INFLOW PREDICTION FOR DECISION SUPPORT SYSTEM OF RESERVOIR OPERATION

An expert system, to assist dam managers for five dams along the Saikawa River, has been developed with a primary objective of achieving swift and accurate reservoir operation decision-makings during floods. The expert system is capable of supporting on decision-makings upon establishment of flood management procedure and release/storage planning. Furthermore, an attempt was made to improve reservoir inflow prediction models for better supporting capability. As a result, accuracy on prediction of inflow up to 7 hours ahead was improved, which is important for flood management of the five dams, using neural network. The neural network inflow prediction models were developed for each types of floods caused by frontal rainfalls, snowmelt and typhoons, after extracting relevant meteorological factors for each.