A Deep Learning Sample Set Construction Method for Commutation Failure Suppression of LCC-HVDC

The traditional fault diagnosis of AC / DC hybrid lines is based on the experience judgment of abnormal electrical quantities, which requires high professional and technical requirements for practitioners and is time-consuming and laborious. With the further development of artificial intelligence, deep learning has attracted the attention of relay protection practitioners with its powerful data processing, complex pattern recognition and nonlinear relationship modeling capabilities. However, the existing fault data cannot support the construction of data sets, which has become a major obstacle. In this paper, a single-ended equivalent model applied to LCC-HVDC lines is proposed, which aims to construct a sample set supporting training in a simulation manner. Compared with the conventional transmission model, the calculation speed is faster under the premise that the accuracy meets the requirements.