Accurate fault localization in power transmission line with wind penetration using variational mode decomposition and CNN-GRU architecture

• Fault localization using VMD with CNN–GRU achieves MAPE of 0.842 %. • MATLAB/Simulink data mimic errors and missing data for robust testing. • DFIG wind farm impact on fault location analysed under multiple cases. The growing integration of renewable energy sources (RES), particularly wind power, introduces significant challenges for the reliable and stable operation of modern transmission systems. Faults on transmission lines can cause severe disturbances, and conventional localization methods often face reduced accuracy due to measurement noise, missing data, and the nonlinear dynamics of RES. To address these challenges, this study proposes a novel hybrid intelligent framework that uniquely integrates Variational Mode Decomposition (VMD) with a Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU). The key advantage of this model lies in using VMD to decompose complex, non-linear fault signals into robust modes, from which the CNN-GRU architecture then efficiently extracts spatial-temporal features, thereby significantly enhancing localization accuracy even under these adverse conditions. A large-scale dataset was developed using the IEEE 24-bus system in MATLAB Simulink, incorporating three Type-III wind farms. The dataset explicitly considers noise and missing data to replicate realistic measurement environments. The results show that the proposed VMD-CNN-GRU framework achieved outstanding performance, with a mean absolute error (MAE) of 0.407 km and a root mean square error (RMSE) of 0.617 km for a 100 km transmission line, which significantly outperforms conventional deep learning approaches.