27 November 2024 in IEEE Open Access Journal of Power and Energy
Space missions would not be possible without an available, reliable, autonomous, and resilient power system. Space-based power systems differ from Earth’s grid in generation sources, needs, structure, and controllability. This research introduces a groundbreaking approach employing digital twin (DT) technology to emulate and enhance the performance of a physical system representing a space-based system. The system encompasses three DC converters, a DC source, and a modular battery storage unit feeding a variable load. Rigorous testing across diverse operating points establishes the real-time high-fidelity DT, with root mean square error (RMSE) values consistently below 5%. The principal innovation leverages this DT to fortify system resilience against unforeseen events, surpassing the capabilities of existing controllers and autonomy levels. The approach offers an invaluable tool for scenarios where the system may not be primed for or physical access to components is limited. This research introduces a modular battery storage solution that seamlessly compensates for power shortages due to dust effects on the Lunar surface or unexpected system faults. This holistic approach validates the DT’s fidelity and underscores its potential to revolutionize system operation, safeguard against uncertainties, and expedite response strategies during unexpected contingencies. The proposed approach also paves the way for future development.
1 January 2025 in Digital Twins and Applications
AI‐driven control, particularly, Reinforcement Learning (RL), is becoming popular in modern power systems with increasing number of DC components. However, training RL‐driven controllers on actual systems may cause interruptions and/or create safety concerns due to random actions that may occur during training. This study evaluates the effectiveness of digital twins as a platform for training such control schemes. We develop an RL‐driven controller for a small‐scale DC microgrid using its digital twin and test it on an actual testbed. The digital twin enables safe and uninterrupted training, demonstrating improved learning efficiency and reduced risk compared to on‐system training. We compare the RL controller's performance with a PI benchmark to assess its advantages and limitations. Results show both benefits and constraints of this approach, contributing to the broader application of AI and digital twins in power system control. This study also highlighting the potential of digital twins in facilitating safe training and testing for developing advanced control schemes.
16 September 2025 in IEEE Access
The increasing shift towards decentralized energy systems have made microgrids (MGs) a key solution for improving energy reliability, resilience, and sustainability. However, managing the complexities of MG operations poses significant challenges. Traditional approaches often struggle to address these complexities, resulting in inefficiencies and compromised reliability. In response, this paper presents a digital twin (DT) framework designed for decision-making and real-time fault detection and diagnosis tested in a lab-scale DC microgrid. The DT model replicates the dynamics of the physical system (PS) under various fault conditions. An AI-driven Long Short-Term Memory (LSTM) model is integrated into the framework to improve health management. The performance of the system is assessed under both normal and emergency conditions using metrics such as Root Mean Square Error (RMSE) and confusion matrices. A comparative analysis between the DT and PS validates the DT’s ability to detect and diagnose faults accurately, demonstrating its potential to enhance microgrid operational resilience.
20 February 2024 in TechRxiv
Controlling the charging and discharging procedures of Lithium-Ion Batteries is of paramount importance as violating safety constraints, such as current deviations, can lead to significant damage to the battery or circuit or interruption in service. Thus, it is crucial to employ a robust controller capable of handling uncertainties and unexpected scenarios. PI controllers have become prevalent in recent years for managing battery dynamics, but they exhibit limited robustness in unpredictable situations. In this paper, we propose a Reinforcement Learning (RL) driven control method as a substitute for the PI controller. The agent is trained using a co-simulation approach with simultaneous employment of Python and Matlab, ensuring an accurate estimation of the environment and, consequently, enhanced performance. A prototype of the proposed controller is developed using dSPACE rapid control prototyper. The performance is compared with the benchmark controller (PI) across different fault scenarios, considering three criteria: overshoot, undershoot, and stabilization time. The comparative analysis reveals that, in most scenarios, the RL agent outperforms the PI controller, exhibiting a remarkable 50% reduction in both overshoot and undershoot compared to the benchmark controller. This research contributes to advancing battery control systems by introducing an RL-based controller that proves to be a more robust alternative, delivering improved performance in the face of uncertainties and fault scenarios.