Compared to conventional synchronous generators, virtual synchronous generator (VSG) enjoys the advantage of flexible controllability. In particular, virtual inertia and virtual damping can have a substantial impact on the stability of VSG. RBF neural network, which enjoys simple algorithm, strong ability of learning and fast learning rate, has good approximation for continuous non-linear function and it can meet the needs of real-time control. Based on the characteristics of the control object, this paper improves the RBF neural network and designs a new adaptive control strategy which uses improved RBF neural network to adjust virtual inertia J of VSG. The neural network algorithm is integrated into the control object to establish an adaptive simulation model in Matlab, and the proposed control strategy is verified by simulation. The results show the adaptive control strategy can effectively improve frequency stability of VSG.