This paper presents and compares two systematic and effective approaches for the hydroelectric generation scheduling problem (HSP). The objective of the HSP is to determine the best substitution of hydro energy for thermal energy in electric generation so that the fuel cost is minimized while meeting the system load and constraints. In the first approach, HSP is formulated as an optimal control problem, for which a multiplier method‐based differential dynamic programming algorithm developed by Chen et al. [4] is adopted as the solution algorithm. The second approach exploits the network structure of the water balance relation and takes linear approximation of the first formulation to formulate HSP as a minimum cost, linear network flow problem. Bertsekas and Tsengs’ RELAX code is applied to solve such a large‐scale network optimization problem. Numerical results indicate that both approaches generate results close to TPC's empirical schedules, are efficient in computation and suggest predictive operation strategies. We address the differences between the two approaches in solution optimality and computational efficiency and provide necessary information for tradeoffs in selecting a suitable approach.