With the rapid development of new energy generation technology， renewable energy sources such as wind energy and photovoltaic not only serve as important active power sources， but their reactive power regulation potentials are also receiving increasing attention. In this paper， an innovative optimization strategy based on the improved Genghis Khan shark optimization （GKSO） algorithm is proposed to address the shortage of virtual power plant （VPP） reactive power sources and the model solving difficulties under high percentage of new energy access. First， a reactive power co-regulation model containing multiple distributed power sources such as wind power， photovoltaic， energy storage and gas turbine is constructed， and the key influencing factors of the uncertainty of new energy reactive power output are revealed through parameter sensitivity analysis. In order to accurately characterize the uncertainty， Latin hypercube sampling （LHS） combined with the scenario generation and reduction technique of Kantorovich distance is innovatively adopted to establish a typical set of scenarios of wind and solar power output. On this basis， a multi-objective optimization model of VPP considering the uncertainty of new energy reactive power is established and efficiently solved using the improved GKSO algorithm. The simulation results show that compared with the particle swarm optimization （PSO） algorithm and seagull optimization algorithm （SOA）， the optimized GKSO algorithm has a significant advantage in solving the VPP reactive power optimization problem， and it is necessary to take the new energy reactive power uncertainty into account in order to reduce the operational risk for large new energy stations with large installed capacity.