FORECASTING THE FLOOD STAGE OF A RESERVOIR BASED ON THE CHANGES IN UPSTREAM RAINFALL PATTERN

Abstract

Flood is among the major disasters in Malaysia. Flood occurs when the existing waterways are unable to support large amounts of water during heavy rain seasons. Reservoirs have been used as one of the flood mitigation approaches in the country. A reservoir can hold excessive water to ensure water flow to the downstream area is under the safe capacity of the waterway. However, due to the needs of the society, a reservoir also serves other purposes such as water supply and recreation. Therefore, reservoir water storage should be maintained to satisfy water usage, and at the same time, the water needs to be released to reserve space for incoming water. This conflict causes problems to reservoir operators when making the water release decision. In this paper, a forecasting model was proposed to forecast the flood stage of a reservoir based on the upstream rainfall pattern. This model could be used by reservoir operators in the early decision-making stage of releasing water before the reservoir reaches its maximum capacity. Simultaneously, the reservoir water level could be maintained for other uses. In this study, the experiments conducted proved that an Artificial Neural Network is capable of producing an acceptable performance in terms of its accuracy.