Aiming at the key challenges of source-load uncertainty

complexity of multi-energy coupling

and insufficient load-side regulation potential faced by multi-energy virtual power plant (MEVPP) in the operation process

this paper proposes a distributed model predictive control strategy based on digital twins. First

a dynamic digital twin resource aggregation model is constructed by combining deep neural network and improved k-means spectral clustering technique

which effectively realizes the dynamic aggregation modeling and uncertainty portrayal of distributed renewable energy sources; second

in order to fully tap the regulation potential of the load-side

a charging and discharging pricing strategy based on real-time tariffs and incentive mechanisms is proposed

and a stepwise dynamic carbon trading scheme is introduced; and lastly. Finally

a digital twin-driven state-space prediction model is established to map the real-time state of the physical system into the digital space

provide a high-fidelity digital basis for distributed model predictive control (DMPC)

and adopt the hierarchical coordination strategy of “day-ahead planning - intra-day rolling - real-time adjustment” to realize the dynamic optimization and control of the multi-energy virtual power plant. Simulation results show that the proposed strategy can effectively cope with the uncertainty and complexity of the multi-energy system

significantly improve the economy and stability of the virtual power plant operation

and provide technical support for the intelligent operation of the virtual power plant under the new type of power system.