CASCADE HYDROPOWER DISCHARGE FLOW PREDICTION BASED ON DYNAMIC ARTIFICIAL NEURAL NETWORKS

Rainy seasons with heavy rainfall in catchment zones cause high potential of flooding at downstream, primarily due to the reservoirs' capacity limit been surpassed. Discharge flow prediction can be used for the hydropower plant to limit downstream flow during rainy seasons. In this study, discharge flow prediction based on the Artificial Neural Network (ANN) is proposed in order to forecast hydropower discharges flow. A cascade hydropower scheme has been selected for this study. Data such as fore-bay elevation, inflow, and discharge flow from the cascade hydropower power plants have been collected and used as an input for the ANN models. The developed models are Feedforward Backpropagation Neural Network, Elman Neural Network, and Nonlinear Autoregressive with Exogenous Inputs (NARX). The models have been assessed with different training methods and the number of hidden neurons to assess their performances. Moreover, the models' flow prediction performances been compared to the conventional Water Balance methodology. The result shows Elman Neural Network demonstrates higher prediction accuracy compared to other techniques based on the statistical error measures.