Flood modelling using Artificial Neural Network

Reliable flood water level prediction is very important in order to achieve good flood prediction system. Flood is a natural disaster that can cause loss in life and property. This paper proposed ANN modeling for flood water level prediction for early warning system using BPNN with NN Inverse Model placed at the output for performance improvement. The Back Propagation algorithm was applied based on dataset obtained from the Department of Irrigation and Drainage Malaysia. The algorithm seeks to minimize the value of error function based on the complexity and performance of the Artificial Neural Network. This is done by adjusting the model parameters values to obtain optimal results. The training inputs used in the algorithm were current values of flood water levels at three upstream river locations. The result produced poor prediction performance. Thus, a NN Inverse Model was proposed to be placed at the output of the BPNN. Significant improvement in performance was observed.