Holistic mutual benefits aware P2P2G market among microgrids in a distribution network: A decentralized data-driven approach
Xiao Liu, Sinan Li, Cuo Zhang, Meng Liu, Jianguo Zhu
Abstract
In contrast to the traditional peer-to-grid (P2G) market, the emerging decentralized peer-to-peer-to-grid (P2P2G) trading can generate enormous potential to reduce the overall operational costs of microgrids (MGs) further. However, it is challenging to incorporate this decentralized market framework directly into the distribution network (DN) trading framework to account for mutual benefits holistically, impeding progress toward future smart electricity markets. This paper proposes an online non-iterative method based on data-driven multi-agent deep reinforcement learning. The decentralized P2P2G trading framework is formulated as partially observable Markov games (POMGs) to consider mutual benefits efficiently and make it compatible for DN operations. It is further integrated with a novel adaptive margin update (AMU) method to protect DN's topology information and return differential rewards to improve training efficiency and operation safety. Comprehensive numerical simulations on a modified IEEE test system demonstrate the superiority of the proposed method, outperforming other data-driven algorithms and a model-based optimization approach in smart electricity market applications. • Proposes a novel decentralized transactive energy framework for P2P2G trading. • Develops a comprehensive approach that simultaneously maximizes economic benefits and ensures the operational safety. • Adaptive Margin Update (AMU) Method: Introduces a AMU method that enhances training by providing differential rewards • Privacy-Preserving Techniques: Utilizes surrogate distribution network representation and critical bus selection techniques to minimize information exposure.