Real-time multi-step ahead reservoir inflow forecasts by artificial neural networks
Journal
Taiwan Water Conservancy
Journal Volume
62
Journal Issue
2
Start Page
26
End Page
34
ISSN
04921505
Date Issued
2014-06
Author(s)
Abstract
Due to Taiwan's special geographical location, an average of 3.5 typhoons attack Taiwan each year. In addition, the particular topography of Taiwan makes rivers short and steep. Rivers flow rapidly during typhoon events, and it, in general, only takes a few hours for rivers to flow from catchments to reservoirs. It will be very helpful and useful for reservoir operation if reservoir inflow information in the next few hours after the arrival of typhoons can be provided. The artificial neural network (ANN) is one of the most prominent and novel methodologies for solving nonlinear problems in recent years. It is effective in data mining that extracts significant features from complex database and is capable of building any complex relationships between inputs and outputs. Therefore, the ANNs have been recognized useful for modeling dynamic nonlinear systems where the physical mechanism may not be precisely understood. This study aims to build forecast models for the Zengwen reservoir inflow using ANNs for investigating the effects of rain gauge precipitation data and radar rainfall data (QPESUMS: Quantitative Precipitation Estimation and Segregation Using Multiple Sensors) on the forecast accuracy of reservoir inflow. The models are evaluated based on three different input combinations: (1) flow information; (2) flow information and rain gauge precipitation information; (3) flow information and QPESUMS. The results indicate that the models with the use of precipitation information perform better than the model only using flow information; in addition, the models with QPESUMS information have the best result among the three input combinations. The study shows that QPESUMS information is contributed to the flow forecasting model.
Subjects
Artificial neural network
Quantitative precipitation estimation and segregation using multiple sensors (QPESUMS)
Zengwen reservoir inflow forecast
Publisher
Taiwan Joint Irrigation Associations
Type
journal article