This paper presents a unit commitment algorithm which combines the Lagrangian Relaxation (LR), Sequential Unit Commitment (SUC), and Optimal Unit Decommitment (UD) methods to solve a general Hydro-Thermal Optimization (HTO) problem. We argue that this approach retains the advantages of the LR method while addressing the methods observed weaknesses to improve overall algorithm performance and quality of solution. The proposed approach has been implemented in a version of PG&E's HTO program, and test results are presented.