Flexibility Trading from P2P Networks to Improve the Grid Frequency Response in the Presence of RTEM and MARL-Based AGC Systems

With the gradual shift from conventional units to renewable energy sources (RES), the modern grid is facing diminished frequency regulation (FR) capabilities, making traditional control methods less effective. In response, this paper proposes an advanced multi-agent reinforcement learning (MARL) control strategy for FR in a RES-dominated power system. The system includes a residential microgrid with diverse distributed energy resources engaged in peer-to-peer (P2P) energy trading at the distribution level, while automatic generation control (AGC) and a dynamic real-time energy market (RTEM) coordinate at the transmission level to balance load and generation. The RTEM dispatches energy every five minutes to support the FR process, complementing AGC operations. Following disturbances, the microgrid utilizes demand-side flexibility alongside AGC and RTEM operations to quickly restore system frequency to its steady state. MARL, implemented using the Deep Deterministic Policy Gradient algorithm, is applied to both AGC and microgrid systems. Here, P2P trading is facilitated through a consumer-prosumer contract participation matrix. The results show that flexibility from P2P networks, in conjunction with AGC and RTEM systems, significantly improves grid frequency response.