Artificial Neural Network-based Flexibility Assessment of the Electric Power System with High Wind Energy Penetration

Abstract

One of the ways to resolve the environmental protection issue is to increase electricity production based on renewable energy sources (RES). High penetration of RES in an electric power system (EPS) induces great uncertainty in EPS operation. In this case, the normal and secure operation of EPS requires information on its flexibility. Algorithms for determining the EPS flexibility in real time are developed in the study. The acceptable time of obtaining the results is ensured by splitting the flexibility computing process into online and offline ones. Offline computation of operating state with minimum flexibility relies on an exact method. Online calculation of operating state with minimum flexibility employs trained artificial neural networks (ANNs). The two machine learning techniques (supervised and unsupervised) are explored and compared. These are the feedforward neural network and Kohonen neural network. In this study, the ANN architecture which fits in with solving the considered problem most is chosen. A new way of interpreting the ANN response is presented. To this end, another ANN is trained to recognize the operating state with minimum flexibility.