An Iot-Enabled Resilient Wireless Sensor Network Framework For Early Forest Fire Detection And Monitoring

Abstract

Forest fires are a serious threat to natural ecosystems, wildlife, human settlements and economic resources and therefore early and accurate detection is essential for timely response to emergencies. Current forest fire detection monitoring systems based on watching towers, satellite images, cameras and traditional wireless sensor networks (WSN) have problems due to fire detection delay, false alarms, limited coverage, high communication overhead, and limited energy for sensor nodes. In an effort to overcome these problems, this paper proposes an Adaptive Fire Risk and Energy-Aware Detection-WSN Framework (AFRED-W) for continuous forest monitoring and early fire identification. The four environmental parameters measured are: temperature, relative humidity, light intensity and concentration of carbon monoxide (CO) by distributed IoT sensor nodes. The acquired measurements are normalized and processed with an Adaptive Fire Risk Index; which classifies the acquired data into low, moderate and high fire-risk levels. The event transmission is regulated by risk, to lower the unnecessary energy consumption. Suspected fire events are then assessed with a weighted machine learning ensemble of Support Vector Machine (SVM), Random Forest (RF) and XGBoost (XGB); sensor localization determines the affected region. From the experimental result, the proposed detection approach is obtained classification accuracy of 98.21% and F1 score of 98.15%. The system level fire detection rate achieved by the proposed method is 98.40% with a false alarm rate of 1.90%. The four techniques of adaptive multi-sensor risk assessment, energy-aware communication, localization, and ensemble learning offer a good framework for forest fire timely detection, energy reduction, and reliable emergency notification in the case of WSNs.