Reducing Power Consuption in Hydrometric Level Sensor Network Using Support Vector Machines

Environmental monitoring is a challeging task for both researchers and technical operators. Data loggers for ultrasonic hydrometric level sensors are compact devices equipped with microprocessor input channels and data storage. One of the critical issues that electronic engineers have to face in designing this kind of sensors is the energy consumption during the sensor startup phase preceding the level measurement. In this paper we propose a new methodology to reduce the power consumption by decreasing the sensor sampling rate when no flood events are occurring. This procedure allows the sampling rate to dynamically self-adapt based on the error between observed and predicted water level time-trend. Support Vector Machines are used to predict the hydrometric level given a limited number of previous samples. The method effectiveness has been tested on a real-world stage-discharge dataset.