An approach based on artifical neural networks (ANNs) is proposed for the scheduling of hydroelectric generations. The purpose of hydroelectric generation scheduling is to figure out the optimal amounts of generated powers for the hydro units in the system for the next N (N = 24 in the work) hours in the future. Input data include system hourly loads and the natural inflow of each reservoir. In the proposed ANN approach, a clustering ANN is first developed to identify those days with similar hourly load patterns and natural inflows. These days with similar load patterns and inflows are said to be of the same group. A total of four groups are used in the work. Then a multilayer feedforward ANN is developed for each group to reach a preliminary generation schedule for the hydro units. Since some practical constraints may be violated in the preliminary schedule, a heuristic rule based search algorithm is developed to reach a feasible suboptimal schedule which satisfies all practical constraints. The effectiveness of the proposed approach is demonstrated by short-term hydro scheduling of Taiwan power system which consists of 10 hydro plants. It is concluded that the proposed approach is very effective in reaching proper hydro generation schedules. Moreover, the proposed approach is much faster than conventional dynamic programming approach.
