With the rapid development of carbon and electricity markets, virtual power plants (VPPs) park have emerged as a novel energy management platform, playing a critical role in achieving low-carbon economic operations. A low-carbon economic optimization scheduling method is proposed for VPPs park under the uncertainties of electricity and carbon markets. A double-layer bidirectional long short-term memory (DBLSTM)-based model is developed to evaluate electricity and carbon price uncertainties. By constructing an electricity-carbon market-coupled optimization model, the study explores the classification modeling and differentiated control strategies of electric vehicle (EV) clusters, proposing a multi-energy system optimization scheme aimed at maximizing revenue. Simulation results demonstrate that the proposed method significantly enhances the system 's economic efficiency and carbon reduction capabilities, improves the flexibility of EV market participation, and achieves maximum comprehensive revenue under more realistic electricity-carbon market conditions, offering a novel approach for the efficient operation of VPPs.