Abstract
This paper investigates the development and application of a daily optimization (DO) model to the short-term operation of the water supply system of Matsuyama, Japan. The main objective of the procedure is to seek the best allocations of water that minimizes the squared deviations between releases and target demands. The daily precipitation for one week ahead is assumed to be deterministic since meteorological short-range forecasts are generally available. The groundwater balance is modeled by Multiple Linear Regression (MLR) and short-term predictions of reservoir inflows are obtained based on the daily precipitations by means of Artificial Neural Networks (ANNs) . System operations using fictitious simulation and the DO model under perfect short-term forecast of inflows are used for comparison. The results of the DO procedure using ANN-based inflow predictions are shown to be equivalent to those obtained by DO under perfect short-term forecast of inflows and superior to the ones found by simulation.
