Large-scale renewable generation bases transmitted via voltage source converter (VSC)-HVDC system (RGBTVS) have emerged as a key approach for renewable power transmission in China's desert regions. The controller parameters of the VSCs significantly affect the small-signal stability (SSS) of RGBTVS. In this paper, a robust controller parameter optimization (RO) model for improving the SSS of a RGBTVS considering the injected power uncertainty is established. A support vector machine regression surrogate model is employed to approximate the highly nonlinear relationship of minimum damping ratio (MDR) and both controller parameters and renewable energy station (RES) output, thereby eliminating the need for matrix inversion and eigenvalue calculation in the SSS constraints. The RO problem is decomposed into two single-layer optimization problems which are solved iteratively using column-and-constraint generation (C&CG) algorithm. The master problem focuses on optimizing controller parameters to improve the system SSS under the worst scenario of RES output, while the sub-problem searches the RES output corresponding to the worst MDR under the specific controller parameters within the uncertainty set. A case study based on a practical RGBTVS in China verifies the correctness and effectiveness of the proposed method.