An Energy-Efficient River Water Pollution Monitoring System in Internet of Things

An important research issue in river water pollution monitoring is to correctly estimate and transfer the pollution data from a river to the base station by consuming minimum energy. In this paper, we propose an energy-efficient river water pollution monitoring system by using deep neural network and long-range communication technology. Firstly, we design a compressed deep neural network for monitoring the river water pollution. Next, we use a knowledge distillation technique to train the compressed deep neural network. The compressed network can be successfully deployed on a limited resources Internet of Things device and achieves an acceptable accuracy. Further, we propose a game theory-based approach to estimate the time duration for using the suitable spreading factor of the long-range network to transmit the river water data. Such game theory-based approach helps in reducing energy consumption and ensures the successful transmission of the data to the base station. Finally, we present experimental and real-world evaluations that demonstrate the effectiveness of the propose system.