Aiming at the virtual power plant multi-body interest game and flexibility resource allocation problem, this paper proposes a two-layer robust optimization method for virtual power plant master-slave game considering mobile energy storage. By constructing a Stackelberg game framework between the virtual power plant operator and multiple subjects, the interaction mechanism between the dynamic pricing strategy and the cooperative scheduling of resources is revealed, and electric vehicles are embedded into the model as flexibility resources to optimise the peak and valley regulation of loads. Robust optimisation is adopted to deal with electric vehicles charging and discharging uncertainties, and Bayesian algorithm is applied to solve the master-slave game model to improve the robustness of the scheduling scheme. Case validation shows that the model can effectively coordinate the interests of multiple subjects, achieve the optimal allocation of electric vehicles flexibility resources, and provide efficient and robust technical support for the economic dispatch of virtual power plant under the power market environment.