Self‐Powered Wireless Temperature Monitor System Based on Triboelectric Nanogenerator with Machine Learning

Triboelectric nanogenerator (TENG) can power wireless, real‐time sensing system with hybrid electromagnetic or piezoelectric power, or directly drive commercial LED without battery. However, it is a great challenge to directly drive wireless real‐time sensing system due to low energy density based on environment energy. Here, a self‐powered smart wireless temperature monitoring system that uses machine learning to accurately measure the ambient temperature is developed. A position modulation‐based TENG‐driven transmitter enables wireless communication and real‐time temperature monitoring. This machine learning‐based wireless sensor can accurately monitor the ambient temperature, with a recognition accuracy of up to 96.2%. This sensor architecture could potentially be used in low‐cost distributed sensors for environmental monitoring.


Introduction
In recent years, climate change has attracted much attention due to its significant impact on ecosystems, [1] biodiversity, [2] agriculture, [3] human health, [4] etc. Distributed wireless sensors could form the basis of future environmental monitoring systems, [5] and such devices require a sustainable, distributed power supply. [6]riboelectric nanogenerators (TENGs) are one of efficient environment energy haversting technologies for using distributed energy to power sensors networks. [7]ENGs can efficiently convert random mechanical energy into electrical energy even at low-frequency, which are low-cost, easy fabrication, and diverse material choices. [8]Therefore, TENG is widely used in energy harvesting, self-powered sensing, blue energy and high-voltage application. [9]any studies use TENG to develop self-powered wireless sensor systems, which have successfully identified different types of environmental information such as air pressure, [10] humidity, [11] and temperature. [12]The system includes TENG, rectifier circuit, energy storage circuit, power management components, microcontroller, sensors, and wireless communication module.In this system architecture, TENG converts the micro-nano energy in environment into electric energy, and stores energy in the storage unit by rectifier and filer circuits, usually controlled by the energy management circuit.Battery-free wireless sensors require low-power energy management, data process, and wireless transmission units. [13]he dramatic increase in computing resources and the development of machine learning are having a profound impact on sensor technology. [14]There is an urgent need to develop a self-powered smart digital wireless sensor enabled by machine learning without integrated circuits and microcontrollers.
Here, we present a TENG based self-powered wireless temperature sensor for online temperature monitoring (Figure 1A).The temperature sensor consists of a TENG, a bimetallic thermometer and an LED array without additional integrated circuits and microcontrollers, and realizes the multifunctional functions of battery-free, temperature sensing, and wireless communication.Moreover, data processing based on machine learning has greatly improved the recognition accuracy, reaching 96.2%.Finally, the wireless sensor integrated with a camera and a computer platform demonstrates a realtime wireless temperature monitoring system that can visualize the ambient temperature.This work proposes a unique sensor architecture that underpins future environmental monitoring system infrastructure.

Design and Integration of Battery-Free Smart Digital Wireless Temperature Sensor
The self-powered wireless temperature sensor was composed of a TENG, transmission module (LED array), and position modulation module (bimetal thermometer), as shown in Figure 1B.The detailed structure of the TENG, as shown in Figure S1A, Supporting Information, consists of a rotor and a stator.As shown in Figure S1B, Supporting Information, the rotor part contains triboelectric material (PTFE), foam, and an acrylic substrate.The stator section, as shown in Figure S1C, Supporting Information, contains a printed circuit board (PCB) and an acrylic substrate.The LED array contains three discrete LEDs whose position signal reflect the characteristic information of the temperature near the sensor (Figure 1C).Polytetrafluoroethylene (PTFE) and copper (Cu) foils were used as friction layer and electrode layer, respectively, to constitute a triboelectric nanogenerator (Figure 1D).The combination of bimetal thermometer and LED array ensures different light output signals at different temperatures.When the temperature around the sensor changes, a unique optical signal will be generated due to the mechanical deformation of the bimetal strip.The data acquisition module (camera) records these signals and transmits them to the computer in real time.The computer performs machine learning and displays the recognition results on the computer screen (Figure 1E).
The working mechanism of TENG is shown in Figure 2A.First, the PTFE film is in contact with the electrode A. Due to the triboelectric effect, negative and positive charges are located on the PTFE surface and the electrode A, respectively.When the PTFE film slides towards the electrode B, the voltage between the two electrodes increases and electrons flow from the electrode B to the electrode A. When the PTFE film is in full contact with the electrode B, the charge rebalances and no charge is transferred to the external circuit.When the PTFE film slides forward, an opposite voltage will be established and drives electrons to flow from electrode A to electrode B. This constitutes a full operating cycle of TENG.
The output performance of TENG at different frequencies is shown in Figure 2B-D.Within a range of 1.1 to 3.1 Hz, the open circuit voltage (V OC ) gradually decreases with the increases of frequency, while it has little effect on the open circuit voltage when the frequency exceeds 3.1 Hz (Figure 2B).The transferred charge tends to decrease as the frequency increases (Figure 2C).As shown in Figure 2B,C, the open circuit voltage and charge remain at values around 673 V and 0.17 nC in the low-frequency state.As the frequency increases, the output decreases due to insufficient contact between the PTFE and the electrode. [15]owever, the short-circuit current (I SC ) increases with frequency (Figure 2D), which may be due to charge transfer in a short time at higher frequencies.In addition, the output power of TENG with different external loads at 5 Hz is shown in Figure 2E.The TENG obtained a peak power of 167.6 μW at 100 MΩ.As shown in Figure 2F, TENG can easily light up LEDs.

Signal Characteristics of LED Array Position at Different Temperatures
The self-powered wireless temperature sensor system driven by TENG is shown in Figure 1D.The wireless temperature sensor consists of a TENG, an LED array, and a bimetallic thermometer.The TENG can easily power three LEDs under mechanical stimulation.By associating the LED array with the bimetal thermometer, the temperature information can be reflected by the position change of the LEDs (Figure 3A).This solution does not require additional power supplies and complex power management circuits.
The sensor can measure the temperature of its surroundings (Figure 3B).The positions of the LED arrays at four different temperatures (T = 0, 10, 20, and 30 °C) are shown in Figure 3C.The image data collected by the camera at four different temperatures is shown in Figure S2, Supporting Information.The position of the LED array reflects the temperature features of the sensor, where a smaller T value represents larger angle (Figure 3D).The angles of the LED array at the temperature of 0, 10, 20 and 30 °C are about 123.4°, 93.2°, 62°a nd 31°, respectively.As the temperature decreases, the angle generated by the LED array increases, which is attributed to the bimetal that converts the temperature change into a mechanical displacement, enabling it to sense the ambient temperature.
The analog data captured by traditional sensors must be converted to digital signals, stored in memory, and then sent to the computing unit. [16]Data conversion and transmission in traditional architectures results in high power consumption and low efficiency.A battery-free wireless temperature sensor can send the collected signal directly to a computing unit.Compared to traditional architectures, TENG-based self-powered wireless sensor systems can quickly create and prototype new applications without spending significant time, cost, and expertise developing custom sensors.

Integrating Machine Learning Enables Temperature Recognition Accuracy up to 96.2%
Differences between temperature are clearly presented by identifying modulated position of these optical signals.Promising results have been produced by machine learning in various tasks such as material recognition and lip recognition. [17]The premise of machine learning is feature extraction. [18]Here, preprocessed methods for feature extraction that have a significant impact on machine-learning prediction accuracy are selected here, including data transformation, imaging enhancement, and data compression.Figure 4A shows the flowchart of TENG-based AI sensors for temperature identification with the assistance of machine learning.Initially, the optical signals images of known temperature were pre-processed.Then, a machine-learning algorithm was established between images features and the labels.The temperature could be identified after the model was established.The linear discriminant analysis (LDA) algorithm, a supervised learning algorithm, was chosen as the temperature identification model. [19]The accuracy of machine learning increased with more training samples (Figure 4B).The feature data of a temperature sensor was shown in Figure 4C.The original data can be collected from a single sensor, thus, machine learning can clearly predict the temperature with a low error.Figure 4D illustrates the dependence of LED array angle (β) on temperature.Previous strategies rely on high output of TENG for an accurately recognition, which demand complicated circuit.With the application of machine learning, high accuracy of identifications can be achieved by the TENGbased AI sensor.In this study, the temperature could be recognized with the ultra-low-power optical signal generated by the TENG; this not only reduce cost of time and money in equipment manufacturing, but also ensure the accuracy of identification.
The detection range and accuracy of the bimetal thermometer are À30 to 50 °C and AE2 °C.The bimetal thermometer uses the bimetal strip, which converts the temperature into the mechanical displacement.The detection range, accuracy, and resolution of the self-powered temperature monitoring system are À30 to 30 °C, AE2, and 0.1 °C, respectively.The response of the LED array angle (β) to temperature changes is almost linear.

A Minimalist Wireless Temperature Sensor Based on TENG Accurately Detects the Ambient Temperature
To demonstrate the real-world feasibility of a TENG-based realtime wireless temperature monitoring system, a high-precision, self-powered temperature sensor was constructed that integrates a TENG, a bimetallic thermometer, and an optical wireless communication module.As a data acquisition module, the camera can precisely capture signals with detailed information about the ambient temperature.The signals were transmitted to cloud computer and the results from machine learning were displayed in real time on a LED monitor.The temperature sensor is integrated with the machine learning platform to establish a real-time wireless temperature monitoring system, as shown in Figure 4E. Figure S3, Supporting Information, shows the temperature identification results.The process of temperature monitoring is shown in detail in Movie S1, Supporting Information.
The workflow of real-time wireless temperature monitoring is shown in Figure 4F.During operation, when the temperature sensor is triggered, the temperature information is transmitted from the LEDs to the camera.Real-time machine learning was applied to the acquired signals and temperature information was intuitively shown on the computer LED monitor.The terminal display interface for temperature monitoring is shown in Figure 4G.The software display interface shows statistical temperature information as shown in the left half, and the right half shows real-time temperature.

Discussion
Herein, we have proposed a smart wireless temperature sensor that does not require integrated circuits and microcontrollers for use in meteorology and agriculture to provide online temperature monitoring.Through machine learning, the temperature detection accuracy was up to 96.2%, and many common temperatures can be accurately detected.In addition, a smart wireless temperature monitoring system was demonstrated, which enables real-time wireless temperature monitoring.
Zhang et al. developed a glove-based human-machine interaction (HMI) system, using ultrastretchable, self-healing conductive hydrogel-based TENGs (PTSM-TENG). [20]This HMI system is capable of implementing functions such as gesture visualization and control of a robotic hand.In addition, the PTSM-TENG-based gloves were able to classify and recognize five objects using learning technology, achieving 98.7% accuracy.Zhang et al. used low-cost flexible pressure sensors to develop a human-computer interaction system for wireless and accurate hand posture detection. [21]Wang et al. developed a self-powered gas and humidity sensitive monitoring system based on a wavedriven TENG to monitor the marine environment. [22]Wang et al. developed a TENG-driven MXene/CuO ammonia (NH 3 ) sensor with excellent response at room temperature and can be used to detect pork spoilage. [23]istributed low-cost wireless sensors could form the basis for future environmental monitoring systems. [24]Traditional wireless sensors that require cables or batteries can be costly to install and maintain. [25]While solar energy enables the operation of sensors that do not require batteries, its use can be limited by factors such as night time, working area and weather conditions, especially in round-the-clock monitoring. [26]Self-powered environmental monitoring methods based on TENG were investigated. [6]However, communication modules based on traditional radio frequency technology tend to consume more energy. [27]Self-powered temperature monitoring systems based on traditional radio frequency technology typically require several to tens of minutes to complete the initial sensing and data transmission process. [28]e introduced optical wireless communication and mechanical modulation for the first time.LEDs powered by TENG can be directly converted into transmitters for optical wireless communication systems. [29]Instead of using traditional methods with electronic modulation circuits, we encode the temperature message as positional features of the visible light signal, which is determined by the bimetallic thermometer.Mechanical modulation not only simplifies circuit design, but also eliminates the waiting time required by the self-powered temperature monitoring system for initial acquisition and data transmission.

Conclusion
In summary, we have developed a self-powered wireless temperature sensor based on a triboelectric nanogenerator.In this wireless sensor, the TENG serves as a power source for the LED array while the bimetallic thermometers act as an information source.These components are combined into a complete wireless transmission unit, eliminating the need for complex power management and signal modulation circuits.The temperature sensing capability is demonstrated by the mechanical position modulation method, which can be considered as the relationship between the angle of the LED array and the temperature.In addition, we have developed a self-powered intelligent wireless sensor system that combines machine learning.This system can collect real-time temperature data for environmental monitoring.This study demonstrates the extension of self-sufficient wireless sensor systems for low-power artificial intelligence applications, achieved by integrating optical communication and mechanical modulation methods.

Experimental Section
The Self-Powered Temperature Sensor: The self-powered temperature sensor integrates an energy harvester, a signal modulation module and a communication module.The energy harvester consists of a wind cup and a TENG.The signal modulation module uses a bimetallic thermometer, which is a mechanical temperature sensor.The communication module consists of three commercial LEDs.The LEDs are integrated respectively on the pointer and face of the bimetallic thermometer and their locations are shown in Figure 3a.The communication is wireless optical communication (LED array) for wireless signal transmission.Machine learning is done on the computer side.
Fabrication of TENG: The TENG consists of a stator and rotor.The rotor substrate has a thickness of 4 mm and a diameter of 8 cm.Likewise, the stator substrate has a thickness of 5 mm and a diameter of 11 cm.Additionally, an 8 mm diameter hole is drilled in the center of the rotor and stator base plates to accommodate the bearings.To improve contact intimacy, two sector-shaped pieces of foam, each 1 mm thick, are glued to the rotor base plate.PTFE with a thickness of 80 μm is then glued onto the foam as a friction layer.A custom-made PCB circuit with two pairs of copper electrodes is attached to the stator base plate.In the middle of the board there is an 8 mm diameter hole to accommodate the drive shaft.
Characterization and Measurements: The output performance (open-circuit voltage, transferred charge, and short-circuit current) of the triboelectric nanogenerator was measured with an electrometer (Keithley 6514).

Figure 1 .
Figure 1.Design and structure of a smart wireless temperature sensor.A) Schematic diagram of the temperature sensing process of the TENG-based smart wireless temperature sensor.B) Structure of the TENG-based smart wireless temperature sensor, which consists of a TENG, a data modulation module, and a transmission module.C) Schematic diagram of the LED array output signal when the smart wireless temperature sensor senses different temperatures.D) Flowchart of the interaction between modules in the smart wireless temperature sensor for temperature sensing.

Figure 2 .
Figure 2. TENG performance characterization.A) Schematic diagram of the working mechanism of TENG.B) Open-circuit voltage (V OC ), C) transferred charge, and D) short-circuit current (I SC ) of TENG at different frequencies.E) Output power of TENG under different resistance.F) Voltage variation of the LED at different frequencies.

Figure 3 .
Figure 3. Temperature response of LED array.A) Schematic of the LED array on the smart wireless temperature sensor.B) Schematic diagram of temperature sensing of the TENG-based smart wireless temperature sensor.C) LED positions when the sensor is at 0, 10, 20, and 30 °C.D) Angles of the LED array at different temperatures.

Figure 4 .
Figure 4. Temperature test.A) Block diagram of machine learning for temperature identification.B) Prediction accuracy for different sample sizes.C) Comparison of predicted and actual temperatures.Data are mean AE standard deviation.For each average, the number of measurements is 50.C) Dependence of LED array angle (β) on temperature.E) Photograph of the smart wireless temperature monitoring system.Through machine learning, the system successfully realizes wireless temperature monitoring and displays the result on the computer screen.F) Workflow diagram of real-time wireless temperature monitoring.G) Computer interface for wireless temperature monitoring.