A feature based distributed machine learning for post fault restoration of a microgrid under different stochastic scenarios

摘要

Stochastic nature of a large scale wind power plant in a power system with insufficient load margin has significant impact on a post fault network. Such probabilistic character of a system makes it quite a challenge to maintain post fault system stability. A short circuit fault under such contingency may introduce power system oscillation resulting in massive voltage fluctuations. One probable solution is to develop a corrective voltage control (CVC) framework in order to maintain sufficient load margin. Standard CVC measures are based on active and reactive dispatch from generating units. However, in a post contingent scenario it is often critical to select appropriate parameters for CVC. This study implements an offline-online data analysis approach using feature selection and machine learning algorithms, as a mean to develop an accurate CVC framework based on supervisory machine control.