Characterization of Artificial Neural Network based Power Transformer Protection

This work strives to develop a high-performance power transformer protection system using Artificial Neural Network (ANN). In achieving this, enhanced back propagation algorithm, the calculation ability of Artificial Neural Network (ANN) and its capability to accurately identify non-linear system were harnessed. A combination of two multi-layer neural networks was used in the implementation of the protection system. The first network was used in predicting the transformer oil temperature using the transformer load current as the input. The second ANN network is used in implementing the classification of the power transformer operating conditions which includes normal and fault conditions. This network has the oil temperature, oil level and dielectric strength parameters of the transformer as its input while after identification and classification gives out at its output the transformer dynamic operation condition. These outputs are Normal operation, Overheating, Dielectric degradation and Critical conditions. The critical condition through a trip logic trips off the power transformer. A total of 1500 data from a working 15MVA power transformer were used in the training of the networks. 70% of the data was used in training, 15% was used in validation while the remaining 15% was used in testing the performance. Both ANNs showed remarkable results. The ANN-1(TEMPEST) was able to predict temperature to an accuracy level of 99.99% while the ANN-2(CLASSFAULT) was able to give a 100% operating condition of the power transformer. These results attest to the fact that ANN could be used effectively and efficiently in the protection of power transformers.