Islanding detection method for microgrids based on CatBoost

The occurrence of unintentional islanding will seriously threaten the stable operation a microgrid (MG). Therefore, detecting an timely is important premise to ensure operates safely and stably. However, problem dead zone exists in traditional detection process because threshold various electrical feature quantities point common coupling (PCC) cannot be determined effectively. To solve this problem, method based on CatBoost proposed for microgrid. novelty lies two aspects: 1) reduce error brought by with weak correlation detection, analysis Spearman coefficient used extract closely related detection. 2) determine more accurately island integrated learning machine dig out correlations between performance verified modified IEEE13-bus system. results example verify that can achieve higher accuracy cases grid-connected interference line faults.