Silikwa N. W.*, Medugu D. W., Danladi A.
Department of Physics, Adamawa State University, Mubi, Nigeria
*Corresponding author’s Email: Silikwawaida45@gmail.com, doi.org/10.55639/607.02010085
ABSTRACT
Wildfires pose recurring environmental and socio-economic challenges in savanna regions of sub-Saharan Africa, particularly during prolonged dry seasons. This study presents a simulation-driven proof of concept for a solar powered wireless sensor network (WSN) integrated with an artificial neural network (ANN) for wildfire risk classification. The proposed system combines low-power environmental sensing of temperature, relative humidity, and smoke concentration with a multilayer perceptron (MLP) classifier to evaluate the feasibility of embedded intelligence for wildfire early-warning applications. A synthetic but region-constrained dataset comprising 6,000 samples with balanced fire and non-fire classes was generated using climatological ranges reported for dry-season conditions in North-East Nigeria. Following data normalization and supervised learning, the ANN achieved a classification accuracy of 92.4%, precision of 91.1%, recall of 93.6%, and a receiver operating characteristic area under the curve (ROC–AUC) of 0.94 on a held out test set. An analytical energy assessment based on component-level current consumption estimated an average node power demand of approximately 148 mW, indicating feasibility for continuous low-duty-cycle operation under solar charging. The study does not claim field deployment or operational early detection performance; rather, it demonstrates the technical feasibility and energy viability of integrating ANN-based wildfire risk classification into a solar-powered WSN node. The results provide a structured baseline for future work involving sensor calibration, real-world data acquisition, and field validation in fire-prone savanna environments.
Keywords:
Wireless
sensor networks,
ANN,
Wildfire risk,
Environmental
monitoring.

