Real-time feature signal extraction by online frequency sweep method for transformer winding deformation

The rate of transformer failures continues to increase, and the value of transformers themselves is expensive. Once an accident occurs, it directly or indirectly causes high losses to the power grid, so transformer detection is of great significance. The frequency response method is one of the important methods to monitor transformer winding deformation, and the difficulty lies in extracting effective response signals from complex noise backgrounds. To address this difficulty, this paper proposes a fast extraction algorithm for the online monitoring signal of transformer winding deformation, which realizes the real-time extraction of the winding deformation signal. The algorithm utilizes the high-frequency filtering characteristics and the low-pass DC characteristics of the dot product operation, and the amount of operations is much reduced compared with the traditional filtering algorithm, which can realize the real-time purification processing of the winding response signal.