Dynamic co-adaptation of nanofluids and machine learning: a closed-loop framework for self-optimizing concentrated solar power systems

Concentrated Solar Power (CSP) systems face persistent challenges in thermal management, where conventional heat transfer fluids struggle to mitigate localized hotspots and dynamic thermal gradients under fluctuating solar irradiance. This study introduces a novel closed-loop framework that pioneers the co-adaptive integration of tailored graphene nanofluids with a hybrid machine learning (ML) architecture for the self-optimizing control of CSP systems. Unlike prior studies focusing on static nanofluid optimization or standalone ML prediction, our framework enables a dynamic, real-time feedback loop where nanofluid performance informs ML control and vice versa. Results demonstrate the tangible benefits of this synergy: graphene nanofluids reduced thermal stratification by 32% compared to conventional metal oxides, while the Neural Network model achieved 94% accuracy in predicting optimal operating parameters. By embedding these ML insights into the system’s control logic, we enabled self-optimizing CSP operation that autonomously adapted to environmental volatility, reducing thermal losses by 18% without additional hardware. This work bridges material innovation and intelligent control, offering a scalable pathway to enhance CSP reliability and efficiency for grid-scale deployments.