This paper presents a new simulated annealing algorithm (SAA) to solve the long-term hydroscheduling problem. A new algorithm for randomly generating feasible trial solutions is introduced. The problem is a hard nonlinear optimization problem in continuous variables. An adaptive cooling schedule and a new method for variables discretization are implemented to enhance the speed and convergence of the original SAA. A significant reduction in the number of the objective function evaluations, and consequently less iteration are required to reach the optimal solution. The proposed algorithm has been applied successfully to solve a system with four series cascaded reservoirs. Numerical results show an improvement in the solutions compared to previously obtained results.