Multiple Input Single Output (MISO) ARX and NARX model of flood prediction system: A comparative study

Abstract

Flood is a most dangerous natural disaster that can cause enormous threat to human life and property. Thus, an accurate flood water level prediction is very prominent prior to develop a reliable flood water level prediction. Despite the widespread use of nonlinear model for flood water level prediction, linear model is still a new model among researchers around the world. Therefore, this paper present MISO linear system identification model namely Autoregressive Exogenous Input (ARX). The river branch treated in this study was Kelang river, located at Petaling bridge that originated from three upstream rivers which were Kelang river at Sulaiman bridge, Kelang river at Tun Perak bridge and Gombak river at Jalan Parlimen. The result obtained was not promising enough and there still rooms for improvement. This is due to the dynamics of flood water level itself is characterized as highly nonlinear. Thus, flood water level prediction using nonlinear model, Nonlinear Autoregressive Model with Exogenous Input (NARX) was proposed and results showed significant improvement.