Chaotic Attentive Recurrent Transformer Network for Intelligent Power Grid Fault Diagnosis

The main objective of this study is to enhance the intelligence level of power grid fault diagnosis systems to address increasingly complex fault scenarios and ensure the overall security, stability, and resilience of modern power grids. Traditional diagnostic methods often fall short in handling high-dimensional, nonlinear, and dynamic data generated in smart grid environments. To overcome these limitations, this research proposes a data-driven framework based on Deep Learning (DL), introducing a novel hybrid architecture called the Chaotic Attentive Recurrent Transformer Network (CARTNet). The proposed method begins with comprehensive data acquisition from various sources, including fault logs, real-time system parameters, weather data, and renewable energy outputs. The data undergoes preprocessing steps such as integration, cleaning, and advanced exploratory analysis to improve quality and extract latent features. CARTNet is specifically designed to model nonlinear dynamics and temporal dependencies in time-series data by synergistically combining chaotic system modeling with attention-based recurrent transformer mechanisms, allowing for more accurate and robust fault identification. The most important results are demonstrated through extensive simulations using Python, where CARTNet achieves a fault diagnosis accuracy of 99.88%, significantly outperforming conventional deep learning models. Its ability to learn complex patterns and adapt to diverse data inputs ensures reliable and timely fault detection. The significance of the obtained results is that CARTNet provides a powerful and scalable solution for intelligent fault diagnosis in smart grids, laying a strong technological foundation for the future of automated and resilient power system operations.