Maximum-Power-Point-Tracking-Optimized Peltier Cell Energy Harvester for IoT Sensor Nodes

Abstract

This paper presents the development of an energization system prototype for IoT sensor nodes using Peltier cells as energy harvesters; its operation is optimized by applying a maximum power point tracking algorithm (MPPT) to capture as much electrical energy as possible, even if the cell temperature conditions have variations. In the IoT sensor node, a power management algorithm that works in accordance with the measurement and transmission operations can extend the node operating time, to obtain a greater amount of information and reducing the need for battery maintenance. The proposed methodology consists of developing an energization system, as well as the IoT sensor node. The energization system consists of a block of Peltier cells to obtain up to 4 V, a SEPIC-type DC-DC converter, and a 3.7 V lithium battery for energy storage. The converter works in a closed loop with the MPPT algorithm and delivers a voltage that guarantees the maximum power transfer to the battery. The sensor node was developed based on the ESP8266 development board, it allows data acquisition of temperature, humidity, light intensity, presence, and sound. The node transmits this information to the Ubidots platform for real-time visualization; to take advantage of its processing capacity, MPPT and energy management algorithms are also implemented. The results showed that to obtain a minimum voltage of 3.3 V in the energization system, a temperature difference of $59\pm 1$ °C between the plates of the Peltier cells is required. The MPPT algorithm allows working at the maximum power point and keeps the power delivered to the battery stable, with small transients when the information is transmitted; however, the overshoot and the settling time are reduced and do not affect the node operation.

1. Introduction

During the last few years, there has been a great advance in the development of wireless technology for information management and communications, so that, regardless of the type of application, a large amount of data can be acquired, processed, and transmitted, and be accessible from anywhere through an internet connection [1,2]. The use of this technology requires structuring a Wireless Sensor Network (WSN), which is composed of a group of devices with information acquisition and transmission capabilities, to a platform in the cloud through a wireless link, so that users can easily access the information, where the only requirement is to have a device with internet access, as shown in Figure 1 [3,4].

When it comes to remote area applications and where there is no access to power lines, such as monitoring environmental conditions in agriculture or protected areas, the nodes are powered by batteries, becoming a limitation on the operation time and reducing the sampling and storage capacity, as well as the transmission range. Therefore, on many occasions, nodes are considered single-life devices, unless the availability of a larger amount of energy is foreseen, as well as considering maintenance for battery replacement and the need to take into account the current environmental constraints on battery production and disposal [5,6,7]. This makes the installation and management of a WSN expensive, which is why it is necessary to include the use of self-sustainable energy sources in the design of nodes [8,9,10,11,12]. One way to solve this problem is with node power management algorithms, so that only the required power is supplied, depending on the operating mode; for example, de-energizing the transmission while data acquisition is being performed, or activating a sleep mode (low-consumption mode) when neither acquisition nor transmission is required [13,14]. This alternative can be combined with dynamic voltage control (DVC), taking into account that, in sleep mode, devices require a low operating voltage, so an energization system is used to deliver the required voltage in each function mode, reducing power consumption and increasing battery life [15,16,17]. An alternative that is being widely used and researched is energy harvesting from physical or environmental conditions to maintain the battery charge, making the node self-sustaining. There are alternatives based on kinetic energy such as piezoelectric harvesters, triboelectric harvesters, and electromagnetic harvesters. There are also alternatives that do not require movement, such as solar harvesters, high-frequency electromagnetic wave harvesters, and thermal harvesters, where the latter is mainly considered in agricultural and environment applications with tall vegetation, where the node is subjected to climatic conditions and it is not favorable to use a solar harvester [18,19,20,21]. Energy harvesting can deliver a small amount of power, and it is common to implement hybrid harvesting systems, as long as the conditions for their use exist, with the most popular combinations being a kinetic harvester with electromagnetic or thermal harvesters, and solar harvester with any any other type of harvester [22].

A thermal energy harvester is a solid-state device, where a cell consists of several junctions in series in a layer of type-N semiconductor material with a type-P layer. The connection is made by copper plates and coated with ceramic plates; in Figure 2, you can see a constructive scheme of a Peltier cell. When there is a temperature difference between the ceramic plates, which translates into a temperature difference between semiconductor junctions, there is a change in the voltage between the junctions according to Equation (1), where $E_j$ is the power difference between the junctions, $S_{AB}$ is the Seebeck coefficient that depends on the material, and $T_H$ and $T_C$ are the temperatures at each junction [23,24].

$$E_j=S_{AB}(T_H-T_C)$$

(1)

This device was proposed in 1909 by Edmund Altenkirch, when he discovered that semiconductors improve efficiency. The most commonly used semiconductors are SiGe, PbTe, and Bi₂Te₃, and their construction consists of hundreds of junctions in series to increase the generated voltage [22,25]. The work developed in [26] experimented with various Peltier cell connection configurations and the results showed that the power delivered by a Peltier-cell-based thermal harvester has a maximum point, as can be seen in Figure 3, which can be obtained with a certain value of voltage and current at its output, depending on the temperature difference between the junctions, which is consistent with the fact that it is a constant power device. It is clear that there is the possibility to optimize the Peltier cell power delivered.

Considering the power curve similarity with a photovoltaic panel, in order for a thermal harvester to work at the maximum power point, a maximum power point tracking (MPPT) algorithm can be used, which will act as an intelligent regulator, optimizing performance in the face of temperature variations. An MPPT algorithm involves forcing the output voltage to the value that aligns with the maximum power condition. An algorithm can be developed using traditional strategies such as perturb and observe (P&0), incremental conductance (IC), as well as modern strategies such as the use of intelligent optimization algorithms. Traditional strategies are suitable when there is one maximum peak in the power curve, while intelligent algorithms are suitable when the power curve has multiple peaks [27,28]. P&0 is a simple iterative algorithm and the most widely used in photovoltaic panels, which changes the output voltage and takes an output power sample to compare with previous data; if this is greater, then the actual data are assumed as the maximum. These actions are repeated until the actual power is lower than the previous power assumed as the maximum. Figure 4 shows the procedure in graphic form. A disadvantage of this strategy is that it does not give an exact solution and keeps oscillating around the maximum, and the oscillation amplitude and speed at which the maximum is found will depend on the perturbation voltage jump [29,30,31,32].

Thermal energy harvester application is a topic that has already been investigated: in [25], a Peltier cell power evaluation of electricity generators, by taking advantage of the temperature difference between a mining deposit and the ambient temperature, was performed. The authors determined that a suitable series and parallel Peltier cell combination was required to obtain 4 V, connected with a DC-DC converter and an electronic regulator to obtain 13 V, with 1 A at the output to charge a 12 V battery. However, the authors recommended that Peltier cells can be better harnessed to energize wireless sensors that require 1.3 V to 3.3 V. In the work [26], the simulation of a thermal energy harvester with MPPT for low-temperature-difference conditions was performed, an identification technique was used to model the Peltier cells, and a BOOST converter was proposed to implement an MPPT algorithm developed with a fractional open circuit voltage technique (FOCV), because this does require continuous power monitoring. The resulted showed that 309 uW with a temperature difference of 16 °C could be obtained. Ref. [24] proposed a thermal energy harvester for water consumption measurement. They evaluated a thermoelectric generator implementation together with commercial DC–DC converters and determined that the best energy harvester system was obtained by connecting the TES1-127060 Peltier cell (PN-Europe, Donauwörth, Germany) with the EM8900+EM8502 converter (EM Microelectronic, Marin-Epagnier, switzerland) to energize the water consumption meter with 3.3 V at a 270 °C temperature difference. The design of a Bluetooth low-energy beacon energized by Peltier cells was presented in [32]. It uses a TEC-12706 Peltier cell (PN-Europe, Donauwörth, Germany) directly connected to the beacon, showing that there was operability when the temperature difference was 200 °C and proposing to add a boost converter to regulate the voltage in case of temperature variations. However, only the design of the converter was simulated, and its operation was not implemented or evaluated. In [22], an IoT node was energized with a hybrid energy harvester, combining solar and thermal principles. The thermal part was implemented with a Peltier cell and the LTC3108 DC–DC converter (EM Microelectronic, Marin-Epagnier, switzerland) to stabilize the voltage with a 1 mF super capacitor. In the research developed in [33], Peltier cells were used to harvest energy from coal fires, and the research concentrated on the transfer heat system construction for the Peltier cells. The energization of IoT sensors with Peltier cells in agriculture applications was investigated in [7], and the work concentrated on the construction of the heat transfer system to take advantage of the temperature gradient between the ground and air when the Peltier cells are directly connected to the sensors. The results showed that 0.464 mW can be obtained with a temperature variation of up to 0.71 °C. Ref. [34] developed a multiparametric monitoring system for aircraft, energized by Peltier cells. It used the temperature difference between a heat storage unit and the exterior aircraft skin, obtaining 0.5 V with a temperature difference of 10 °C, connected to an energy management unit with a storage battery. In the research in [35], a long-range wireless real-time steam leak detector for industrial pipelines was energized with a thermoelectric generator to avoid the use of batteries, and a house and mechanical radiator were designed for temperature transfer from pipelines to Peltier cells, and these were connected to a boost converter to obtain 3.3 V and store the energy in a 5 F supercapacitor. The research concentrated on evaluating the transmitted data quality, instead of the harvested energy.

In this article, an energization system prototype for IoT sensor nodes using Peltier cells as energy harvesters is developed, and a SEPIC converter is used to optimize energy harvesting through the MPPT algorithm with the P&O strategy. The main contribution is the implementation of the whole system as a prototype and the evaluation of its performance; in addition, it is combined with an energy management algorithm in the IoT node to reduce energy consumption. This research was motivated by the shape of the Peltier cells power vs. voltage curve, which is very similar to that obtained in photovoltaic panels. Thus, using an MPPT algorithm to optimize its efficiency seems to be a good option for this energy harvester, and the results will allow establishing a practical alternative for IoT systems and remote sensors, contributing with a scientific and practical application that expands the field of use of the MPPT algorithm, and taking into account that no works have been found in which this optimization was used.

This article is structured in four sections, the first exposes the theoretical foundations, as well as a review of the state of the art on the use of Peltier cells as energy harvesters. Section 2 details the design of the prototype used to evaluate the MPPT algorithm in a Peltier cell. Section 3 presents the tests performed on the prototype, the results of which are analyzed to validate its operation, and finally, Section 4 summarizes the main findings in the conclusions.

2. Materials and Methods

To evaluate a Peltier cell energization system with MPPT for an IoT sensor node, an electronic prototype was developed according to Figure 5. The energy harvester was designed with a Peltier cell combination to obtain enough voltage to charge the battery; the power block acts as an interface between the energy harvester and storage block, being a power converter in a closed loop with an MPPT algorithm to establish the maximum power transfer with temperature variations. The storage block was constructed with a lithium battery, to take advantage of their fast charging and energy density, making them suitable for storing the harvested energy. The IoT node was integrated using a WiFi transmitter, sensors, and a controller in charge of data acquisition, which also executes control actions for the power block and is in charge of data transmission via WiFi to the Ubidots platform, where a graphical interface was developed for data visualization.

2.1. Energy Harvester

The energy harvester was implemented based on the TEC12706 Peltier cell (PN-Europe, Donauwörth, Germany), which, according to the datasheet, has a nominal voltage of 12 V and a nominal current of 6 A. However, these specifications are given for a heating–cooling application. With this precedent, its use as a voltage generator was evaluated at an average temperature difference of 50 °C, obtaining approximately 0.14 V. The connection of 14 devices in series was necessary to obtain 2 V, and a larger number of cells was not used, to avoid increasing the size of the harvester, which would make it difficult to ensure the same temperature in each cell. The cells were connected in an aluminum plate of 32 cm × 14 cm distributed as shown in Figure 6; the distribution was obtained experimentally, looking for a uniform temperature distribution and each Peltier cell generated the same voltage.

2.2. Power Block

The main power block requirement is the ability to obtain higher or lower output voltages compared with input voltage such as the Peltier cell voltage, so that the converter works at the operating point defined by the MPPT algorithm. Topologies that are only elevating or only reducing such as the Buck and Boost were discarded, and converter topologies such as Buck-Boot, Sepic, Cuk, or Zeta were considered as appropriate. Among these, the Buck-Boost and Cuk were discarded because the power switch is subjected to high voltage levels and reverses the polarity of the output voltage, complicating the integration with the load and control circuits. The Zeta converter is more efficient when acting as a step-down and, due to the current loop at the input, requires a more complex control, while the Sepic is efficient in a wide range of voltage variations and, thanks to the coupling capacitor C2, has isolation against transient faults of the source or load, plus the diode arrangement provides protection against short circuits in the load. With this background, a power block was developed with a SEPIC type DC–DC converter, whose topology can be seen in the Figure 7. This converter allows obtaining 4 V at the output, as well as working in conditions of variable temperature difference, due to its characteristic of an elevator–reducer.

SEPIC converter design equations for each component were obtained through steady-state analysis [36], which allowed obtaining voltage and current waveforms for each element from the analysis of the equivalent circuit in close and open conditions of $S1$ and $D1$, as shown in Figure 8. With an analysis of the load stage for each element and considering the energy balance for the converter continuous-conduction condition, where the average current of the inductor as well as the average voltage on the capacitor were zero, Equations (2)–(6) were derived. The converter operating conditions defined the design specifications shown in

Table 1. SEPIC converter design conditions.

Parameter	Symbol	Value
Input voltage	$V_{in}$	$2.0$ V
Output voltage	$V_o$	$4.0$
Output voltage ripple	$\Delta V_o$	10 mV
Inductors current ripple	$\Delta i_L$	$5\%$
Switching frequency	$f_{sw}$	10 KHz

. A toroidal inductor of 7 mH was selected for $L1$ and $L2$, and capacitors of 470 $\mathsf{\mu }$F 470 $\mathsf{\mu }$F were selected for $C1$ and $C2$, respectively, these values guaranteed the operation in continuous current condition.

$$V_o={\displaystyle \frac{V_{1}k}{1-k}}$$

(2)

$$L1\ge {\displaystyle \frac{V_{1}k}{\Delta i_{L1}{f}_{sw}}}$$

(3)

$$L2\ge {\displaystyle \frac{V_{1}k}{\Delta i_{L2}{f}_{sw}}}$$

(4)

$$C1\ge {\displaystyle \frac{i_{o}k}{\Delta V_{C1}{f}_{sw}}}$$

(5)

$$C2\ge {\displaystyle \frac{i_{o}k}{\Delta V_{C2}{f}_{sw}}}$$

(6)

The SEPIC converter works in a closed loop with power delivered as a feedback signal, which is calculated with the converter output voltage and current. The power reference is the power value considered the maximum and the error signal enters the MPPT algorithm to determine the duty cycle to deliver the maximum power; Figure 9 shows a scheme of the closed loop, which was developed digitally on the ESP8266 card. The MPPT was developed with a P&O strategy; This consisted of calculating the output power ($Po$) by means of voltage and current sensor readings ($Vo,Io$) for a given duty cycle ($\delta$); then, Po is compared with the preset maximum power ($Pmax$), and if it is higher, $\delta$ is increased with the jump (${\delta }_j$), otherwise $\delta$ is reduced with ${\delta }_j$ and the process is repeated from the start, the flowchart of the implemented program can be seen in the Figure 10.

2.3. IoT Node

This block required a data acquisition and transmission system for the variables commonly required in environmental studies, meteorological studies, and industrial studies, and with the capacity for transmission via WiFi, prioritizing the selection of devices with the capacity to work in low-power-consumption mode. The node was developed with a temperature and humidity sensor DHT11 (Adafruit, New York, NY, USA), a presence sensor HC-SR501 (Adafruit, New York, NY, USA), and a sound sensor Ky-038 (Adafruit, New York, NY, USA), and for the illumination level, a photoresistor was used. These sensors were chosen taking into account their accuracy, measurement range, and communication interface, so that the measurements are comparable with the parameters established by the national institute of meteorology and hydrology. The sensors were connected to an ESP8266 development board (Espressif Systems, Shanghai, China), selected for the availability of analog and communication ports for all sensors, as well as the availability of an integrated WiFi transmitter; a schematic of the components and their connections is shown in Figure 11.

The node operation includes an energy management algorithm based on a timer, which determines the acquisition time, transmission time, and low-power time; the timer generates an interruption and executes a subroutine that performs the respective actions, energizing only the devices that will operate in each mode and activating the low-power mode for other devices.

2.4. Real-Time Visualization

The Iot node sends the information via WiFi to the Ubidots platform, where an interface is implemented to visualize the data in real time using numerical indicators and historical data curves, as shown in the Figure 12, so that the information can be viewed through any device with internet access.

3. Results

3.1. Temperature Test

This test was to validate the MPPT algorithm operation in the power system, and the tests were performed in two stages. The first stage of testing was performed without the MPPT algorithm, it can be said that the SEPIC converter worked in open-loop. The second stage of testing was performed with the algorithm or in closed-loop. Both stages were performed under controlled conditions for the plate temperature difference of the energy harvester, and for this purpose a thermal resistor energized by an auto-transformer with a commercial on–off controller for electric oven, to maintain the temperature in a 5% variation range, was connected to the hot plate, while the cold plate was kept at ambient temperature, isolated from the heat of the resistor, as shown in Figure 13. The IoT node was not energized, to avoid damage due to voltage variation; instead, a 100 m$\Omega$ resistor was used.

3.1.1. Open-Loop Test

This test allowed determining the existence of the maximum power point at different temperature variations between plates, and the working temperatures were 38–40 °C, 58–60 °C and 78–81 °C; intervals in which it was possible to keep the temperature difference stable during the tests. For each temperature interval, a manual variation in the SEPIC converter duty cycle was performed and the output power data were registered. This information allowed validation of the operating point determined by the MPPT algorithm. The results obtained can be seen in Figure 14.

Figure 12 shows that there was a maximum power point for a given duty cycle value in the SEPIC converter, and it was noticeable that, after the maximum, the power decreased rapidly, and this coincided with the SEPIC converter behavior, where the output voltage growth was exponential as the duty cycle increased. Figure 15 shows the maximum power values and the duty cycle at which the maximum occurred for each temperature interval. It can be seen that there was an increase in the maximum power when the temperature difference was greater, while the duty cycle at which the maximum occurred decreased, and both behaviors appeared to be linearly related. Another interesting observation was that the SEPIC’s maximum power point was in the region where it worked as a reductor.

3.1.2. Closed-Loop Test

This test was performed to validate the MPPT algorithm operation. The test was performed under the same conditions as the open-loop test, to ensure that the results were comparable. Figure 16 shows the power curves obtained. The action of the algorithm is notable, because the power went up until it reached a maximum value, then it remained relatively constant. The closed-loop system had a maximum overshoot of 25%, and it had a reasonable settling time and a maximum steady-state position error of 20%, taking into account that temperature is a slow process. The variations reflected the oscillating characteristic of the algorithm, because they stayed around the maximum.

It can be seen that the power curve had the shape of an underdamped system, with overshoot and settling time. Considering that the algorithm had to find the maximum power point, the steady-state value difference with maximum peak indicated the existence of a position error, and it was equal to the overshoot; the values that reflect this behavior are shown in

Table 2. Closed-loop test important values.

Temperature Range	MP [W]	tss [s]	Maximum Stable Power [W]	Duty Cycle	Vo [V]
38–40 °C	0.04	120	0.21	0.3	2.23
58–60 °C	0.03	125	0.33	0.25	3.51
78–81 °C	0.035	80	0.36	0.23	3.85

.

The tabulated data demonstrate the effective performance of the MPPT algorithm; although there was a maximum error of 25%, the error was produced by the steady-state position error of the closed-loop converter with the MPPT algorithm, and this could be reduced with the inclusion of a controller in future works. Regarding the steady-state output voltage, it was determined that to charge a 3.3 V battery, the temperature difference had to be between 58 °C and 60 °C.

3.2. Battery Charge Test

This test evaluated the energy harvester operation to charge the battery while the IoT node was energized. It is necessary to clarify that the node had a voltage regulator to avoid overvoltages due to the operation of the MPPT algorithm. For this test, the IoT node worked by acquiring sensor data and transmitting them in 5 min intervals. Then, all devices were put in low-power-consumption mode. Two conditions were evaluated: one in which the battery was discharged, and one in which the battery was charged. Due to the fact that the current was variable while charging and discharging the battery, the performance of the harvester with the MPPT algorithm was analyzed through the SEPIC converter output voltage, and the results are shown in Figure 17 and Figure 18.

In Figure 17, it can be noticed that the converter delivered 3.7 V and remained constant while the IoT node was in low-power mode; after 5 min, there was a voltage drop that coincided with node data transmission. However, the algorithm responded in a short time, returning to deliver the adequate voltage, and this demonstrated the operation of the harvester with the MPPT algorithm, delivering the maximum power available. Figure 18 shows a similar behavior, with the difference that the output voltage was 8.5 V, which is reasonable, because the battery did not receive much current and the algorithm increased the voltage to work at the maximum power point. Additionally, there was an overshoot, which shows the behavior of the MPPT algorithm looking for the maximum. As in the previous case, there was a voltage drop when the node transmitted data, but the algorithm responded quickly.

According to the behavior in Figure 17, the MPPT kept the output voltage constant, with 0.27 in duty cycle, and this suggests that a constant duty cycle, without the need to run the MPPT algorithm, could have similar results. This alternative was evaluated under the conditions of the previous tests, and the results in Figure 19 were obtained.

The results show that voltage did not remain constant; instead, it increased in each interval, and this may have been due to the behavior of the battery current while charging. The experimental results demonstrated that this mode of operation is impractical and validated the operation of the MPPT algorithm, as it is responsible for maintaining the maximum power point, optimizing the use of the available energy to charge the battery.

3.3. IoT Node Test

This test was performed to validate the IoT node operation; sensor data acquisition was tested, as well as the transmission to Ubidot’s real-time interface. The node was configured to acquire and transmit data at 5 min intervals and the Ubidots information was monitored by accessing from a computer. Figure 20 shows the real-time display interface results. It is observed that the IoT node was working correctly, the data transmitted were valid, and there was no evidence of a loss of information.

The results obtained cannot be compared with related works, because even though they used Peltier cells in the power stage, it was not evaluated, and related work focused on the prototype functionality with respect to the application and left the operation of the power system in the background.

4. Conclusions

The research results obtained indicate that it is possible to use Peltier cells as energy harvesters for low-power electronic devices, finding a specific application in IoT nodes in remote areas, and allowing node operations for long periods of time without battery maintenance or replacement. With the design of the temperature sensing structure, as well as the power electronics, it was possible to obtain enough energy to charge a lithium battery, as long as a temperature difference of $59\pm 1$ °C is achieved, and the energy harvester was able to recharge the energy consumed during the transmission of information in a short time.

The behavior of the energy harvester delivering power based on Peltier cells allows the application of optimization strategies such as the MPPT algorithm. For the case study, it was sufficient to implement the P&O strategy, since there were no multiple maximum power points and this allowed maximum detection with a maximum error of 25%. The MPPT algorithm allowed thermal energy harvester optimization, and it is ideal for battery charging rather than for direct energization, because temperature difference conditions could occur that would cause an overvoltage and damage node devices. Use of the MPPT algorithm can also help when there is not a proper temperature distribution in the Peltier cells, because even if there are cells that deliver less energy, the algorithm works as a system and delivers the maximum power. A non-uniform temperature distribution may cause multiple maximum power points, in which case it would be necessary to use modern techniques such as heuristic algorithms.

This research focused on the evaluation of an energy harvester, and the temperature difference was achieved experimentally with help of specific devices. Future work will include the identification of applications where there is high temperature generation, such as steam pipelines or areas where the ambient temperature is hot enough to obtain usable energy.

The design of the harvester was based on several Peltier cell groupings, so optimization requires emphasis on the mounting structure design, which must ensure that the temperature is uniformly distributed, and thus that all cells deliver the same energy with a better use of temperature; it should also include heat sinks that can help to have a greater temperature difference.

The results obtained showed that the MPPT algorithm is applicable in energy harvesters based on Peltier cells, establishing an optimization strategy that allows taking full advantage of the power generated by the temperature gradient. This can be used as an alternative energization in IoT applications, agriculture, meteorology, and in general in remote measurement systems. This energization system can be optimized in the future by means of modern control strategies that reduce the position error, reaching the maximum power point in less time and reducing oscillations. All this with a mechanical design that allows a temperature gradient.