Multichannel consecutive data cross-extraction with 1DCNN-attention for diagnosis of power transformer

TL;DR

This paper introduces a novel multichannel data cross-extraction method combined with 1DCNN-attention for power transformer diagnosis, effectively utilizing sequential data for improved condition assessment.

Contribution

The paper proposes the MCDC structure and 1DCNN-attention mechanism, enhancing feature extraction and diagnosis accuracy for power transformers using multichannel sequential data.

Findings

MCDC outperforms existing algorithms in transformer diagnosis.

1DCNN-attention demonstrates superior stability and efficiency.

The approach shows strong generalization on real operational data.

Abstract

Power transformer plays a critical role in grid infrastructure, and its diagnosis is paramount for maintaining stable operation. However, the current methods for transformer diagnosis focus on discrete dissolved gas analysis, neglecting deep feature extraction of multichannel consecutive data. The unutilized sequential data contains the significant temporal information reflecting the transformer condition. In light of this, the structure of multichannel consecutive data cross-extraction (MCDC) is proposed in this article in order to comprehensively exploit the intrinsic characteristic and evaluate the states of transformer. Moreover, for the better accommodation in scenario of transformer diagnosis, one dimensional convolution neural network attention (1DCNN-attention) mechanism is introduced and offers a more efficient solution given the simplified spatial complexity. Finally, the effectiveness of MCDC and the superior generalization ability, compared with other algorithms, are validated in experiments conducted on a dataset collected from real operation cases of power transformer. Additionally, the better stability of 1DCNN-attention has also been certified.