Single-Phase Ground Fault Identification Model via Feature Extraction and AdaBoost Model

The paper presents single-phase grounding fault calculation algorithm is prone to influences from operation modes and configuration parameters. To solve such problems, this paper proposes a novel calculation method based on physical model machine learning compound drive. First, the feature project is established based on the zero-sequence circuit. Second, to solve the different number ratio of actual failure and non-failure samples, and "dimensions of disaster" of feature engineering brought about by actual situations, the up-sampling and principal component dimensionality reduction techniques are introduced to realize high-dimensional space equivalent representation of feature engineering. Further, by combining AdaBoost machine learning and receiver operating characteristic curve, assessment effect is improved. The PSACD simulation validates the accuracy of the proposed physical-data based approach.