ABSTRACT
Wireless Sensor Network (WSN) is gaining popularity day by day in a large area of applications. However, the operation of WSN is facing a multitude of challenges, mainly in terms of energy consumption since WSN nodes operate with battery power and changing the batteries is a complicated task, as networks may include hundreds to thousands of nodes. In this context, it is very crucial to know the remaining energy value in the battery of the sensor node to take required actions before losing sensor's function. Sending these measurements is very expensive in terms of energy and reduces the battery lifetime of the sensor and thus of the entire network. In this paper, we are interested in defining a probabilistic approach which aims to estimate these monitoring energy values and optimize energy consumption in WSN. Our approach is based on hidden Markov chains and includes two phases namely a learning phase and a prediction phase. Our approach is implemented as a web service. We illustrate our approach with a sensor-based health-care monitoring case study for COVID-19 patients. To evaluate our approach, we carry out experimentations based on the AvroraZ(a) simulator with a test for different types of applications and for different energy models: mu AMPS-specific model, Mica2-specific model, and Mica2-specific model with actual measurements. These experimentations demonstrate the accuracy and efficiency of our approach. Our results show that periodic WSN applications i.e. applications which send monitoring data periodically, tested with the mu AMPS-specific model perform an accuracy of 98.65%. In addition, our approach can perform a gain up to 75% of the battery charge of the sensor with an estimation of three-quarters of the remaining energy values.