Cost estimation of transformer main materials using Artificial Neural Networks

This paper presents a method of calculating the cost of transformer main materials using Artificial Neural Networks. In initial phase of project, having knowledge of estimated weights and hence subsequently, the cost in a short time can prove to be beneficial. As the major amount of transformer cost depends on the weight of its main materials, its cost estimation is of vital importance. A Multi-Layer Perceptron (MLP) neural network has been proposed for predicting the weights of aluminium, iron and transformer oil (which are neural network outputs). Once the weights of main materials are known the total cost can be calculated. The inputs to MPNN are kVA rating, maximum flux density, maximum current density and volt per turn. A MATLAB program has been developed to train the neural network. The data required for training the neural network has been obtained from the transformers made by Jyoti Transelect Company, Bhuj, India.