Transformer Multi-Feature Intelligence Protection Principle Based on BP Neural Network

Power transformer is one of the key equipment in power transmission and distribution system whose protection algorithm is of great prerequisite and guarantee for the safe and reliable operation. Through the rapid detection and identification of transformer faults by artificial intelligence technology, the protection device of transformer will not refuse to operate in case of internal faults, and will not misoperate under the conditions of inrush current, which can effectively improve the selectivity of transformer protection. This paper presents a transformer multi-feature intelligence protection principle based on BP Neural Network. Firstly, based on studying the operation characteristics of transformer, five characteristic quantities, i.e. initial closing phase angle, harmonic content, dead angle, waveform symmetry and equivalent magnetization curve, are adopted as the typical characteristics of transformer. Then, it is analyzed that the corresponding relationship between these characteristic quantities and the operation state of transformer. Secondly, BP Neural Network with typical characteristic quantities as input and operation state of transformer as output is constructed. Finally, PSCAD is used to verify the proposed algorithm. The results show that the proposed protection algorithm can accurately judge the operation state of the transformer, and the correct recognition rate of the simulation data reaches 100%.