PROGNOZA CENA NA TRŽIŠTU ELEKTRIČNE ENERGIJE KORIŠĆENJEM MAŠINSKOG UČENJA U PLEXOS SOFTVERU

Abstract

Machine learning is a part of artificial intelligence that uses specialized computer algorithms to create a model based on samples of input data. Once the model is successfully trained, it can be used to make predictions or decisions without being explicitly programmed to do so. In this paper, machine learning is used to predict electricity market prices in Western Australia, using publicly available data from the internet. The whole process was done in the PLEXOS software, in which machine learning was implemented from version 9.1. First, the available data that can influence the forecast value are being analysed, and then only a few are selected as input data for machine learning. The next step is the pre- processing of input data for use in PLEXOS and creation of a database in PLEXOS. Next, the machine learning model is trained with selected samples of historical data, and after successful training, the model is ready to forecast the selected value. Historical data was also taken as input for the forecast, so the results depend only on the quality of the machine learning model without the influence of forecast errors. The paper will show the procedure for selecting optimal input data and changing in model accuracy when changing the set of parameters used for training the model.