OPTIMIZATION OF HYDROPOWER RESERVOIR OPERATION USING ARTIFICIAL NEURAL NETWORKS

After development of any optimization model a post-optimization simulation is needed for two purposes: 1) Checking and evaluating of system performance and 2) Computing performance criteria. The common rule is developing a simulation model in the form of a computer code. In this paper, using stochastic dynamic programming, the optimum operating rule of a hydropower system is derived. Then capability of artificial neural networks (ANNs) has assessed as a substitution of simulation model. The optimization model has divided to 50 classes of discrete storage volumes and 8 classes of inflows in each time period. Derived optimum rules then have applied for ANN model training. Then the optimum releases of a 43 year historical record has determined with simulation model and ANN model, as model testing. The results show that the optimum releases of reservoir in both ANN and simulation models are very close together.