The integration of artificial intelligence with floating photovoltaic (FPV) systems has reached a new milestone. Researchers at Cranfield University have developed an innovative digital twin framework capable of predicting system performance with remarkable accuracy. By harnessing data from 155 physical experiments, this system employs a two-tier artificial neural network (ANN) to create both a high-fidelity model and a reduced-order model, achieving R² values of 0.9996 for PV surface temperature and 0.9189 for power output.
Advanced Digital Twin Framework
The research team has created a cloud-based digital twin that mirrors the behavior of physical FPV systems deployed on water surfaces. This system utilizes real-time data collected by sensors from the physical twin, which is then transmitted to the cloud. The digital twin not only simulates system behavior but also forecasts performance under various environmental conditions. It is designed to predict potential risks, allowing for proactive management instead of reactive responses.
Experimental Methodology
To develop the digital twin framework, researchers conducted a series of experiments in Cranfield University’s specialized wave tank, measuring 30 meters long, 1.5 meters wide, and 1.5 meters deep. A floating catamaran structure was centrally positioned, equipped with a 50 W solar panel and illuminated by a solar simulator positioned 40 cm above it. This setup allowed for the collection of time-series data on key parameters, including hydrodynamic motion, mooring line forces, PV surface temperature, and power output.
Data was gathered under varying conditions, including five different incident angles (90°, 75°, 60°, 45°, and 30°), two wave amplitudes (0.025 m and 0.0375 m), and 15 wavelengths ranging from 1.5 m to 5.0 m. All data points were sampled at a rate of 10 Hz, forming a robust dataset for training the digital twin models.
Two-Tier Neural Network Model
The digital twin framework incorporates a two-tier ANN composed of a high-fidelity model and a reduced-order model. The high-fidelity model is initially trained using the comprehensive experimental data, allowing it to accurately capture the complex behaviors of the FPV system. To enhance computational efficiency, a second, reduced-order model is developed. This model is trained on carefully selected outputs from the high-fidelity model, along with supplementary experimental data, enabling it to replicate key behaviors while requiring less computational power.
Impressive Predictive Performance
The high-fidelity model demonstrated exceptional predictive accuracy, with R² values of 0.9996 for PV surface temperature and 0.9189 for power output. It effectively modeled oscillatory behavior in surge and pitch motions, rapid variations in mooring forces, and transient power fluctuations, achieving root mean square errors as low as 0.1986 W for power and 0.1526° for PV temperature. The reduced-order model maintained strong performance, with R² values of 0.9073 for power output and 0.9660 for temperature.
Insights from Performance Mapping
The research team also generated three-dimensional performance maps, revealing significant nonlinear interactions between environmental inputs and system behavior. Notably, the heave motion reached peak levels under specific wave lengths of 2.5–3.5 m and higher amplitudes, indicating resonant conditions. Optimal power output occurred when solar irradiance surpassed 340 W/m² at a 90° incidence angle, with PV temperature exceeding 75°C under similar conditions. These insights enable predictive optimization and deepen the understanding of FPV performance amidst varying sea states.
Collaborative Research Efforts
The study, titled “Digital twins for a floating photovoltaic system with experimental data mining and artificial intelligence modelling,” showcases the collaborative efforts of scientists from Cranfield University and Beijing University of Posts and Telecommunications. This groundbreaking research paves the way for more efficient and effective management of floating PV systems.
Key Takeaways
- The new digital twin framework for FPV systems enhances predictive capabilities using AI.
- High-fidelity and reduced-order models provide a balance between accuracy and computational efficiency.
- The framework allows for proactive risk management and performance optimization under varying environmental conditions.
In conclusion, this innovative digital twin framework represents a significant advancement in the management of floating photovoltaic systems. By leveraging artificial intelligence, researchers have created a tool that not only simulates but also forecasts system performance, enabling more efficient and effective energy production. The potential applications of this technology could revolutionize the FPV landscape, promoting sustainable energy solutions worldwide.
Read more → www.pv-magazine.com