Real time inflow prediction in the Paute River Basin

Hydropower generation, which depends on reservoir inflows, is generally the cheapest power within a country’s generation facilities. To apply for electricity production quotas an accurate forecast of those inflows is essential. Artificial Neural Network (ANN) models have become a popular choice among the non-linear flow forecasting methods, because of its ability to learn from examples. This paper applies ANN to forecast the river flow at the Amaluza reservoir in the Paute river basin for one lead day, based on river flow and rainfall measured at various upstream stations and the Amaluza inflow. The first approach uses measurements from the online stations available in the river basin. As to reduce model complexity, based on the correlation analysis of the measurements at the online stations the data of some stations were eliminated. The research revealed that the existing station network does not cover adequately rainfall in the catchment. Therefore some additional non-online stations were included in the model, enabling a better description of the meteorological regimes in the basin. In doing so, the model produces adequate forecasts of the Amaluza reservoir inflow.status: publishe