Developing an Intelligent Reservoir Flood Control Decision Support System through Integrating Artificial Neural Networks

Abstract:

Typhoons and storms hit Taiwan several times every year and cause serious flood disasters. Because the mountainous terrain and steep landform rapidly accelerate the speed of flood flow, rivers cannot be a stable source of water supply. Reservoirs become one of the most important and effective floodwater storage facilities. However, real-time operation for reservoir flood control is a continuous and instant decision-making process based on rules, laws, meteorological nowcast, in addition to the immediate rainfall and hydrological data. The achievement of reservoir flood control can effectively mitigate flood disasters and store floodwaters for future uses. In this study, we construct an intelligent decision support system for reservoir flood control through integrating different types of neural networks and the above information to solve this problem.

This intelligent reservoir flood control decision support system includes three parts: typhoon track classification, flood forecast and adaptive water release models. This study used the self-organizing map (SOM) for typhoon track clustering, nonlinear autoregressive with exogenous inputs (NARX) for multi-step-ahead reservoir inflow prediction, and adaptive neuro-fuzzy inference system (ANFIS) for reservoir flood control. Before typhoons landfall, we can estimate the entire flood hydrogragh of reservoir inflow by using SOM and make a pre-release strategy and real-time reservoir flood operating by using ANFIS. In the meanwhile, NARX can be constantly used real-time five-hour-ahead inflow prediction for providing the newest flood information. The system has been successfully implemented Typhoons Trami (2013), Fitow (2013) and Matmo (2014) in Shihmen Reservoir.