Akademik Veri Yönetim Sistemi
International Graduate Research Symposium – IGRS’22., İstanbul, Türkiye, 1 - 03 Haziran 2022, cilt.1, ss.485-495
Microgrids are structures consisting of renewable energy sources, energy storage units and local loads, which can operate simultaneously connected to the power grid when needed, or can operate independently and autonomously from the grid for economic and structural reasons. This autonomous operation feature of microgrids necessitates a strong energy management system infrastructure. Today, microgrids are sized using classical installation methods and controlled with expert system-based management systems. This type of management both limits the efficiency that can be obtained at the time of operation of the microgrid and has difficulty in following new trends in energy storage technologies. The idea that current energy storage technologies are used effectively in the system and that today's artificial intelligence technology forms the basis of the energy management system of microgrids constitute the importance of the study. An artificial intelligence (AI)-based energy management system should be able to predict the power generation of the solar power plant and the power generation of the wind turbines in the microgrid. In addition, it should be able to predict the dynamic price data and the load level to be requested from the microgrid. Accordingly, the system needs prediction models with high accuracy. In this study, solar power generation and wind power generation in a microgrid are forecasted by ensemble learning and other machine learning algorithms. While live price and load demand data are forecasted by deep learning algorithms, the performances of these prediction models are examined and compared. The suitability of all these prediction models for an AI-based energy management system has been investigated and analysed.
Keywords: