Identification of excessive neutral-to-ground voltage in secondary distribution system using deep learning method.

Excessive neutral-to-ground voltage (ENTGV) in power distribution systems poses a critical challenge to the integrity and reliability of electrical networks. This thesis undertakes a comprehensive exploration to address this issue by focusing on model development, factor classification, and localization techniques. A detailed electrical circuit model is developed to characterize a normal neutral-to-ground voltage (NTGV) profile within a secondary distribution system (SDS), taking into account load conditions, grounding components, and the incorporation of ground return current. The model serves as a benchmark for understanding baseline NTGV behaviour and is intended for validation using future real-world data. Its performance is rigorously evaluated against existing models by using empirical measurement data, demonstrating improved alignment with observed system behaviour. To classify the contributing factors of ENTGV, a deep learning (DL) approach is proposed, leveraging raw waveform inputs without the need for manual feature extraction.