1. Introduction and Related Works
Wireless sensor network (WSN) is increasingly being used to improve applications such as military surveillance, medicine, transportation, environmental conservation, agriculture, home health care, and industrial process control, among others. With the emergence of the Internet-of-Things (IoT), many more low-power sensors will be deployed in different environments to collect, process, analyze, and monitor environmental factors in real time. The power supply of sensor nodes becomes a challenge as sensor nodes are normally battery-powered [
1]. Batteries used include primary batteries (non-rechargeable batteries) and secondary batteries (rechargeable batteries). Primary batteries are typically used as the power source for many sensor networks, and the lifetime of these sensor nodes is the time taken to discharge below minimum charge level that a sensor node requires [
2]. Although batteries have high energy density, they have a limited lifetime. For long-term deployments, regular battery replacements are required [
3]. Battery replacement can be expensive and not feasible in a case like remote environmental monitoring [
4,
5]. The limited energy also determines some important performance parameters in a sensor network system, such as the sampling frequency and the maximum transmission distance between sensor nodes [
6]. With the sampling frequency and transmission power increasing, the sensor node will consume more energy. Three different ways are commonly used in energy-harvesting systems to provide the energy: a supercapacitor, a battery, and a combination of both [
7]. To achieve long-term operation, energy harvesting with secondary batteries (or supercapacitors) as an alternative storage element is essential.
Self-powered sensor systems, which are very active research topics, use various energy-harvesting methods, such as solar energy using solar cells [
8,
9,
10] and solar thermoelectric generators [
11,
12], mechanical energy [
13,
14], and wind energy [
15]. As provided in [
16], solar cells can provide the highest power density (15 mW/cm
2), which is a lot higher than vibration (200 µW/cm
2), thermoelectric (60 µW/cm
2), and solar thermoelectric (16 µW/cm
2) methods [
11]. The extraction of solar energy is difficult in nonstationary environments; for example, cloudy days, dust on the solar cell surface, or the varying illumination levels at different times at the day can cause the degradation of performance [
17]. Therefore, an energy storage/buffer unit and a solar controller are required. Harvested energy must be stored in an energy buffer. Some works use a supercapacitor as the primary power supply and a battery as the backup source, as studied in [
18]. Some others choose rechargeable batteries only. For example, the work in [
19] uses one 150 mAh NiMH rechargeable battery as the primary source and a 2200 mAh Li-ion battery as the backup source.
When transferring energy from harvesting sources to the energy buffer, some works use the maximum power point tracking (MPPT) technique [
8,
20] to keep the solar panel at the maximum power point. Without MPPT, it will become difficult when using a small-sized solar panel to charge the energy buffer. Some commercial chips like LTC3105 [
21] and bq25504 [
22] can easily support the MPPT function on board.
Supercapacitors have almost unlimited charging cycles, high power density, high charge–discharge efficiency (97%–98%), and they do not release any thermal heat during the discharging process, while rechargeable batteries have lower charging cycles and high energy density [
16,
18]. These outstanding features make supercapacitors a suitable primary source for sensor nodes in a sensor network system. Research in [
23] shows that different supercapacitors have voltage losses from 5%–60% over 2 weeks’ time, and as the temperature increases, the self-discharge rate increases as well.
A few existing wireless sensor networks use solar energy harvesting. For example, the work in [
6] uses a 54 mm × 42 mm solar cell, with a 2.3 V 50 F supercapacitor and rechargeable batteries together as the energy-harvesting unit. The rechargeable battery acts as the backup buffer. With this unit, the power system is enough to extend the operational life of the network to 1 year. They also indicate that having both supercapacitor and battery in the system will increase the feasibility of the network design architectures.
The work in [
8] presents an efficient, small-sized, low-cost MPPT circuit for a WSN. The solar energy is transferred from solar panels to batteries, and then the power is supplied to the sensor nodes. This solar-powered WSN monitors the temperature and luminosity information of the marine ecosystem at different depths. A solar-powered power management unit operating with battery packs is implemented to maximize the power efficiency and lifetime.
Work in [
3] presents a WSN for bridge structural health monitoring (SHM). The researchers installed solar panels (11.4 cm × 21.0 cm) and rechargeable batteries with 10,000 mAh capacity in eight nodes, where changing the battery is difficult. They found that for nodes with good sunlight, battery voltage can be well maintained at around 4.15 V to keep the system working; for a node with only indirect or reflected sunlight, the battery voltage keeps decreasing and the solar panel cannot charge the battery.
SensorScope [
19,
24] presents WSNs for monitoring harsh environments. Solar energy-harvesting has been integrated into each of the stations to power up the sensor nodes and enables the long-term deployment of the system. The system is able to last for 180 days.
Work in [
25] also uses solar energy harvesting, whereby it is only used in the gateway node because the gateway consumes more power compared to the sensor node. The gateway is placed inside a birdhouse that is quite useful. This paper also reports that under direct sunlight, it is hard to charge Li-ion rechargeable batteries with a charging temperature ranging from 0–45 °C. The gateway battery voltage keeps decreasing without charging once the open-air nature temperature increases. This is an important issue that needs to be taken into account while selecting the energy storage unit and choosing the casing. In the work in [
26], researchers use a solar radiation shield to protect the electronic components from direct sunlight exposure and rain, which may be useful in solving this issue.
Many works utilize solar panels that can generate relatively higher power: for instance, a 1.5 W solar panel is used in work [
9] and a 3 W solar panel is used in works [
10,
27]. The large output-power solar panels can charge rechargeable batteries more easily compared to the solar panel with only a few hundred milliwatts of output power. However, these solar panels normally require a larger area—more than 100 cm
2—and could be used together with the bigger-sized sensor nodes.
The work in [
10] deploys a solar-powered traffic-sensing network in urban and desert areas, which uses an XBee module as the communication module. As described before, the sensor node is powered by a 3 W solar panel, which is far enough for powering a 90 mW sensor node. However, the author finds that in practice, the sensor node ends up running out of energy quite fast due to dust accumulation on the solar cell and shadows from surrounding buildings. As the temperature gets high, the Li-ion batteries used in the node are also affected severely.
The systems mentioned above use large solar panels and do not study the performance of the sensor network with solar energy harvesting in detail. We have designed a solar energy-harvesting unit and integrated it with EN-Nets (a low-power environmental sensor network system developed at Monash University, Australia). The proposed energy-harvesting mechanism can successfully supply the required power to sensor nodes in real time, extending the lifetime of the WSN. Supercapacitors or rechargeable batteries are used as the storage unit to hold the required amount of harvested energy for continuous operation. The proposed technique provides an energy solution to keep sensor nodes active for a whole day. Each sensor node includes multiple sensors, one microcontroller, one XBee module, and one energy-harvesting unit. The system has been validated in both field and laboratory environmental conditions.
The remainder of this paper is organized as follows:
Section 2 outlines hardware implementation;
Section 3 discusses some results and evaluations, and a brief conclusion and possible future improvements are given in
Section 4.
2. Hardware Implementation
A simplified schematic diagram of the self-powered sensor node is shown in
Figure 1. The energy-harvesting system sources power from a small-sized solar panel and then provides the power to the power management unit in a sensor node. Power management includes one low drop-out (LDO, MCP1700 from Microchip Technology Inc., Chandler, AZ, USA) voltage regulator to regulate the sensor node voltage, and one microcontroller with three MOSFET (metal-oxide semiconductor field-effect transistor) switches to turn the different sensors on and off for a specific time to reduce power consumption (only two switches are shown in the figure because only two switches were used in this experiment). The DC–DC converter helps to keep the voltage constant and charges rechargeable batteries or supercapacitors. The wireless sensor node operates with a small duty cycle, D = 1.25% (a 15 s wake-up time in a 20 min period). Therefore, the aim of the energy-harvesting system is to provide sufficient power to run for at least 1 day once the supercapacitor has been charged to a certain capacity. The size of the harvesting prototype board is 3 cm × 3 cm and it can be placed inside the sensor box together with the sensor node and supercapacitors. A detailed description is provided below.
Figure 2 shows the sensor node with the solar panel.
2.1. Energy Harvesting Unit
A 55.0 mm × 67.5 mm × 3.2 mm solar panel (from Shenzhen Chuanningsheng Electronics Co., Ltd., Shenzhen, China) is used in the system. The solar panel features open voltage at 3 V and short-circuit current at 150 mA. The maximum power it can produce is around 310 mW with a voltage of 2.6 V. The characteristic diagrams are measured under direct sunlight, on a cloudy day, 1 h before sunset, or under heavily cloudy conditions using different load resistors, which are shown in
Figure 3. The current is measured by connecting a current-meter in a series with the load resistor. The voltage is measured as the voltage difference across the resistor.
LTC3105 is used in the circuit to track the maximum power point of the solar panel. This device contains a high-efficiency step-up DC–DC converter with a low start-up voltage (250 mV) that can support small-sized solar panels and features a maximum power point controller (MPPC) function. The maximum power point (MPP) is set by changing the resistance of the MPPC resistor to suit the solar panels’ MPP, which is around 2.5 V. The output voltage is also user selectable. In our design, the output is set at 4.65 V when charging the supercapacitor and at 4.1 V when charging the rechargeable battery.
Setting MPP makes the solar panel only work at maximum power point 2.5 V. Therefore, the solar panel works at around 300 mW under direct sunlight, 150 mW on a partially cloudy day, and around 23 mW 1 h before the sunset. During a heavily cloudy day, the output power of the solar panel is near 0 at 2.5 V.
Initially, our sensor nodes were designed to be powered by batteries. Therefore, a LDO voltage regulator was added to the sensor node to regulate the input voltage. It requires at least 3.6 V in order to keep the sensor node running at a nominal voltage of 3.3 V.
For supercapacitors, the energy is stored electrostatically and does not involve any chemical reactions. One drawback of a supercapacitor is that it has a higher self-discharge current, which is 0.28 mA for the one used in the work (HP-2R7-J107UY LR from Jinzhou Kaimei Power Co. Ltd. [
28]). While supercapacitors are normally restricted to low voltages ranging from 2.5 V to 2.7 V for a higher capacity, they must be connected in series to get higher voltages. We used two 100 F × 2.7 V supercapacitors in series to obtain 50 F capacitance with a maximum voltage of 5.4 V. For rechargeable batteries, 18650 rechargeable batteries (ICR18650-26F from Samsung SDI Co., Ltd., Yongin, South Korea) were chosen, which each have 2600 mAh capacity.
2.2. Brief Descriptions of Sensor Node
The main components of the sensor node include one XBee module, one microcontroller unit (MCU), and four sensors on the sensor board in this testing. XBee-Pro 900HP (from Digi International Inc, Minnetonka, MN, USA) is a low-power module that consumes only 2.5 µA during sleep time. The MCU in this work is ATmega328p (from Atmel), which features low power, low cost, and high performance.
The four sensors were carbon dioxide (CO2), carbon monoxide (CO), temperature, and humidity. The temperature and humidity sensor were both soldered on the board, while CO2 and CO sensors were removable. The temperature sensor (MCP9700 from Microchip) is an analog sensor that can be connected to the analog-to-digital conversion (ADC) pin of the MCU. It can be powered from 2.3–5.5 V with only 0.6 µA current consumption. The humidity sensor (HIH5030 from Honeywell International Inc, Morris Plains, NJ, USA) is an analog low-voltage sensor, which operates from 2.7 V to 5.5 V with 200 µA current consumption. The CO sensor (MICS-5121WP from SGX Sensortech Limited) is also an analog output sensor, which has a wide detection range from 1 to 1000 ppm. It is the most power-hungry sensor in our design, which consumes 30.7 mA current. The CO2 sensor (GC-0012 from CO2 Meter) is an ultralow power consumption (3.5 mW) sensor which can measure 0–10,000 ppm CO2 concentration. This sensor supports digital output and is read by the MCU. Temperature, humidity, and CO2 sensors were controlled by the SW2 and woke up for 15 s, while the CO sensor was controlled by the SW1 and was turned off immediately after the reading had been taken to save power.
4. Conclusions and Possible Future Improvements
An autonomous IoT network system using solar energy harvesting has been presented in this work. The proposed technique provides an energy solution to keep sensor nodes active and reliable for a whole day.
For supercapacitor-powered nodes, during the day with sunny or partially cloudy weather conditions, the sensor node could run at least 8 h without interruption. At night, the 50 F supercapacitor could support the sensor node for the whole night until dawn the next day. On the second day of testing, the weather was partially cloudy and the voltage could still be maintained at 4.6 V in the day. However, if there is going to be heavy cloud coverage for a several days, the 50 F supercapacitor-powered node may be offline because the power extracted from the solar panel is close to 0 in this weather condition. To avoid the sensor node being offline, a much higher supercapacitor value is needed in order to provide the required energy to the node during such days. When the supercapacitor is fully charged to the programmed level, the energy-harvesting board almost stops sourcing energy from the solar panel. Some future improvements to utilize the solar energy better, in case of excessive energy, could involve increasing the sampling frequency and transmission power in order to have a more continuous monitoring sensor network and improve the network coverage area. The energy stored in the supercapacitor (E) was equal to ½ CV2. Therefore, when the supercapacitor was disconnected at 3.6 V, there still remained ½ × 50 × 3.62 = 324 J of unused energy. A practical solution to maximize the utilization of the energy would be integrating a buck-boost voltage regulator.
For battery-powered nodes, a small-sized solar panel could successfully charge the large-capacity rechargeable battery, which was 2600 mAh. The battery voltage could gradually increase from the time we deployed the sensor nodes and could be maintained at a certain level during the night. This also fulfilled our expectations of charging the battery and keeping the sensor nodes running without disruption.
Battery-powered nodes can start to communicate with the base station right after the deployment because a new rechargeable battery has an initial voltage about 3.8 V Nevertheless, for supercapacitor-powered nodes, there is a delay in network connection if the supercapacitor is deployed at night. For the network that we set up in this experiment, it took approximately 11.5 h (8:20 p.m. to 10:00 a.m.) to join and synchronize with the whole network. To enhance the performance of supercapacitor-powered nodes, all the supercapacitors can be charged before the deployment.
The maximum power that can be extracted from the solar panel is 300 mW (sunny day), 150 mW (partially cloudy day), and 23 mW (1 h before sunset). The sensor node only draws an average power of 4.9 mW. In the worst month (June), the daily peak sunlight hours are 2.4 h in Melbourne [
29]. Considering that the harvesting board is running at 300 mW for 2.4 h only, the average power from the solar panel will be (300 × 2.4)/24 = 30 mW. Although 30 mW is still oversized for sensor nodes, the sensor node can utilize more solar power if needed when the duty cycle of the network is increased.
A better solution for the WSN would be having two separate energy supply units. One would have a supercapacitor as the main source and a rechargeable battery as the backup source. Whenever the supercapacitor’s voltage drops below a certain level, the control circuit will switch the power supply from the supercapacitor to the battery so that the network will keep on running and prolong the lifetime of the sensor node. Our future goals for the WSN will be improving the energy-harvesting unit so that it can have a more robust control circuit to manage the power supply unit and improving the sensor box so that direct sunlight will not cause the internal temperature to increase too much.