Wireless Sensor Network Pre-Deployment Algorithm for Reliable Forest Fire Detection

Forest fires are amongst the significant causes threatening our environment degradation nowadays. Current fire detection systems lack the capability of fire triggering detection from its early stages. Wireless sensor networks (WSN) as a solution letting to gather sensory data measurements, such as humidity and temperature, from all points of forest continuously, and, provide reliable and dependable data transmission to the control center. However, sensor networks face serious drawbacks such as its limited energy resources and the potential fault vulnerability due to cruel environmental conditions. Our framework includes proposals for a pre-deployment algorithm based Bayesian approach allowing determining the optimal parameters required to pre-deploy the WSN while achieving its fault detection. In our work, the sensor fault probability (the fact that the sensor itself can be defected) is introduced into the event detection process. For the Bayesian approach principle the optimal detection error resulting on both measurement error and sensor fault is exponentially decreased when the sensor neighborhood size increases. For a given detection error upper bound which resumes the fault tolerance in the WSN a minimum neighborhood size does exist. Through extensive simulations, we show that our proposed algorithm can provide a best choice of pre deployment parameters ensuring the WSN fault detection reliability while minimizing energy consumption. Keywords--Wireless Sensor Networks; Forest fire detection; Energy minimization; Bayesian approach; Sensor nodes predeployment; Reliable fault detection.