A digital twin based forecasting framework for power flow management in DC microgrids

www.nature.com/scientificreports OPEN A digital twin based forecasting framework for power flow management in DC microgrids Kerry Sado1, Jarrett Peskar2, Austin Downey2,3, Jamil Khan2,3 & Kristen Booth1 The ability to forecast system conditions is integral to the definition and functionality of digital twins. While forecasting methods have been explored for use in digital twin systems, the integration of feedback mechanisms for real-time forecasting and in-situ decision-making in DC microgrids has not been extensively investigated. This research develops a modular forecasting framework tailored for digital twins in DC microgrids to enable real-time monitoring, online forecasting, and decision-making. DC microgrids, characterized by dynamic load variations, benefit from advanced predictive capabilities to maintain stability and operational efficiency. The proposed digital twin-based forecasting framework addresses these challenges by providing real-time predictive insights based on dynamic system conditions and a forecasting window defined by a decision-maker, facilitating proactive management strategies. Leveraging real-time sensor data, the digital twin forecasts system behavior under varying load conditions, enabling proactive management through real-time decision-making within operational constraints. As a proof of concept, the framework incorporates an electro-thermal digital twin designed to manage power flow based on thermal constraints in power distribution cables. Experimental validation using a simplified three-bus DC microgrid testbed demonstrates the effectiveness of the framework in enabling timely adjustments to power flows and preventing thermal overloads. Keywords Digital twin, Forecasting, Power systems, Electric ships, Decision-making A digital twin (DT) is a comprehensive digital replica of a physical asset that integrates multiphysics, various scales, and probabilities to mirror and forecast the life-cycle of its counterpart1. Essentially, the digital twin acts as a comprehensive and faithful representation of a physical system or subsystem, referred to as the physical twin2. In this context, a physical twin refers to an actual real-world entity which can vary in complexity. It might be a machine, a piece of infrastructure, a single system component, or an entire complex system. The physical twin encapsulates specific attributes, characteristics, and functionalities that define the entity. The definition of digital twins is often misused in the literature with many studies failing to emphasize the feedback loop that differentiates digital twins from traditional models. The bidirectional flow of data between the physical asset and its digital counterpart is integral to the digital twin paradigm as it provides real-time insights, updates, and enabling dynamic decision-making3,4. This mechanism enables the digital twin to adaptively update its state in real-time, driven by sensor data from the physical twin. Furthermore, the bidirectional nature of the feedback loop allows the digital twin to receive information about the current operational parameters from the physical twin, enabling synchronized adjustments in the virtual representation. The bidirectionality also allows the decision-making entity within the digital twin to execute instructions on the physical twin based on forecasted insights generated by the virtual representation. This capability ensures synchronized operations and dynamic adjustments. Without the bidirectionality, a model cannot be considered a true digital twin but rather a static digital representation. The omission of this crucial aspect in research undercuts the potential benefits and applications of digital twin technology5. The significance of digital twins is rapidly gaining recognition from both academic and industrial sectors6. In the automotive industry, digital twins are used to monitor, simulate, and optimize production and operational 1Department of Electrical Engineering, University of South Carolina, Columbia, SC, USA. 2Department of Mechanical Engineering, University of South Carolina, Columbia, SC, USA. 3Department of Civil and Environmental Engineering, University of South Carolina, Columbia, SC, USA. 4This work was supported by the Office of Naval Research under contracts N00014-22-C-1003 and N00014-23-C-1012. “The views expressed are those of the authors and do not reflect the official policy or position of the Department of Defense or the U.S. Government. DISTRIBUTION STATEMENT A. Approved for public release distribution unlimited. Approved, DCN# 2025-1-31562”. email: Scientific Reports | 2025 15:6430 | https://doi.org/10.1038/s41598-025-91074-0 1 www.nature.com/scientificreports/ performance7. In the energy sector, general electric (GE) has implemented a digital twin for a wind farm, enhancing reliability by gathering real-time data on weather, performance, and service8. Similarly, Siemens developed a digital twin for the Finnish power grid, leading to improved safety, reliability, and resource savings9. Digital twin technology has applications beyond industrial and infrastructure sectors. In healthcare, digital twins are being explored to model cancer patients, providing patient-specific clinical decision support and enabling large-scale virtual clinical trials10. In manufacturing, digital twin-driven systems facilitate the integration of cyber-physical systems for smart workshops, enhancing efficiency and customization capabilities11. Recent studies also highlight the role of digital twins in renewable energy. For example, Sehrawat et al.12 proposed a machine learning-based digital twin framework for solar irradiance forecasting, demonstrating its capability to enhance operational efficiency and predict performance in renewable energy systems. Similarly, Guo et al.13 presented a digital twin framework integrating fuzzy logic for wind energy forecasting, offering high accuracy and robust predictions over long-term horizons. These studies illustrate the adaptability of digital twin frameworks to address dynamic challenges in energy systems. Several studies, including those by Nwoke et al.14 and Di Nezio et al.15, commend digital twins for their effectiveness in monitoring and predictive maintenance, specifically for power electronic converters. These studies focus on reducing data latency and enhancing system parameter estimation through sensor data. However, they primarily address component-level applications and fall short of extending their findings to system-wide forecasting or proactive power management. Wileman et al.16 developed a component-level digital twin to monitor the health of power converters using physics-based models. This demonstrates the precision that digital twins can achieve at the component-level. However, it also highlights a significant research gap in applying digital twin technology to broader, system-wide applications, where the ability to forecast across the entire sys (...truncated)