Digital Twin Paradigm for Fault Ride Through in Grid-Integrated Distributed Generation

The integration of distributed generation (DG) systems into the power grid presents both opportunities and challenges for grid stability. Ensuring smooth DG operation during grid anomalies, such as islanding events and faults, is vital for a stable power supply. To address these concerns, a comprehensive approach by combining islanding detection based on transfer learning with fault ride-through mechanisms using Model Reference Adaptive Control (MRAC) is proposed in this paper. The methodology is developed and evaluated through simulations of a grid-connected DG system. The performance of the islanding detection system is thoroughly assessed across various scenarios, including normal grid conditions and isolated instances. The proposed model, trained with open neural network exchange (ONNX) and transfer learning, achieves an average accuracy of 98.7%, validated with unseen data to yield 96% accuracy with minimal misclassifications. This study also examines the response of the grid-connected DG system to simulated faults and its interaction with an MRAC controller. Analysis of post-fault behaviors provides insights into aspects like AC current, DC-link voltage regulation, and power transfer dynamics, offering a valuable understanding of system resilience and performance during challenging scenarios.