Transformer insulation diagnosis using hybrid artificial intelligence-modified particle swarm optimisation

A hybrid neural network (NN)-modified particle swarm optimisation technique was proposed in this work for determination of the fault type in power transformer insulation system. The artificial NN (ANN) characteristics were optimised by modified particle swarm optimisation (PSO) with coefficient of time-varying acceleration (TVAC). The improvement of the modified PSO-TVAC over other methods including ANN without optimisation, the existing diagnosis methods and previous works was validated by comparison between each method. It was observed that the proposed hybrid ANN-modified PSO-TVAC managed to achieve the largest correct fault type identification than other methods. Hence, this shows that the proposed method can be applied in fault type determination of actual power transformers.