Electricity demand forecasting with artificial intelligence algorithms: The case of Tekirdağ, Çerkezköy

Electricity is widely used in industry and various other fields today. In industrial applications, electricity is utilized for cooling, heating, lighting, electronic, and mechanical operations depending on workforce requirements, the scale of the production site, and seasonal conditions. In this thesis study, a case analysis of the Tekirdağ, Çerkezköy region was conducted to perform short-term demand forecasting using the least squares method, k-nearest neighbors, decision tree, support vector machine, and artificial neural networks algorithms. In this thesis, hourly consumption data for the Tekirdağ, Çerkezköy region covering the period from January 1, 2021, to December 31, 2023, were obtained. Hourly meteorological data for the same period were provided from the Meteoblue database. To ensure that the forecast outputs are interpretable and comprehensible in graphical form, the dataset was reduced. For the least squares and k-nearest neighbors algorithms, a 72-hour dataset covering the period from May 1, 2021, 00:00 to May 4, 2021, 00:00 was used. For the decision tree, support vector machine, and artificial neural networks algorithms, a 240-hour dataset covering the period from April 24, 2021, 00:00 to May 4, 2021, 00:00 was utilized. The 240-hour dataset was divided into 70% training and 30% testing sets. The results obtained in MATLAB were analyzed based on error metrics, including mean absolute percentage error, correlation coefficient, and root mean square error. Following these processes, the methods used in the study were ranked based on their performance, and the method with the lowest error rate was identified.