Search

This paper aims to propose a novel data-driven approach for exploring the partial discharge inception voltage (PDIV) of turn-to-turn insulation in inverter-fed motors. To this end, the gas discharge theory-based finite element method is investigated to simulate the state of turn-to-turn insulation, and a comprehensive database is established concerning the structural and environmental factors. Subsequently, a bisected deep belief network (BDBN), which combines with deep belief networks (DBN) and the bisection method, is proposed to explore PDIV. Specifically, DBN is designed to distinguish the discharge state of turn-to-turn insulation, and the testing applied voltage is adaptively adjusted to explore PDIV by the bisection method. The experiment result demonstrates the effectiveness and reliability of BDBN in exploring PDIV of turn-to-turn insulation in terms of precision. [ABSTRACT FROM AUTHOR]