Çok ajanlı sistem ve takviye öğrenimine dayalı akıllı şebeke enerji yönetimi

The energy sector is undergoing fundamental shifts the depletion of fossil fuels and considerations for the environment have allowed it to use renews like solar and wind energy resources. A micro- grid constitutes a component of an intelligent grid and is ready to play an important role in the generalization of renewable resources. However, because renewable energy is intermittent in nature, it affects the dynamics and stability of the micro-grid and thus new approaches for coordination and control are required when integrated in the micro-grid. The existing systems lack run-time adaptive behavior and suffer from communication overhead. To meet these challenges and to achieve an optimal balance between generation, energy storage and load demands, we need to incorporate efficient communication and control strategies into micro-grid monitoring. In this thesis a, Reinforcement learning is implemented for optimal energy management of a micro-grid. It is extended into Multi-Agent Reinforcement Learning (MARL) for distributed optimization of micro-grids. The performances of the reinforcement learning methods are compared with conventional methods. The effectiveness of MAS and MARL are investigated in micro-grid for autonomous adaptation in the dynamic environment, and for economic and environmental optimization.