Guoxing WU / Shenzhen Power Supply Bureau Co. Ltd.
Peng WANG / Shenzhen Power Supply Bureau Co. Ltd.
Yonghao REN / School of Electrical Engineering, Southwest Jiaotong University
Yuanda SONG / School of Electrical Engineering, Southwest Jiaotong University
Sheng Lin / School of Electrical Engineering, Southwest Jiaotong University
Characteristics of transformer magnetizing current which is difficult to obtain can be uesed to assess transformer under DC bias effectively. Considering that the distortion of magnetizing current is directly affected by the DC current in the transformer neutral point,it is critical to research on fitting calculation method based on neural network which taking DC current as input and several characteristics of magnetizing current as output. Firstly, typical characteristics of magnetizing current that reflect distortion effectively are selected as output of the fitting calculation. Then, the neural network algorithm is used to fit the DC current and several typical characteristic of magnetizing current respectively, and the corresponding relationship between the DC current and the magnetizing current is analyzed. Finally, a transformer simulation model has been built to verify the effectiveness of this fitting calculation method which provides the calculation basis for evaluating the state assessment of transformer under DC bias.