The work cycle of profiling floats can be divided into four phases: descent, drift, ascent, and communication ( Figure 2 , Table 1 ). During the drift phase, the control system transitions to sleep mode, while other systems are temporarily powered off. Consequently, power consumption during this phase is minimal and it is often omitted from motion energy analysis.

Statistical analysis indicates that the primary energy consumption of profiling floats during a work cycle ( Figure 2 ) is mainly allocated to hydraulic drive, satellite communication, sensor measurement, and system control ( Table 2 ). Accordingly, the research on low-power consumption strategies has been primarily focused on these four aspects ( Figure 3 ). Among these, the hydraulic system constitutes the most energy-intensive component, thereby becoming a major focus of optimization efforts. For APEX floats, which are the most widely deployed type in the Argo program, the hydraulic system consumes approximately 28% of the total energy consumption, whereas, for Deep SOLO floats operating at depths of up to 6,000 meters, this proportion reaches as high as 78% (Roemmich et al., 2019b).

As an autonomous platform, the energy consumption of each component of the profiling float is interdependent to a certain extent ( Figure 3 ). For example, increasing sensor sampling frequency not only raises the power consumption of individual sensors but also generates larger volumes of observational data. This increase in data volume imposes additional processing requirements on the control system (e.g., for data compression) and increases the burden on satellite communication. Given the limited bandwidth of satellite communications, transmitting large volumes of data may necessitate extended communication durations, thereby significantly increasing the energy consumption of the satellite communication system. Consequently, effective energy management requires a system-level approach to minimize overall power consumption.

To develop the low-power strategy, it is typically necessary to first analyze the motion model of the float and subsequently construct an energy consumption model to analyze the factors affecting energy consumption under different motion states of profiling floats (Yang et al., 2019). When constructing a motion model, most researchers tend to simplify the analysis by disregarding the influence of lateral currents or the vertical variation of seawater density, focusing primarily on the forces of gravity, buoyancy, and drag in the vertical direction (McGilvray and Roman, 2010; Barker, 2014; Dologlonyan and Grekov, 2018; Zou et al., 2018; Yang et al., 2019; Guo, 2020; Liu et al., 2020; Wang et al., 2020b; Wu, 2021; Wen et al., 2022). However, due to inhomogeneous mass distribution within the float, lateral flow perturbations induce an inevitable tilt angle. In this regard, Si et al. constructed a single-profile energy consumption model for the 4000-m Deep-Argo Otarriinae, incorporating 19 parameters (Si et al., 2020). It was found that the gliding angle, diving depth, and gliding speed had the greatest impact on energy consumption, while the influence of other parameters was relatively limited (Si et al., 2020). Therefore, applying a simplified model that neglects the effects of horizontal flow may lead to significant errors when researching optimizing float energy consumption.

Commercially available profiling floats typically utilize a highly reliable hydraulic system as the buoyancy engine to achieve ascent and descent. This system mainly consists of an internal reservoir and external bladder, solenoid valves, a hydraulic pump, and pipelines. During the float’s movement, the hydraulic pump drives oil to flow between the internal reservoir and the external bladder, thereby changing the float’s volume to achieve ascent or descent. Due to the need for the hydraulic system to actively discharge oil to counteract the high pressure of the deep sea during the ascent phase, energy consumption during this phase is greater than in others. Therefore, researchers have often focused on modeling and analyzing the energy consumption of the hydraulic system during the ascent phase.

When constructing an energy consumption analysis model for hydraulic systems ( Table 3 ), studies generally employ numerical analysis methods such as the Runge-Kutta method, to solve differential equations and obtain solutions that minimize power consumption (Chen et al., 2017; Zou et al., 2018; Wang et al., 2020b). Some researchers have also developed simulation models (Dologlonyan and Grekov, 2018; Yang et al., 2019; Guo, 2020) and applied the Non-dominated Sorting Genetic Algorithm II (NSGA-II) (Liu et al., 2019b) for energy consumption analysis.

After conducting an energy consumption analysis of the hydraulic system, researchers primarily focused on low-power technology in areas such as controlling operating speed, managing oil discharge frequency, and optimizing the hydraulic system’s structural design.

The energy consumption of the sensor system and satellite communication system of profiling floats occurs during the processes of data generation, processing, and transmission. Thus far, low power consumption has been primarily achieved by reducing data volume and decreasing working time.

The sensors on profiling floats can be divided into ocean observation sensors and auxiliary operational sensors. The latter typically utilize mature micro-sensors, which consume minimal power. However, the continuous operation of ocean observation sensors during the ascent process can lead to increased energy consumption. For instance, the energy consumption ratio of the sensors on the SOCCOM APEX float, equipped with multiple biogeochemical sensors can reach as high as 41.4% (Riser et al., 2018). Furthermore, the volume of data collected during observations is directly related to the energy consumption of subsequent satellite communications.

The observation sampling frequency of profiling floats depends on their design, sensor type, and task requirements, and different floats may have different sampling frequencies at varying depths. To reduce power consumption, it was proposed that extracting a set of characteristic values from multiple groups of data measured within a certain depth interval to represent the true values within that interval would decrease the subsequent data transmission volume (Liu, 2020). However, reducing the volume of data may lead to the inability to observe high-frequency changes in the ocean and the omission of critical ocean data. To balance data recording and energy consumption, the Argo program suggests some sampling strategies for several float types (Argo data management, 2022). For the Provor Bio 5.0 Argo float equipped with a Seabird CTD and Wetlab Saraover optical sensor, the primary sampling (CTD sampling) is typically set to occur every 25 dbar from the bottom to 200 dbar and every 10 dbar from 200 dbar to the surface, while the secondary sampling (chlorophyll sampling) is typically set to occur every 10 dbar from 1000 to 300 dbar and every 1 dbar from 300 dbar to the surface (Argo data management, 2022).

Once the data collection of the profiling floats is completed, communication and positioning are conducted on the sea surface. During this phase, energy consumption is primarily influenced by the transmission efficiency of the satellite system, operational duration, and hardware power consumption. Initially, the profiling floats primarily utilized the ARGOS satellite system (0.06 kByte/min) to obtain location and transmit data. Its unidirectional transmission mechanism necessitated prolonged sea surface floating durations (6–12 hours) for data transmission and positioning operations, thereby significantly elevating the risk of encountering surface obstructions (such as ships, sea ice, or flotsam) by 3–6 times (Morris et al., 2024). Since 2005, the profiling floats have adopted GPS for location acquisition and have utilized Iridium satellites for bidirectional communication in either SBD (approximately 1 kByte/min) or RUDICS (approximately 12 kByte/min) mode (André et al., 2020). This significantly reduced communication time, with the transmission of 1000 temperature/salinity (T/S) standard profile data taking only about 7 minutes (Gruber et al., 2007). For instance, an APEX Argo (7553) float equipped with the Daytona 9522A Iridium module had a power consumption of 4.2 W during communication connection, was capable of transmitting 160 bytes per second, and its GPS module had a power consumption of 0.221 W, with a typical GPS positioning time of 120 seconds (Gordon, 2017). As of February 2025, the usage of the Iridium communication system in the global Argo observation network had reached 97.54%. ³

The COPEX float, HM2000 float, and FUXING float developed in China all utilize the BeiDou Navigation Satellite System as their communication method, which features positioning, navigation, timing, and short message communication services (Lu et al., 2016; Liu et al., 2019a; Wang et al., 2022). Currently, the latest BeiDou-3 system provides a short message communication service capable of handling 10 million communications per hour for China and surrounding regions, with a receiver transmission power of 1–3 W, and supports a maximum of 14,000 bits per single communication, while the global short message communication service supports a maximum of 560 bits per single communication (China Satellite Navigation Office, 2019). Furthermore, a low-energy transmission terminal based on the BeiDou system was developed for Argo floats through strategic power-state modulation. By deactivating hardware components during non-transmission intervals and leveraging multi-stage STOP-mode transitions, the system demonstrated a 77.282% decrease in single communication energy demand compared to continuous power-dissipation configurations (Li et al., 2024).

Due to the limited satellite bandwidth, transmitting large volumes of data typically necessitates a prolonged transmission period, thereby leading to increased energy consumption. Researchers have frequently employed lossless data compression methods to enhance transmission efficiency while maintaining data integrity and accuracy. For example, a lossless compression rate of 26% was achieved using the LZSS method (Xie, 2020), and a 25.9% compression rate was realized with a Differential coding-Huffman method (Li et al., 2024). Additionally, a novel block-wise lossless compression method was developed for Argo floats data, integrating bidirectional LSTM networks, multi-head self-attention mechanisms, and multilayer perceptrons. This method achieved a compression rate of 12.11% on both PC and Jetson Nano platforms (Guo et al., 2024).

Profiling floats are autonomously operated instruments, and control algorithms are required to achieve precise buoyancy adjustment, motion control, data collection and transmission management, and energy allocation.

The most classic control algorithm employed is proportional-integral-derivative (PID) control (Borase et al., 2021), especially in control systems that require maintaining a set point. It is simple and easy to use, thus widely utilized in the depth control of underwater vehicles (McGilvray and Roman, 2010; Morales-Aragón et al., 2025). However, the PID algorithm parameters are fixed, while the float experiences fluctuations in pressure and flow during the diving process. Therefore, a fixed-parameter PID controller may not be able to adjust the control signal in some situations, leading to slower system response, overshoot, and increased energy loss (Mu et al., 2019; Li et al., 2023). In addition, there is a trade-off between power consumption optimization and operational performance metrics during the hydrodynamic adjustment phases (Carneiro et al., 2024). The PID controller is typically optimized for a single objective, making it difficult to achieve global optimization and thus unable to attain the best energy-saving effect.

Some research efforts have focused on developing control algorithms to further reduce energy consumption. Studies have shown that advanced control algorithms based on PID have yielded limited improvements in energy efficiency (Mu et al., 2019; Yang et al., 2019). Therefore, alternative control algorithms have been proposed. For example, motion planning based on genetic algorithms was implemented for a 4000-meter profiling float, reducing energy consumption to 1/50 of that of dual closed-loop fuzzy PID control and RBF-PID methods, while maintaining the same accuracy (Zheng et al., 2021). A speed closed-loop control strategy was proposed for a 4000-m profiling float, resulting in an 18% improvement in energy efficiency compared to the traditional control strategy when the set speed was 1 m/s (Liu et al., 2020).

The PCM system stores thermal energy through phase transitions: at the ocean surface, warmer seawater causes the PCM to transition from solid to liquid, thereby storing thermal energy. As the float descends into colder waters, the PCM returns to its solid state. This cyclical phase change drives a hydroelectric generator through pressure differentials created by volume changes, facilitating the conversion of thermal to electrical energy (Wang et al., 2019b). Alkanes are extensively adopted as PCMs due to their superior volumetric latent heat capacity, substantial solid-to-liquid volumetric expansion coefficients, and inherent chemical inertness, which eliminates phase segregation risks and material corrosion concerns (Wang et al., 2020a).

In addition to thermal energy harvesting technology, the integration of other energy harvesting technologies with profiling floats has also been explored. However, unlike PCM-based thermal recharging profiling floats, these technologies remain in the conceptual and preliminary validation phases.

In recent years, wave energy has been recognized as a high-density (2–3 kW/m²) renewable energy source that is readily accessible (Lopez et al., 2013). Several patents have been granted for the integration of hydraulic turbines (Liu et al., 2023b) and inertial electromagnetic power generation technology (Ren et al., 2019b) with profiling floats to capture energy from their wave-induced oscillations on the sea surface. An innovative profiling float design utilizing near-surface wave energy has been proposed ( Figure 6A ). Unlike traditional profiling floats, which rely primarily on hydraulic systems for buoyancy adjustment, this design harnesses wave-induced motion through a heave plate connected to turbines via stiff and elastic tethers (Shomberg et al., 2022). These components generate power through the flow induced by the pod’s oscillations and enable depth adjustment by modulating drag to facilitate data collection (Shomberg et al., 2024). This concept has been experimentally validated, demonstrating an electrical output of 10 W under simulated conditions (Shomberg et al., 2024).

In addition to harnessing wave energy at the sea surface, a current energy converter was designed for deep-sea profiling floats ( Figure 6B ). This system employs a spiral involute blade to capture radial current energy in shallow waters and axial relative current energy generated by the float’s autonomous vertical motion (Wu et al., 2018). The turbine’s minimum self-starting flow speed is approximately 0.3 m/s, while stable operation requires speeds greater than 0.6 m/s. Higher flow speeds, smaller load factors, and smaller buoy inclinations could improve performance, but they also amplify load fluctuations and increase the probability of damage (Wu et al., 2018). Therefore, load and speed must be optimized to ensure stable self-starting and sustained operation (Wu et al., 2018).

As a different development direction, multiple patents have described conceptual designs for solar-powered profiling floats (McCoy, 1993; Wang et al., 2006; Ren et al., 2019a). For instance, McCoy proposed a self-powered autonomous profiling float that integrates solar and thermal energy harvesting (McCoy, 1993). In this design, solar cells are installed within the float’s transparent housing to facilitate solar energy collection. Additionally, a thermoelectric device is positioned inside the float, with one side in thermal contact with a thermal mass(e.g., water or ethylene glycol) and the other in thermal contact with the surrounding water medium, creating a temperature gradient that facilitates electricity generation through the Peltier effect (McCoy, 1993). However, to date, there are no existing reports documenting the actual fabrication of solar-powered profiling floats.

In addition to harvesting external energy, the integration of a hydraulic motor into the buoyancy-driven system has been proposed to enable energy recovery for Deep Argo (Xue et al., 2020). It has been demonstrated that this design can recover 11.35% of the energy consumed by the hydraulic pump and reduce its energy consumption by 2.77% (Xue et al., 2020). However, several challenges remain, including mitigating inertia-induced fluctuations in the hydraulic motor’s rotational speed and output torque during energy recovery (Xue et al., 2020). Given the limited amount of energy recoverable by profiling floats in this process, extending its operational lifespan through energy recovery may be less effective than through energy harvesting.

Edge Artificial Intelligence (Edge AI) enables intelligent management in resource-constrained devices such as profiling floats. Unlike traditional AI, which relies on cloud-based data processing, Edge AI deploys algorithms and models directly onto low-power microcontrollers and edge devices (e.g., Nvidia Jetson Nano), utilizing techniques such as model compression, pruning, and optimization (Gibbs and Kanjo, 2023), thereby ensuring efficient operation on constrained hardware. For more demanding tasks, dedicated AI accelerators can be integrated. For instance, an RNN architecture implemented on a Lattice ICE40UP5K FPGA was shown to consume only 360 µW at a clock frequency of 146 kHz (Bartels et al., 2023). The combination of Edge AI’s computational efficiency and low power consumption makes it particularly suitable for energy management, data processing, and intelligent decision-making in profiling floats, with significant potential for advancing autonomous ocean observation.

In intelligent energy consumption prediction and management, Edge AI enables real-time forecasting of energy usage, dynamically adjusting computing resource allocation and power distribution based on environmental variations and data collection frequency (Nammouchi et al., 2021; Lv et al., 2023). This capability facilitates the optimal scheduling of component operation and energy consumption in profiling floats. For data processing, Edge AI can locally analyze seawater parameters such as temperature, salinity, and dissolved oxygen, enabling real-time anomaly detection and accurate trend identification (Tran et al., 2022; Zhao et al., 2022). Furthermore, in response to the complex and variable marine environment, Edge AI can enhance autonomous decision-making by integrating multi-sensor data to adjust sampling depth, data transmission frequency, and other parameters, allowing profiling floats to operate independently under diverse conditions (Wang et al., 2019c). These capabilities not only enhance system responsiveness and stability but also significantly reduce communication costs and energy consumption, providing a viable approach for the long-term, efficient, and autonomous operation of profiling floats with broad application potential.

Traditional solar cells face challenges in profiling float applications due to their large space requirements and susceptibility to corrosion (McLeod and Ringwood, 2022). However, recent advancements in flexible solar cell technology have opened new possibilities for integration into profiling floats. These novel solar cells are lightweight, portable, and flexible while exhibiting good bending cycle stability (Liu et al., 2023a). In addition, an autonomous underwater vehicle (AUV) design incorporating flexible solar cells has been proposed (Byford and Wood, 2019). Power generation performance was estimated, with outputs ranging from 424-1700Wh/m²/day at a depth of 1 ft below the surface, thereby validating the feasibility of this conceptual approach (Byford and Wood, 2019).

However, traditional solar energy harvesting methods remain predominantly constrained to surface or near-surface operations, thereby limiting their efficacy for underwater applications. Indeed, recent studies have demonstrated that solar energy is still accessible underwater, with irradiance values reaching approximately 130 W/m² even at a depth of 20 m (Qian et al., 2024; Lv et al., 2025). Consequently, underwater photovoltaic (PV) technology has garnered increasing attention. Accordingly, extensive research has been conducted on the selection and fabrication of solar cells optimized for underwater light conditions (Enaganti et al., 2020; Enaganti and Goel, 2021; Röhr et al., 2023). It has been demonstrated that wide-bandgap solar cells, particularly gallium indium phosphorus (GaInP) and organic solar cells with bandgaps exceeding 1.5 eV, outperform silicon solar cells (1.1 eV) (Röhr et al., 2023). These cells have exhibited operational efficiencies of 14% or higher even under low-light underwater conditions (Röhr et al., 2020, 2022; Samantaray et al., 2023; Qian et al., 2024). Furthermore, their low cost, high efficiency, and flexibility make them well-suited for underwater environments. However, at present, underwater photovoltaic technology has predominantly garnered attention within the field of photovoltaic materials. Future research will need to further explore its potential for powering underwater marine equipment, with particular emphasis on shallow-water applications.

Wave energy harvesting technology enables the conversion of mechanical energy from seawater vibrations into electrical energy through various mechanisms, including electromagnetic, electrostatic, and piezoelectric methods. Among these, electromagnetic generators (EMGs) are the most widely used, characterized by high power output and operational stability (Xu et al., 2022). However, EMGs generally exhibit low efficiency under mild wave conditions at sea (Xu et al., 2022).

Given that profiling floats are generally characterized by their small size and limited oscillation range, these constraints highlight the need for more compact and efficient systems. Triboelectric Nanogenerators (TENGs), characterized by their potential for miniaturization, simple design, affordability, high power density, and strong adaptability to low-frequency waves (Choi et al., 2023), have emerged as a promising solution for profiling floats. TENGs operate on the principle of the triboelectric effect, where charge transfer occurs between two distinct materials, generating an electric potential that can be harnessed as they make contact and subsequently separate (Shan et al., 2024).

In 2019, a high-output multilayered TENG was integrated into a self-powered intelligent buoy system to harvest wave energy on the surface, successfully providing a steady DC voltage of 2.5 V to the load (Xi et al., 2019). Additionally, a hybridized energy harvesting system integrating TENGs and EMGs was designed and evaluated on a buoy in the Jialing River and an AUV in the Huanghai Sea (Gao et al., 2020). It was found that the device’s resonance frequency was reached exclusively at a water wave frequency of 1 Hz, yielding peak voltages of 11.15 V for the TENG and 3.1 V for the EMG, respectively (Gao et al., 2020). Moreover, TENG units demonstrated the ability to sense angular attitude changes based on variations in output energy (Gao et al., 2020). However, most TENGs remain at the prototype stage and are currently undergoing laboratory examination.

PCM-based thermal recharging profiling floats have been commercially introduced. However, challenges remain, including low energy conversion efficiency, the poor thermal conductivity of PCM, and its substantial weight (Wang et al., 2020a). In addition, the power generation process requires the coordinated operation of multiple modules, including the hydraulic system, mechanical system, and batteries, which increases system complexity and vulnerability to malfunctions (Wang et al., 2020a). Additionally, the reliance on temperature gradients for phase transitions imposes operational limitations on PCM-based thermal recharging profiling floats, particularly in shallow waters and high-latitude regions.

In response to these limitations, alternative thermal energy harvesting technologies, such as shape memory alloys (SMAs), thermoelectric generators (TEGs), and thermodynamic cycles, have been actively explored. These technologies, along with their potential applications in UUVs, have been extensively reviewed (Wang et al., 2020a; Jung et al., 2022). Key features of these technologies are summarized as follows:

In summary, while these technologies offer promising alternatives, they are still in the conceptual and theoretical stages for UUVs. To date, no studies have explored their application in profiling floats. Consequently, further research and development are required to evaluate their potential in this context.

Given the complexity and variability of the marine environment, power generation technologies based on a single energy source or operating principle often exhibit limited effectiveness under fluctuating sea conditions. For instance, the performance of solar cells is significantly diminished during cloudy weather and at night, while the efficiency of thermal energy conversion is constrained by the magnitude of the temperature gradient. To address these limitations, the integration of multiple energy harvesting technologies(e.g., solar, wave, and thermal energy) can offer a more stable and continuous power supply, independent of transient environmental conditions (Wang et al., 2020a). Furthermore, as the spatial distribution of energy resources varies across different marine regions, with some areas being more suitable for wave energy harvesting while some for thermal energy, a multi-source power generation system can optimize energy harvesting by dynamically selecting the most appropriate energy source based on geographic and environmental characteristics, thereby enhancing overall efficiency.

In 2022, a hybrid nanogenerator (TPEPT-HG) was proposed, integrating TENG, PV, EMG, piezoelectric nanogenerator (PENG), and thermal energy (TG) units (Xue et al., 2022). This system was deployed on an intelligent ocean buoy for sustainable energy generation. Laboratory evaluations demonstrated that at an excitation frequency of 2.4 Hz, the maximum peak-to-peak power outputs of the TENG, PENG, and EMG reached 0.25 mW, 1.58 mW, and 13.8 mW, respectively (Xue et al., 2022). Additionally, the PV unit achieved a maximum open-circuit voltage of 1.33 V and a short-circuit current of 49 mA, while the TG unit produced 5 V and 15 mA (Xue et al., 2022).

Compared to single-source energy conversion systems, hybrid power generation technologies mitigate the temporal and environmental constraints associated with individual energy sources, thereby significantly improving overall energy harvesting efficiency. However, the practical deployment of such systems necessitates addressing key challenges related to energy integration, system coordination, and real-time power management. The development of advanced energy management strategies is crucial for ensuring efficient regulation, seamless switching among different power sources, and the reliable operation of hybrid power generation systems in complex marine environments.

In summary, the energy limitations of autonomous profiling floats are expected to be addressed through advancements in system-level energy consumption optimization, the integration of intelligent technologies, and breakthroughs in solar, thermal, and wave energy harvesting, coupled with the enhanced capabilities of hybrid power generation technologies. These developments will drive ocean observation toward greater efficiency, intelligence, and sustainability.