Generator Coherency Identification using a Machine Learning Based Technique

Authors

  • Ahmed Isnaadh , Ir. Jacqueline Lukose

Abstract

— The key aim of this project is to design and develop a machine learning based system for
optimizing the clustering of generators based on their coherencies. In this proposed method, PowerWorld was
used to simulate faults on an IEEE 9-bus system and acquire rotor angles of generators as data. Using 90 sets
of unique data, Adaptive Neuro Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) are
tested to determine a suitable technique, by comparing the Root Mean Squared Error (RMSE) values. It was
determined that the Triangular membership function was most suitable in constructing the ANFIS, and the
Levenberg-Marquardt function to train the ANN. Using these optimal configurations, the systems were
constructed, trained and tested. ANFIS showed a lower RMSE value of 0.02373, while ANN showed an
RMSE value of 0.32425. The training of ANN was almost instantaneous, mainly due to it being a simple
constructed network, while ANFIS required an average of 3.092 seconds to train with 3 epochs. Overall, the
ANFIS is the optimal machine learning technique to be implemented for clustering generators based on their
coherencies.

Published

2020-01-31

Issue

Section

Articles