Transformer winding state diagnosis method based on short-circuit impulse response and multi-feature joint

Abstract

Listen

Translate article

Abstract To investigate the relationship between the physical characteristics of transformer short-circuit impulse signals and winding deformation, this paper proposes a transformer winding state diagnosis method based on the combination of short-circuit impulse response and multiple features. Firstly, based on the relative displacement model of a single degree of freedom system, a winding short-circuit impulse response spectrum is constructed, and the energy entropy characteristics of different frequency bands are extracted using the Hilbert energy spectrum. Secondly, the energy spectrum is converted into a time-frequency matrix, and its energy density distribution and short-term energy attenuation changes are quantified. Then, we perform support vector regression (SVD) decomposition and calculate the proportion of singular value vectors. Finally, multiple feature quantities are combined to construct a feature set, which is then input into the SVD model optimized by the particle swarm algorithm (PSO) for diagnosis. The experimental results show that the proposed method can accurately extract the physical characteristics of short-circuit impulse signals, and the error between the obtained transformer winding deformation diagnosis value and the actual state value is only 3.6%, which can provide guidance for online monitoring of the transformer winding state.