Treffer: Deep Reinforcement Learning-Based Adaptive Control of a Concentrated Solar Thermal Plant.

• Adaptive control strategy to dynamically optimize temperature regulation in solar plants. • A real-time adaptation of PI controller's parameters using Reinforcement learning algorithm. • A comprehensive simulation scenario for validating the effectiveness of the proposed approach. • Significant improvements using DDPG based PI over traditional PI controller. Effective temperature control of solar thermal plant is crucial for optimizing energy production and prolonging the lifespan of plant components, especially given the highly variable and unpredictable nature of solar energy. Although proportional-integral-derivative controllers remain widely adopted due to their simplicity and ease of implementation, their reliance on fixed parameters poses significant limitations in handling system nonlinearities and disturbances, often resulting in suboptimal performance. To address these challenges, this study introduces an adaptive temperature control framework for a concentrated solar thermal plant based on deep reinforcement learning. The proposed control scheme utilizes the deep deterministic policy gradient algorithm to continuously adjust the parameters of a proportional-integral controller, enabling rapid adaptation to changing operational and environmental conditions. The effectiveness of the framework is validated through comprehensive simulations in Matlab®/Simulink®, using an experimentally verified model of a linear Fresnel plant. To ensure robustness, the controller is tested across various operating conditions, including different setpoints scenarios, abrupt load changes and rapid solar irradiance transients. The results demonstrate substantial improvements in control performance, achieving reductions of up to 99.5 % in mean squared error and 94 % in mean absolute error compared to the conventional proportional-integral controller. Under abrupt irradiance disturbances, the controller reduces overshoots by more than 83 %, achieves significantly faster settling times, and maintains a steady-state error below 0.004 °C, underscoring its superior transient response and robust disturbance rejection capabilities. These findings demonstrate the effectiveness of deep reinforcement learning in improving the control performance and operational efficiency of concentrated solar thermal plants, contributing to the advancement of sustainable and resilient energy systems. [ABSTRACT FROM AUTHOR]

Copyright of Applied Thermal Engineering is the property of Pergamon Press - An Imprint of Elsevier Science and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)