ANN design for fast security evaluation of interconnected systems with large wind power production

Abstract

This paper presents the performed steps to design an Artificial Neural Network (ANN) tool, able to evaluate, within the framework of on-line security assessment, the dynamic security of interconnected power systems having an increased penetration of wind power production. This approach exploits functional knowledge generated off-line, the Linear Regression (LR) variable selection stepwise method to perform automatic Feature Subset Selection (FSS) and ANN to provide a way for fast evaluation of the system security degree. In order to choose the best input/output set of variables for the ANN tool, a comparative analysis is performed, regarding the obtained predicting error, by performing a statistical hypothesis test. The reduced error results confirm the feasibility and quality of the derived security structures.