|
Abstract
|
Hybrid solar systems, particularly concentrated photovoltaic-thermal (CPV-T) systems, are gaining prominence for their dual capacity to generate electrical and thermal energy. Improving thermal power remains a critical challenge, with tracking-based mirror configurations emerging as a key solution. This paper proposes a real-time improvement method for thermal power enhancement in CPV-T systems by dynamically adjusting heliostat mirrors. A feedback control strategy is proposed, combining a proportional controller with an online regression-based machine learning correction term to adaptively refine mirror angles under varying operational conditions. The proposed approach addresses both static and dynamic inefficiencies inherent in tracking systems, ensuring sustained thermal output despite environmental and mechanical perturbations. Experimental validation is conducted on a custom CPV-T testbed featuring three coaxial mirrors actuated by dual-axis motors, enabling precision adjustments in elevation and azimuth. The real-time results demonstrate that the proposed method outperforms conventional configurations, achieving approximately 500% improvement in thermal power stability under transient solar irradiance and thermal load variations. Integrating machine learning with classical control principles offers a scalable framework for enhancing CPV-T performance, with implications for industrial solar cogeneration systems seeking robust, adaptive energy harvesting.
|