QI-ResAFNet: Deep Learning Architecture for Intelligent Fault Diagnosis in Integrated Power Transmission System

The integration of energy sources into modern power transmission networks generated new challenges for accurate and adaptive fault diagnoses due to increased operational uncertainty, nonlinear disturbances, and varying data values. This study proposes a novel hybrid intelligent diagnostic architecture using a Quantum-Infused Residual Attention Fusion Network (Qi-ResAFNet) for high classification accuracy across various operational grid environments. Prior to extracting features, a systematic preprocessing of raw transmission-level dataset is performed including data cleansing, evaluation of data quality, and normalization prior to carrying out feature creation process. A comprehensive Exploratory Data Analysis (EDA) is then conducted to determine statistical characteristics, evaluate fault vs non-fault physical separation, and establish correlation patterns among the datasets. The feature engineering process is employed to create distinctive attributes from electrical signals in both time and frequency domains, thus providing improved representation of the signals for classification. The proposed Qi-ResAFNet is a new combination of quantum feature encoding, dual-attention fusion and deep residual network design to capture complex spatiotemporal correlations and quantum-enhanced interactions among features. The automated hyperparameter adjustment of the Heron Optimization Algorithm (HOA) significantly improve the generalization and convergence performance of the proposed architecture. Based upon the results produced by Python simulations of many fault-types under fluctuating energy supply, the proposed architecture achieves the highest accuracy, precision, recall and F1-Score of 0.99. This demonstrates that the proposed architecture provide rapid, reliable and flexible diagnostic capabilities of future intelligent transmission grids.