity” Mars rover that was launched in 2011[1].

In [3], the authors present the usage of thermo-generators in combination with medical devices that are powered using the heat from the human skin.

Engineers are working on improving the performance of thermoelectric materials that can be evaluated through the

figure of merit Z, defined as: 2SZ

kσ= , where σ is the

electrical conductivity, S the Seebeck coefficient and k the thermal conductivity. A higher value for Z is an indicator of

higher efficiency, for a TEG that means higher output power at the same temperature gradient.

In this paper, the authors propose and simulate an energy harvesting system that recovers unused heat, converts it to electrical energy and stores it into a rechargeable battery. A specialized LTC 3105 DC-DC converter [4] is used to boost the TEGs output voltage to a desired level needed to charge the battery. A dynamic maximum power point tracking system that takes advantage of the DC-DC converter built-in MPPT feature is proposed and simulated.

II. THE PROPOSED ENERGY HARVESTING SYSTEM We have designed and built an energy harvesting system

that is based on TEGs. Its purpose is to gather unused heat, convert it to electrical energy and store it into Li-Ion rechargeable batteries. The energy harvesting system is comprised of two major units. The main unit contains the mechanical parts – two TEGs sandwiched between a copper plate which are mechanically attached to the heat power source and the cooling unit that employs a thermosyphon or a heat-pipe. The second unit contains the electronics necessary to convert the electrical power generated by the TEGs and store it into rechargeable batteries (Fig 1.).

Figure 1. General overview of the energy harvesting system

Fig. 2 presents the TEG’s main unit exploded view where

an electrical heater (power resistors) is used to emulate the heat power source.

The thermal insulation board separates the hot side of the TEG, blocking the heat flux from reaching the cold side. The hot side of the TEG must be separated from the cold side as good as possible in order to achieve the highest performance; otherwise the heat will reach the cold side of the TEG without passing through the internal semiconductors.

978-1-4673-1176-2/12/$31.00 ©2012 IEEE

Figure 2. Our TEG main unit exploded view

The electronics from the second unit is built around the

LTC 3105 Step-Up DC-DC converter. This device allows a low start-up voltage of only 250mV with adjustable output voltage from 1.5 to 5.25 V.

An important function of the LTC3105 is the built-in maximum power point tracking controller that enables the operation from low voltages and high impedance sources such as TEGs. This circuit allows the user to set the optimal input voltage operating point for a given power source [4], controlling the inductor current to maintain Vin at the voltage set on the MPPT pin.

The LTC DC-DC converter will be used to charge a 3.7V Li-Ion rechargeable battery. The battery model is reproduced as in [5] and is based on the work presented in [6], where the authors proposed a rapid test procedure that can be used to derive the parameters of a battery model.

The model, as presented in [5] was scaled down to a 3.6V Li-Ion cell. The temporal mismatch simulation problem (it takes hours to charge or discharge a real battery) was resolved by scaling the battery time by a factor of 3600 that changes hours into seconds. This battery model emphasizes the State of Charge (SOC) parameter. The model needs the battery open circuit voltage values in order to function. In our case, the voltage – SOC dependency was experimentally determined for a Li-Ion of 1000mAh capacity.

We have modified the model and scaled it to permit the battery to charge and discharge with 1C current, where C represents the battery capacity. A limiter was introduced in order to prevent charging the battery to a capacity above 100%.

III. THE PROPOSED DYNAMIC MPPT SYSTEM The LTC 3105 converter allows the user to control the

maximum harvested power from the device he uses (photovoltaic panels, TEGs, etc). An internal current source of 10μA, together with an externally connected resistor (RMPPC) generates a voltage that represents the minimum value the Vin pin can drop off to extract the maximum power from the energy source. For a photovoltaic panel, as its temperature increases, the generated voltage is reduced. In this way, a diode that is thermally coupled with the photovoltaic panel can be used as indicated in [4] to track the maximum power. As we experimentally determined in [7], in the TEG’s case, the output voltage increases linearly with temperature.

The authors propose two original solutions for the dynamic MPPT. In order not to perturb the internal current source of 10μA, an operational amplifier that sinks this current must be used.

The first solution (Fig 3.) is based on a digital approach and uses the microcontroller from the energy harvesting device. A lookup table is used to store the TEGs output voltage – temperature dependencies. Having this information, at any moment the maximum power extracted will occur when

2noLoad

inVV = , where Vin is the input voltage and VnoLoad is the

output voltage of the TEG when no load is connected to it. The microcontroller’s internal PWM is used to generate the desired MPPT pin voltage based on the current thermistor’s temperature and on the internal lookup table. The voltage is filtered and fed to a repeater that can sink the 10μA the MPPT pin is providing. The Rst represents a startup resistor that permits the LTC circuit to start working from the input voltage of 2 10Rst Aμ⋅ ⋅ . After startup, the microcontroller will be power by the LDO output of the LTC converter and the desired voltage value for the MPPT pin can be generated.

Figure 3. Proposed dynamic MPPT system that takes advantage of the

microcontroller

Figure 4. Proposed autonomous dynamic MPPT system that works

independently

The second solution has the advantage of being completely autonomous and is microcontroller free.

The proposed schematic is presented in Fig 4. The idea is to use the NTC thermistor and linearize its variation with R1 and R2 resistors. The necessary offset adjustment is performed by

modifying the gain of OP2. In order to determine the correct slope for the output voltage function of temperature, a NI LabView VI application (Fig. 5) was built to dynamically select the components values. The upper left corner graph from Fig. 5 presents the output voltage corresponding to the positive input pin of OP1 repeater (Fig. 4). In the lower left corner the thermistor’s characteristic function of temperature is presented. The graphic on the right side of Fig 5 shows the output voltage of OP2 which is the prescribed voltage for the MPPT pin. The voltage output variation is represented as a function of temperature. The offset and slope values can be adjusted from the resistors presented in Fig. 4 (R3, R4, R5, Rr). The resistors values are represented as slides in the VI application from Fig. 5. The values are modified until the voltage - temperature variation corresponds to the experimentally determined characteristic.

Figure 5. LabView VI front panel describing the voltage-temperature

characteristic from the output of OP2 operational amplifier

IV. SPICE IMPLEMENTATION The setup described in Fig 2, together with the topology

presented in Fig 4. are simulated by the authors in Linear Technology’s LTspice simulation software.

The TEG main unit simulation consists of two steps. In the first, using the thermal to electric analogies, we have calculated the thermal resistances and capacities of the mechanical parts (heat sinks, aluminum and copper plates) and expressed them as electrical components.

Our experiments show that if we consider the two sides’ mechanical parts as homogenous blocks, the errors can reach up to 50%. The way the Biot number indicates if it is sufficient to use the so called lumped capacitance method to represent the system is explained in [8]. In our system, due to the geometry of the mechanical parts, the spatial effect must be considered and a digital discretization must be made.

The mechanical parts have been split into 16 smaller parts and for each the thermal resistance and capacitance have been calculated taking into account the material, density and volume. These smaller blocks were then assembled in series and parallel to reconstruct the real mechanical blocks.

The thermal – electrical analogies are presented in [9]. The thermal watt is expressed with a current source, the temperature through a voltage source and the thermal resistance [/W] through a resistor.

The second step of the TEGs main unit simulation consists in modelling the TEG.

In [7 and 10] we have described the steps for measuring the internal parameters variation with temperature (Seebeck coefficient, internal resistance, thermal conductance). The parasitic elements – inductance and capacitance have also been investigated and modelled in our previous works [11].

Based on all these results, an improved thermoelectrical TEG model (that replicates with accuracy the functioning of a real Everrredtronics TEG127-40A within a ∆T temperature gradient of up to 30) has been built. The model is based on the previous work of Lineykin and Ben-Yakov [12].

The SPICE thermoelectrical TEG model that also employs the TEGs main unit mechanical parts thermal simulation is presented in Fig. 6.

Figure 6. TEG main unit model (thermal model of the mechanical parts and

TEG thermoelectrical)

The SPICE model for the DC-DC converter, the proposed dynamic MPPT adjustment and the battery model are presented in Fig. 7. The LTC’s SPICE model is offered by Linear Technologies on their website [4].

Figure 7. SPICE model for the dynamical MPPT, LTC converter and battery

For the autonomous dynamic MPPT adjustment, a 10KΩ

NTC thermistor was used and its variation with temperature was simulated using a lookup table.

V. RESULTS The necessary temperature values used in simulation for

heating the hot and cold sides of the TEGs main unit have been experimentally determined for 5000 seconds. These temperature values are the only input to the system. In simulation, the time has been scaled down 1000 times, 1 simulation second meaning 1000 seconds in reality.

In Fig. 8, the VMPPC voltage is set to follow the temperature only until the hot side temperature reaches 310K, after which the voltage remains constant. It can be seen that the output voltage remains at this value although the temperature increases and the TEG is capable of sourcing more power. This is also reflected in the output power, the current through D4 diode that charges the battery remains at a constant value.

Figure 8. SPICE simulation results for dynamic MPP tracking up to 310K

In Fig. 9, the output generated by the dynamic MPPT is set

to follow the temperature up to 328K. As a result, it can be seen that the output power increases as the input voltage is following the VMPPC setpoint voltage.

Figure 9. SPICE simulation results for dynamic MPP tracking up to 328K

VI. CONCLUSIONS In this paper we have built a SPICE model that simulates

the complete energy harvesting system we’ve proposed. The hot and cold side temperatures of the TEG’s main unit were prescribed and experimentally determined. Based on our previous results we have built an improved TEG thermoelectrical model that takes into account the parameters

temperature variations. We have discretized the mechanical system in order to increase the model’s accuracy. An LTC3105 DC-DC converter was used to harvest the energy supplied by the TEG and charge the Li-Ion battery. The LTC model was obtained from Linear Technology website and an existent battery model was modified to fit our needs. We have proposed two solutions for the dynamic maximum power point tracking system and simulated the autonomous analog one.

The model of the complete energy harvesting system was implemented in LTspice and the simulation results prove the correct behavior of the system. The proposed dynamic MPPT system proves effective in extracting the maximum power generated by the two TEGs. The simulation of the TEG’s main unit is according to the experiment. The electronics - LTC converter, along with the dynamic MPPT system and the battery follows to be experimentally implemented and the result compared with the current simulation.

ACKNOWLEDGMENT This work was partially supported by the strategic grant

POSDRU/88/1.5/S/50783, Project ID50783 (2009), co-financed by the European Social Fund – Investing in People, within the Sectoral Operational Programme Human Resources Development 2007-2013.

REFERENCES

[1] John McCoy, “Thermoelectric Technology: Materials, Processes, Devices & Systems”, Hi-Z Technology, San Diego, 2012.

[2] John W. Fairbanks, “Thermoelectric Developments for Vehicular Appliations”, Diesel Engine-Efficiency and Emissions Research Conference, Detroit, 2006.

[3] M. Lossec, B. Multon, H. Ben Ahmed, C. Goupil, “Thermoelectric generator placed on the human body: system modelling and energy conversion improvements”, The European Physical Journal Applied Physics 52, 2010.

[4] LTC 3105 Step-Up DC/DC Converter, Linear Technologies Datasheet, http://www.linear.com/product/LTC3105, 2011.

[5] Intusoft Newsletter, Personal Computer Circuit & System Design Tools, Issue 78, Nov. 2005.

[6] Suleiman Abu-Sharkh, Dennis Doerffel, “Rapid test and non-linear model characterisation of solid-state lithium-ion batteries”, Journal of Power Sources 130 (2004) 266–274, 2004.

[7] Mihail Cernaianu, Adrian Cernaianu, Cosmin Cirstea, Aurel Gontean, “Thermo Electrical Generator Improved Model”, International Conference on Power and Energy Systems - ICPES, Hong Kong, 2012.

[8] Jan L M Hensen, Abdullatif E Nakhi, “Fourier and Biot Numbers and the Accuracy of Conduction Modelling”, Proceedings of Bep 94 Conference “Facing the Future”, pp. 247-256, 1994.

[9] Alain Laprade, Scott Pearson, Stan Benczkowski, Gary Dolny, Frank Wheatley, “A New PSPICE Electro-Thermal Subcircuit For Power MOSFETs”, Fairchild Semiconductor Application Note 7534, July 2004.

[10] Mihail Cernianu, Aurel Gontean, “Thermoelectric Modules Thermal Conductance Measurement System”, to be published.

[11] Mihail Cernaianu, Aurel Gontean, “Parasitic Elements Modelling in Thermoelectric Modules”, to be published.

[12] Simon Lineykin, Sam Ben-Yaakov, “Modeling and analysis of Thermoelectric Modules”, IEEE Transactions on Industry Applications, Vol. 43, pp. 505-512, 2007.