ML Based Anomaly Detection in VSC

Abstract

This thesis provide novelty by establishing simple and iterative workflow which can be further optimized to be adopted in actual site operation for achieving an extra layer of safety in a VSC operating in smart grid. The utilization of RF and SVM classifiers for analysing data yielded by emulation of actual hardware also showcase the constraints that are not directly related to concurrent ML technology but can prove to be the limitations for adoption of discussed study in real time system and are discussed accordingly in the implementation section. Finally, this thesis is concluded by reporting the advantages of SVM over RF observed through the set performance indicators and some areas where RF classifier outperforms the SVM ML model outputs.