Optimal Scheduling Strategy for Port Virtual Power Plant Based on DQN Algorithm: Considering the Dual Impact of Electricity–Carbon Trading

Abstract In order to address the challenge of cooperative optimization of port virtual power plants (VPPs) participating in both electricity and carbon trading markets, this paper proposes a two-stage day-ahead–intraday optimal scheduling strategy using the deep Q -network (DQN) algorithm strategy, which can achieve both economic and environmental benefits for the port. First, a physical model of a port VPP was constructed in which the wind and photovoltaic (PV) power, energy storage systems (ESS), and multiple types of loads are integrated. Second, a detailed carbon accounting model was developed to refine current carbon accounting methods and quantify in detail the carbon emissions from the port’s energy system, i.e., gas-fired power generation, grid electricity purchases, and other sources. Based on this, a two-stage scheduling framework is proposed. During the day-ahead stage, the DQN algorithm is employed to derive the optimal day-ahead electricity sales declaration curve. In the intraday stage, according to real-time deviations, the DQN algorithm is reapplied to dynamically adjust the charging and discharging power of energy storage and the output of gas-fired units, while incorporating an ε -greedy strategy to address uncertainties. Finally, the effectiveness of the developed method was verified with a comprehensive set of case studies. Experimental results show that the developed method can achieve considerable economic benefits as well as remarkable reduction of carbon emissions. The DQN algorithm has strong adaptability to renewable power fluctuation and load uncertainty, which provides an effective solution for port VPP in multimarket cooperative trading.