Flood Disaster Warning System on the go

Abstract

Floods are one of the top natural disaster that affects many regions around the world, harming human lives and lessening economy growth. Therefore, it is crucial to build an early warning system that forecast flow rate and water level to reduce the casualties of flood disaster. The objective of this paper is to design a flood monitoring system which integrates both flow and water level sensor and use two class neural network to predict the flood status from stored data in the database. A laboratory experiment was carried out to simulate the system and a pressure gauge was utilized to measure the pressure of inflowing water. A NodeMCU ESP8266 enables transmission of sensor data to Thingspeak channel for real-time visualization and storing the data in database. Furthermore, two class neural network module built in Microsoft's Azure Machine Learning (AzureML) was used to predict flood status according to a pre-define rule. The result of the 2-class neural network showed that using 3 hidden layers has the highest accuracy of 98.9% and precision of 100%.