WIRELESS FLOOD MONITORING USING INTEGRATED HYDROLOGICAL SENSORS AND FLOOD PREDICTION VIAARTIFICIAL NEURAL NETWORK

Flooding is a natural phenomenon that is very difficult to model into an equation because of its nonlinear characteristics. As a result, early warning flood prediction systems are seldom developed and often rely on meteorological satellites and hydrological maps. However, in the advent of technology, randomness and nonlinearity can now be modelled using artificial neural network. The goal of this study is to develop a wireless flood monitoring and prediction system using artificial neural network, specifically the Nonlinear Autoregressive Network with External Inputs (NARX) neural network that can be used in a small community as flood early warning system. The flood monitoring system was developed by integration of different hydrological sensors such as rain gauge, float sensor, flow meter, soil resistivity meter, air humidity and temperature sensors. The wireless communication was achieved by the use of Zigbee modules. Training of ANN was done via the backpropagation algorithm and an MSE of 0.0032 was achieved using seven epochs having the fourth epoch having the best validation. During the field testing, an average prediction rate accuracy of 98.65% was achieved. A two-sample t-test was done to see if the actual field test is different from the predicted values and the result was there is no significant difference between the two that validates the accuracy of the predictio