EVALUATION OF REAL-TIME FLOOD MONITORING AND EARLY WARNING SYSTEM USING ARTIFICIAL NEURAL NETWORK

Authors

  • Dimson I.C. Dept. of Electronic and Computer Engineering, Nnamdi Azikiwe University, Awka, Nigeria.
  • Obianyo O.R. Dept. of Computer Engineering, Madonna University of Nigeria, Akpugo Campus, Enugu State, Nigeria.
  • Onyeyili T.C. Dept. of Electronic and Computer Engineering, Nnamdi Azikiwe University, Awka, Nigeria

Keywords:

Flooding, Artificial intelligence, Artificial Neural Network, Simulink.

Abstract

This research paper analyzes real-time flood detection and monitoring using Artificial Intelligence. The research was motivated by the need to develop an intelligent system that can pre-warn the inhabitants of flood prone localities on the likelihood of flooding at a specific time to initiate a process of safeguarding their lives, properties and evacuation to a safe location on time. A computer-aided software engineering methodological approach was deployed in achieving the aim and objectives of this work through characterization of a PID based conventional flood detection and monitoring system to establish the parameters that improve flood detection and monitoring system. Using the established parameters, a nonlinear model of Ugwuaji River in Enugu State (Nigeria) was developed. The sensing device that is meant to acquire real-time data from the environment was designed using pressure as the sensing element. A nonlinear model predictive control system that utilizes previous process control behaviour to foretell the future response of the system was modeled and implemented in Simulink. The developed mathematical models were transformed into a discrete form using Laplace transform to establish the transfer functions for the development of the Simulink model for real-time simulation. The model predictive control system network was trained offline using BFGS quasi-Newton Back Propagation Algorithm during simulation. Simulation results show that the proposed system achieved regression of 1 after several iterations. Results also show that the proposed system responded very fast to flood signal within 8.44s seconds as against 22 seconds achieved by the conventional PID sensor. The percentage increase in the new system performance is 21.6%. The comparative delay time between detection and prediction is 24.6s for the characterized sensor and 2.84 for the new sensor, the percentage improvement in the delay performance is therefore 88.4%.

Downloads

Published

2024-10-22 — Updated on 2024-10-22

Versions