Energy Management of Networked Microgrids with Real-Time Pricing by Reinforcement Learning

Coordinating the microgrids (MGs) in distribution network is a critical task for system operator (DSO), which could be achieved by setting prices as incentive signals. The high uncertainty of loads and renewable resources motivates DSO to adopt real-time prices. MGs require reference price sequences long time horizon advance make generation plans. However, due privacy concerns practice, may not provide adequate information build closed-form model. This causes challenges implementation conventional model-based methods. In this paper, framework coordination through proposed. bi-level framework, sets signals, based on charging plan. model-free reinforcement learning (RL) applied optimize pricing policy when response behavior unknown DSO. To deal with large action space problem, incorporated into RL algorithm efficiency improvement. numerical result shows that minimized cost obtained developed close method while private preserved.