Price forecasting in the electricity spot markets with artificial intelligence and time series models

The increasing energy demand in the global market has caused a number of rapid reactions within our country. As a result of the reactions, a new structure has emerged in which all the influencing and affected factors in the market are together. This situation has led to privatization and competition in the economic dimension of the business. It is necessary to address this in an efficient and sustainable way in market conditions while meeting the energy need. Energy production methods, grid systems, distributed energy systems and the control of these systems, which can vary rapidly, have brought about a versatile structure. It is aimed to present the inferences obtained with modern analysis methods to the bidders in a way that will eliminate the gaps in the literature and the market, and then to provide the necessary environment for the determination of the approximate costs and plans for the institutions and persons that will serve the end user. Markets have to evaluate many criteria to ensure the optimum price. Accurate analysis will be a great reference for risk management in the spot market. In parallel with the developments in the world, the electrical energy market in Turkey is also divided into different sub-markets. In the study, short-term day-ahead and intraday price inferences were made for electricity spot markets by using traditional and machine learning-based models of time series with the data obtained from the EPİAŞ transparency platform. Then, the models were turned into a web-based application for the market user to dynamically repeat these transactions. As a result of the literature comparison and tests conducted with transparent data, it has been determined that the best model for the Intraday Market is ARMA; and the best model for the Day Ahead Market is Prophet. The work will be used for commercial activities, and it is also intended for obtaining multiple international academic papers and patents.