Intelligent wireless sensor nodes in water monitoring systems

A wireless sensor network (WSN) consists of spatially distributed autonomous nodes (sensors) to cooperatively monitor environmental or physical conditions. The considerable reduction in size and energy consumption of CMOS circuitry together with the self-organizational capability of network nodes are leading WSNs to be ubiquitous. WSNs operate under very strict energy constraints and the power supply is usually the most expensive component of the node being designed. This study proposes a methodology to reduce the power consumption by means of an energy-efficient management of node resources and tasks. We focus our attention on hydrometric level sensors. Support Vector Machines (SVMs) are used to predict the nonlinear dynamics of hydrometric level given a limited set of previous measurements. The procedure takes the error committed by the predictor and the battery charge status as input parameters. Based on the input state, fuzzy logic (FL) is used to infer the most appropriate sensor sampling rate and the need to transmit real-time data to the gateway node. The idea is that when battery charge is low and no flood events are occurring, then data transmission can be deferred and sampling rate can be decreased. The method efficiency has been tested in a real-world environment.