Machine learning-based real-time outage fault detection for distribution networks in smart grid

Smart grid distribution networks are crucial for managing power dispersion, peak demand, and outage recovery. However, their integration with consumer networks and the main power grid makes them prone to faults caused by hardware failures and erroneous data inputs. This study addresses the challenge of detecting such faults by introducing a novel outage fault-detection scheme based on soft generative learning (SGL). SGL combines soft computing and generative adversarial networks to detect distribution faults by modeling power surges and outages through generator and degenerator components. The generator maps expected peak surges, while the degenerator identifies unexpected outages and their propagation across the grid. Fault detection is achieved by analyzing the imbalance between these two functions over time intervals. The proposed method performs cyclic computations to enhance the accuracy and stability of the detection process. It classifies faults as surge outages or distribution failures using distribution maps generated by SGL. Experimental evaluation shows that the proposed scheme improves detection accuracy by 12.38 % in hours, 13.91 % in generated power, 13.62 % in demand, and 12.58 % in device count under time-based assessment. These results demonstrate the effectiveness of the proposed approach in enhancing real-time fault monitoring and operational efficiency in smart grid distribution networks. • Proposes a fault detection method using soft generative learning (SGL). • Detects outage faults via generator–degenerator imbalance modeling. • Improves fault detection accuracy under power and time factors. • Enhances smart grid stability through real-time surge analysis.