Self-learning for Day-night Mode Energy Strategy for Solar Powered Environmental WSN Nodes

Environmental data is in many cases acquired in remote locations that are difficult to access for sensor maintenance. Therefore, efficient use of available energy is crucial, particularly in systems that use energy harvesting devices, such as solar panels. This study presents a hybrid energy management strategy implemented in an environmental wireless sensor network (EWSN) controller. The control unit employs a model-free Q-learning algorithm during the day and linear energy discharging at night. A three-component Q-learning reward signal along with 7 actions and 11 energy states are designed for the system to achieve optimal performance in terms of data sensing and transmission operation and to minimize the amount of failures due energy storage depletion.