Demand forecasting using artificial neural networks for power transformers

Beholding the historical processes, the human instinct to predict the future has always been dominant. In the modern era, scientific and mathematical models have given civilizations an important tool in predicting the future. By the last quarter of the 20th century, artificial neural network studies gained momentum and gave a new perspective to forecasting techniques. Demand forecasting is one of the forecasting issues that come to the fore with industrialization. In this study, a demand forecast for power transformers has been carried out with the multilayer perceptron method, a type of artificial neural network. The factors affecting the demand have been determined by taking an expert opinion from a regional producer, and the data covering past 40 years regarding these factors were obtained from official sources. The artificial neural network created during implementation has produced consistent results at the rate of 97.99%. It was observed that it outperformed the ARIMA model created in the study for comparison. It was concluded that the multi-layer artificial neural network created in the study is a useful tool for demand forecasting of power transformers.