Transformer Health Index Sensitivity Analysis using NeuroFuzzy Modelling

K. Ibrahim; R.M. Sharkawy; H.K. Temraz; M.M.A. Salama

Authors

K. Ibrahim Department Electrical and Control Engineering, Faculty of Engineering, Arab academy of Science and Technology (AAST), Cairo, Egypt
R.M. Sharkawy Department Electrical and Control Engineering, Faculty of Engineering, Arab academy of Science and Technology (AAST), Cairo, Egypt
H.K. Temraz Department of Electrical Power and Machines, Faculty of Engineering, University of Ain Shams, Cairo, Egypt
M.M.A. Salama Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Canada

Keywords:

Asset management, condition monitoring, Health Index (HI), transformer tests, Particle Swarm Optimization (PSO), Self-Adaptive NeuroFuzzy Inference System (ANFIS)

Abstract

In this paper a transformer Health Index (HI) sensitivity analysis study is presented. A HI prediction model is developed using a Self-Adaptive Neuro-Fuzzy Inference System (ANFIS). The ANFIS model is tuned using the Particle Swarm Optimizer (PSO). The utilized measurements are a combination of actual field measurements including Carbon Monoxide (CO), Acetylene (C2H2), Ethane (C2H6), Interfacial Tension (IFT) and Furans content (FFA) for 724 working transformers within a network of an industrial facility. Results show that the PSO based ANFIS model is capable of obtaining good and reliable results. The model response is tested and it is able to predict the HI numerically with high accuracy. Furthermore, it was found that the model yielded a
good response in predicting HI change with respect to the change of transformer’s measurement values.