Catching Every Wave: AI-Enhanced Self-Powered Systems for Real-Time Ocean Monitoring and Disaster Prevention
The vast, mysterious oceans are not only sources of life and resources but also formidable forces capable of unleashing immense destruction. With the global economy increasingly expanding into marine territories, the risk of coastal disasters has grown significantly, contributing to approximately one-third of all natural disaster-related economic losses. Consequently, the ability to accurately monitor and predict oceanic conditions, including everything from wave heights to sea level pressure, has become critically important for maritime safety, disaster mitigation, and the sustainable management of our planet's resources. While digital technologies and artificial intelligence (AI) have significantly advanced the precision and timeliness of these monitoring and early-warning systems, traditional approaches have faced inherent limitations.
Conventional ocean monitoring often relies on systems that use "single-mode triboelectric nanogenerators" (TENGs), which are essentially devices that convert mechanical energy (like vibrations from waves) into electrical energy. The fundamental flaw of these single-mode systems is their dependence on acquiring only one type of signal, leading to a critical lack of accuracy in recognizing different ocean states and a lower reliability in providing early warnings. Moreover, the devices used for collecting hydrological data, such as oceanographic buoys, are typically powered by traditional energy sources like solar power, offshore wind energy, or batteries. While these sources are common, the harsh, corrosive marine environment drastically shortens their lifespan, necessitating frequent equipment replacements and incurring high maintenance costs. This situation has driven researchers to explore more resilient and self-sustaining power solutions, particularly those that can harvest the abundant wave vibration energy in the marine environment.
Into this critical need steps a groundbreaking innovation: the Self-powered AI-enhanced Monitoring System (SAMS). Proposed by researchers to address these challenges, SAMS is a highly integrated, multimodal system designed for diverse ocean state monitoring. This sophisticated system moves beyond single-mode limitations by combining two types of TENG interactions—solid-solid and liquid-solid—and incorporating three distinct triboelectric conversion mechanisms within a single, self-sustainable unit. The integration of artificial intelligence, specifically deep learning, allows SAMS to not only harvest energy but also to significantly improve wave level recognition accuracy and provide real-time warnings, representing a major leap forward for intelligent marine monitoring in complex environments.
At the heart of SAMS's ingenious design is its ability to harvest energy directly from the ocean's movements while simultaneously using those same movements to sense and interpret wave conditions. Triboelectric nanogenerators (TENGs) are an emerging power generation technology known for their high efficiency, light weight, greater power density, and lower manufacturing costs compared to other energy harvesting methods. TENGs work on the principle of contact electrification and electrostatic induction – essentially, generating electricity when certain materials come into contact and then separate, or slide against each other. While many TENGs traditionally rely on solid-solid contact, which can limit the effective power generation area due to material rigidity, SAMS leverages the fluidity of liquids to allow for larger effective contact areas by integrating solid-liquid TENGs. The breakthrough lies in developing multimodal TENGs that combine both solid-solid and solid-liquid contact effects within a single structure, a concept that was previously largely unexplored.
SAMS is conceptualized as a spherical framework, inspired by the multiple pentagonal and hexagonal panels on the outer surface of a football. This spherical design allows it to interact with various types of waves and collect omnidirectional wave vibration energy. The system cleverly integrates three different electret generation modules: the Double-Electrode Electret Generator (DEG), the Freestanding-layer Electret Generator (FEG), and the Spiral Electret Generator (SEG). Each module is strategically placed and designed to capture distinct information related to ocean wave activity.
Let's break down these three critical components:
1. The Freestanding-layer Electret Generator (FEG): Detecting Subtle Movements The FEG is located on the lower surface of the spherical SAMS. It is specifically designed to detect subtle wave vibrations through continuous liquid-solid contact. When wave intensity is low, the FEG is the primary module activated, coming into continuous contact with the water. This liquid-solid interaction generates induced charges on its surface, and the continuous water flow causes these charges to vary periodically, producing voltage outputs. The FEG's structure consists of an FEP film, copper electrodes (in pentagon and hexagon shapes), and a polyimide (PI) substrate, all fabricated using flexible printed circuit board (FPCB) technology to ensure it can conform to the spherical surface.
The charge transfer mechanism in the FEG involves the FEP surface continuously inducing stable negative charges, while an electric double layer forms at the contact area between the FEP and water. Electrodes within the system induce a positive charge to maintain electrical neutrality. As the SAMS moves up and down or rolls on the water surface, the solid-liquid contact area changes, causing induced charges to continuously transfer between the electrodes. A larger difference between the maximum and minimum contact areas typically results in a higher output voltage, which can be used to characterize wave height. For instance, during up-and-down motion, a peak voltage of approximately 0.04V is generated, while in a rolling state, a repetitive alternating signal of 0.02V peak-to-peak is observed. Even during "rushing" waves, the FEG generates a transient high voltage of 0.12V. These distinct outputs under various wave conditions are crucial for multimodal sensing.
2. The Double-Electrode Electret Generator (DEG): Capturing High-Intensity Splashes Positioned on the upper surface of the SAMS, the DEG is engineered to sensitively capture intermittent liquid-solid interactions that occur under high-intensity waves, such as splashes and scours. This module comes into play when wave intensity rises further, leading to splashes or rushing water striking the SAMS surface. The DEG features a double-electrode electret generator, enhanced via oxygen plasma treatment, which improves its sensitivity. This treatment makes the FEP surface hydrophilic (water-attracting), reducing the water droplet contact angle from 115° to 85° and enhancing the adhesion of water droplets. This is vital for effectively capturing the energy from intermittent water flows.
The charge transfer in the DEG is a three-step process: pre-charging, electric double layer formation, and charge transfer. When a water droplet impacts the FEP surface, the FEP's strong electronegativity allows it to capture electrons, becoming negatively charged. As the water leaves, a positive charge is induced on the electrode to maintain balance. When another droplet arrives, H₃O⁺ ions accumulate on the FEP surface, forming an electric double layer. Finally, as the water droplet spreads and contacts a surface electrode, the induced charges transfer, generating a signal. The DEG can produce significant voltage outputs, reaching up to 80V from rushing water and 40V from splashing water, demonstrating its excellent performance under intermittent water flow conditions. While effective for high-intensity interactions, the DEG generates weak output during continuous contact (like being immersed in water) and no output at all for low-level waves, highlighting the need for a multimodal approach.
3. The Spiral Electret Generator (SEG): Broadening Detection Range with Internal Vibrations Inside the spherical SAMS, a spiral electret generator (SEG) is installed to further broaden the range of detectable wave levels. This module is activated as wave intensity increases from low to moderate levels. The SEG is characterized by its dual-spiral structure and a mass block at its center that amplifies deformation. This design allows it to generate both in-plane and out-of-plane vibrations under low-frequency wave excitations, delivering outputs of up to 100V. For instance, it can exceed 300V at an in-plane vibration amplitude of 3cm and over 300V at an out-of-plane amplitude of 10cm. The SEG's spiral torsion spring structure offers good elasticity and self-rebound properties, making it highly sensitive to external vibrations.
The SEG operates by inducing capacitance changes: in-plane vibrations (swinging motion) cause varying distances between electrodes, while out-of-plane vibrations (up-and-down motion) change the area between electrodes. To enhance its performance, corona polarization is used to inject charge onto the FEP film surfaces, boosting triboelectric and electrostatic induction effects. The SEG's output voltage increases with vibration amplitude, and it can continue to vibrate with decreasing amplitude for about 2 seconds after a single excitation due to its elasticity. It can achieve a maximum output power of 1.18 mW in in-plane mode and 125 μW in out-of-plane mode. Crucially, the out-of-plane mode primarily exists during weak waves, while the in-plane mode predominates under stronger waves, offering distinct characteristics for wave level recognition. However, very large movements at the "Rough" level can restrict the mass block's displacement, reducing SEG output.
The Power of Multimodal Sensing Enhanced by AI
The true strength of SAMS lies in its triple-modal design, which enables the simultaneous generation of signals from all three channels. This comprehensive data collection is then fed into a powerful deep learning framework, specifically a 2D Convolutional Neural Network (CNN), to overcome the limitations of manual signal recognition from complex wave conditions.
Researchers categorized wave intensities into four levels: "Calm," "Smooth," "Slight," and "Rough". Each SAMS module responds differently at these levels, providing unique data signatures. For example, at "Calm" levels, only the FEG and SEG produce regular outputs; at "Slight" levels, the DEG starts generating weak signals from splashes, and the SEG shows more intense in-plane vibrations.
The CNN is trained to extract features from these collected wave environment signals and perform supervised classification. The signals from the electret generators are transmitted through different channels into the CNN. For the training, 1000 voltage values from each channel are treated as feature inputs, creating a dataset of 3000 voltage features from the three channels. A total of 200 data sets are collected, divided into the four wave categories, with 60% used for training and 40% for testing, resulting in a dataset of 600,000 data points. Data normalization is performed to account for performance variations between generators (e.g., SEG output in volts, others in millivolts), ensuring all inputs are within a consistent range.
The impact of using different signal sources on recognition accuracy was thoroughly studied. Single-mode structures showed limited accuracy: DEG at 75.00%, SEG at 77.5%, and FEG at a mere 41.25%. The FEG alone, for instance, struggled to accurately recognize ocean waves. Combining generators significantly improved performance; for example, combining SEG and FEG yielded 78.75% accuracy. However, the most dramatic improvement was observed with the triple-layer structure, combining FEG, DEG, and SEG, which achieved an impressive recognition accuracy of 96.25%. This clearly demonstrates that the signals collected by the three generators, operating through different mechanisms, complement each other, allowing for a more comprehensive extraction of wave information. The multimodal SAMS can recognize splash sizes at high wave levels and detect surface vibration amplitude variations at low wave levels, enabling it to monitor a wider range of wave intensities.
Building on this, an AI-enhanced ocean wave-level monitoring platform was developed to visualize these prediction results. This platform can display the three-channel signals and provide real-time warnings based on the CNN model's predictions. When the wave level is predicted as "Calm" or "Smooth," a green light indicates normal conditions. As the wave level reaches "Slight," a yellow light turns on, signaling an approaching dangerous level. Finally, if "Rough" waves are detected, a red light illuminates, indicating abnormal and dangerous wave levels, thus enabling real-time monitoring, recognition, and early warning.
In conclusion, the proposed SAMS represents a significant breakthrough in intelligent marine environment sensing. By integrating three distinct electret-based energy harvesting modules—DEG, SEG, and FEG—each designed to capture specific information about wave vibration amplitude, swinging motion, and splash extent, SAMS achieves comprehensive wave level monitoring. The crucial aid of two-dimensional convolutional neural networks (2D-CNN) transforms this collected data into actionable insights, dramatically improving recognition accuracy from a mere 41.25% in single-mode operation to an outstanding 96.25% in triple-mode. The development of a real-time monitoring platform further solidifies SAMS's potential to provide reliable solutions for accurate wave level monitoring and early warnings. This innovative self-powered, AI-enhanced system offers a vital tool for enhancing maritime safety, mitigating marine disasters, and supporting the sustainable management of our oceans in an increasingly complex world.
Researchers:
Dr. Xinhui Mao
Dr. Jiyuan Zhang
