3 Hours ahead of time flood water level prediction using NNARX structure: Case study pahang

Flood is defined as an overflow of large amount of water beyond its normal limits. Therefore, it has become threat to people's life and can cause damages to properties. However, in Malaysia, the only existing flood warning system are the alarming system which only notify residents nearby flood location to evacuate only when flood occur. Thus, flood water level prediction is very much needed in order to prevent flood disaster to happen. One of the effective techniques which frequently used to solve nonlinear cases such as flood is Artificial Neural Network (ANN). Therefore, this paper proposed 3 hours flood water level prediction using Neural Network Autoregressive model with Exogenous Input (NNARX) technique. The area involved in this study was along Pahang river basin where the flood location is situated at Mentakab. Four input parameters were fed in to the NNARX model to predict flood 3 hours ahead of time. The inputs were carefully selected during flood events. The samples used for training, validation and testing stage are 1553, 1997 and 4000 samples respectively. The NNARX flood prediction model developed using Matlab Neural Network Toolbox. Result shows satisfactory performance with low error measures.