A particle swarm optimiser with passive congregation approach to thermal modelling for power transformers

This paper employs an intelligent learning technique based on a particle swarm optimiser with passive congregation (PSOPC)lgorithm to identify the thermal parameters of a simplified thermoelectric analogous thermal model (STEATM) for transformers, based upon only a few on-site measurements instead of experimental methods. The model outputs deliver good agreements with the on-site data based upon a single set of parameters obtained from the PSOPC learning with a fast convergence rate. The simulation results are compared with that obtained using an artificial neural network (ANN) approach.