Power swing prediction for out-of-step mitigation

Abstract

This paper explored the possibility of accurately predicting the classification of developing power swings. The notion of machine learning was employed, and tested the application of Decision Tree (DT) algorithms to wide area power system protection schemes. The novelty of the designed Wide Area Protection (WAP) scheme was portrayed by the WAP’s ability to adaptively and accurately predict the classification of developing successive power swings. DTs being a Data Mining (DM) technique, a transient stability analysis was performed on an IEEE 39 bus test system in Dig SILENT®. The learning sample from the Phasor Measurement Unit (PMU) data was organized and stored in a data base in Microsoft Excel® 2010. The CART analysis and DT model design was done using Salford Predictive Modeller-CART® v6, trial licence. The results of this investigation were quite accurate and gave DT algorithms ‘thumbs-up’ in terms of classification prediction.