Fast and Accurate Estimation of Power System Loading Margin using Extreme Learning Machine with Reduced Input attributes

This paper presents an Extreme Learning Machine (ELM) approach for a fast and accurate estimation of the power system loading margin for multiple contingencies with reduced input attributes. Active and reactive power flows of all load buses are chosen as the input features to the ELM. The training data for the ELM model are generated by using the Continuation Power Flow (CPF) method. The proposed method is Mutual Information (MI) based dimensionality reduction technique to reduce the input dimension and for improving the performance of the developed network with less training time, which makes ELM approach applicable to large scale power system. IEEE 30-bus system and IEEE 57-bus system are considered for the demonstration of an effectiveness of the proposed methodology under various loading conditions on the single line contingencies. Simulation results validate the proposed ELM with reduced input features for fast and accurate on-line voltage stability assessment.