The short-term unit commitment (UC) problem of hydrothermal generation systems is a mixed-integer nonlinear programming (MINLP), which is difficult to solve efficiently, especially for large-scale instances. The perspective relaxation (PR) is an effective approach to constructing tight approximations to MINLP with semi-continuous variables. In this paper, the PR of UC problem is formulated as a mixed integer second-order cone programming (SOCP) model because the quadratic polynomial cost function of the UC problem is SOCP-representable. The proposed model is implemented by using the commercial optimization software IBM CPLEX 12.4. Extensive numerical studies have been conducted to verify the advantages of our proposed method. Instances of the test system vary from 10 to 1000 units. Our results indicate that the proposed method performs better than the existing methods in terms of production cost savings and faster computational times, especially for large systems. (C) 2013 Elsevier Ltd. All rights reserved.