Artificial neural network for transformer insulation aging diagnosis

Abstract

Transformer insulation aging diagnosis is important for over all power industry. It is difficult and complicated to assess transformer insulation ageing status. In China, with the vigorous development of new technology, high performance liquid-phase chromatography (HPLC), recovery voltage method (RVM), polarization/depolarization current (PDC), degree of polymerization (DP) and Dielectric Loss tgo, etc are adopted widely in transformer insulation aging diagnosis. They belong to high-new field, such as furfural concentration and polarization current (PDC). This paper describes the structure and specific features of transformer insulation ageing diagnosis artificial neural network (TADANN)in detail. It describes the structure and specific features of TADANN in detail. TADANN has great potentialities. Its design includes selection of input, network topology, synaptic connection weight and output. There are ten feature elements (i.e., DP, C5H4O2, PDC, RVM, CO, CO2, tgo, etc) in its input layer. A four output nodes network is adopted, with normal, oil aging, solid insulation aging, and transformer life nodes. It can be used for diagnosis transformer insulation ageing & on-line monitoring & life-assessment. The feasibility and effectiveness of TADANN for transformer insulation ageing diagnosis are explained by some examples.