The reliability of the ultra-high voltage direct current (UHVDC) lines with large capacity is important to the power system security. The existing studies about the UHVDC reliability is for long-term purpose, and can not show the short-term change of the reliability after component failures. In this paper, based on the detailed structure of the converter station of UHVDC, instantaneous state probability and steady-state aggregation algorithm, a transient hierarchical equivalence algorithm is proposed to derive the transient state space model of the converter station. The time-dependent reliability indices for the bipolar availability and bipolar outage rate are proposed. Based on the average probability algorithm, the average state probability and reliability of the converter station with a period of time are found. The proposed models extend the reliability model and indices of the UHVDC system, which is beneficial to evaluate its short-term risk. Compared with the transient state probability algorithm, the validity of the proposed transient hierarchical equivalent algorithm is verified. The results show that the transient state probability of UHVDC converter station is different from its steady-state value, and converges from different initial states to the steady-state value.